Islem Rekik

Karim Lekadir, Aasa Feragen, Abdul Joseph Fofanah et al. · eth-zurich

Despite major advances in artificial intelligence (AI) for medicine and healthcare, the deployment and adoption of AI technologies remain limited in real-world clinical practice. In recent years, concerns have been raised about the technical, clinical, ethical and legal risks associated with medical AI. To increase real world adoption, it is essential that medical AI tools are trusted and accepted by patients, clinicians, health organisations and authorities. This work describes the FUTURE-AI guideline as the first international consensus framework for guiding the development and deployment of trustworthy AI tools in healthcare. The FUTURE-AI consortium was founded in 2021 and currently comprises 118 inter-disciplinary experts from 51 countries representing all continents, including AI scientists, clinicians, ethicists, and social scientists. Over a two-year period, the consortium defined guiding principles and best practices for trustworthy AI through an iterative process comprising an in-depth literature review, a modified Delphi survey, and online consensus meetings. The FUTURE-AI framework was established based on 6 guiding principles for trustworthy AI in healthcare, i.e. Fairness, Universality, Traceability, Usability, Robustness and Explainability. Through consensus, a set of 28 best practices were defined, addressing technical, clinical, legal and socio-ethical dimensions. The recommendations cover the entire lifecycle of medical AI, from design, development and validation to regulation, deployment, and monitoring. FUTURE-AI is a risk-informed, assumption-free guideline which provides a structured approach for constructing medical AI tools that will be trusted, deployed and adopted in real-world practice. Researchers are encouraged to take the recommendations into account in proof-of-concept stages to facilitate future translation towards clinical practice of medical AI.

Imen Jegham, Islem Rekik

Decrypting intelligence from the human brain construct is vital in the detection of particular neurological disorders. Recently, functional brain connectomes have been used successfully to predict behavioral scores. However, state-of-the-art methods, on one hand, neglect the topological properties of the connectomes and, on the other hand, fail to solve the high inter-subject brain heterogeneity. To address these limitations, we propose a novel regression graph neural network through meta-learning namely Meta-RegGNN for predicting behavioral scores from brain connectomes. The parameters of our proposed regression GNN are explicitly trained so that a small number of gradient steps combined with a small training data amount produces a good generalization to unseen brain connectomes. Our results on verbal and full-scale intelligence quotient (IQ) prediction outperform existing methods in both neurotypical and autism spectrum disorder cohorts. Furthermore, we show that our proposed approach ensures generalizability, particularly for autistic subjects. Our Meta-RegGNN source code is available at https://github.com/basiralab/Meta-RegGNN.

Mehmet Yigit Balik, Arwa Rekik, Islem Rekik

Graph neural networks (GNNs) have achieved extraordinary enhancements in various areas including the fields medical imaging and network neuroscience where they displayed a high accuracy in diagnosing challenging neurological disorders such as autism. In the face of medical data scarcity and high-privacy, training such data-hungry models remains challenging. Federated learning brings an efficient solution to this issue by allowing to train models on multiple datasets, collected independently by different hospitals, in fully data-preserving manner. Although both state-of-the-art GNNs and federated learning techniques focus on boosting classification accuracy, they overlook a critical unsolved problem: investigating the reproducibility of the most discriminative biomarkers (i.e., features) selected by the GNN models within a federated learning paradigm. Quantifying the reproducibility of a predictive medical model against perturbations of training and testing data distributions presents one of the biggest hurdles to overcome in developing translational clinical applications. To the best of our knowledge, this presents the first work investigating the reproducibility of federated GNN models with application to classifying medical imaging and brain connectivity datasets. We evaluated our framework using various GNN models trained on medical imaging and connectomic datasets. More importantly, we showed that federated learning boosts both the accuracy and reproducibility of GNN models in such medical learning tasks. Our source code is available at https://github.com/basiralab/reproducibleFedGNN.

Furkan Pala, Islem Rekik

A central challenge in training one-shot learning models is the limited representativeness of the available shots of the data space. Particularly in the field of network neuroscience where the brain is represented as a graph, such models may lead to low performance when classifying brain states (e.g., typical vs. autistic). To cope with this, most of the existing works involve a data augmentation step to increase the size of the training set, its diversity and representativeness. Though effective, such augmentation methods are limited to generating samples with the same size as the input shots (e.g., generating brain connectivity matrices from a single shot matrix). To the best of our knowledge, the problem of generating brain multigraphs capturing multiple types of connectivity between pairs of nodes (i.e., anatomical regions) from a single brain graph remains unsolved. In this paper, we unprecedentedly propose a hybrid graph neural network (GNN) architecture, namely Multigraph Generator Network or briefly MultigraphGNet, comprising two subnetworks: (1) a many-to-one GNN which integrates an input population of brain multigraphs into a single template graph, namely a connectional brain temple (CBT), and (2) a reverse one-to-many U-Net network which takes the learned CBT in each training step and outputs the reconstructed input multigraph population. Both networks are trained in an end-to-end way using a cyclic loss. Experimental results demonstrate that our MultigraphGNet boosts the performance of an independent classifier when trained on the augmented brain multigraphs in comparison with training on a single CBT from each class. We hope that our framework can shed some light on the future research of multigraph augmentation from a single graph. Our MultigraphGNet source code is available at https://github.com/basiralab/MultigraphGNet.

Bobby Azad, Reza Azad, Sania Eskandari et al.

Foundation models, large-scale, pre-trained deep-learning models adapted to a wide range of downstream tasks have gained significant interest lately in various deep-learning problems undergoing a paradigm shift with the rise of these models. Trained on large-scale dataset to bridge the gap between different modalities, foundation models facilitate contextual reasoning, generalization, and prompt capabilities at test time. The predictions of these models can be adjusted for new tasks by augmenting the model input with task-specific hints called prompts without requiring extensive labeled data and retraining. Capitalizing on the advances in computer vision, medical imaging has also marked a growing interest in these models. To assist researchers in navigating this direction, this survey intends to provide a comprehensive overview of foundation models in the domain of medical imaging. Specifically, we initiate our exploration by providing an exposition of the fundamental concepts forming the basis of foundation models. Subsequently, we offer a methodical taxonomy of foundation models within the medical domain, proposing a classification system primarily structured around training strategies, while also incorporating additional facets such as application domains, imaging modalities, specific organs of interest, and the algorithms integral to these models. Furthermore, we emphasize the practical use case of some selected approaches and then discuss the opportunities, applications, and future directions of these large-scale pre-trained models, for analyzing medical images. In the same vein, we address the prevailing challenges and research pathways associated with foundational models in medical imaging. These encompass the areas of interpretability, data management, computational requirements, and the nuanced issue of contextual comprehension.

Atefe Hassani, Islem Rekik

A fundamental challenge in federated learning lies in mixing heterogeneous datasets and classification tasks while minimizing the high communication cost caused by clients as well as the exchange of weight updates with the server over a fixed number of rounds. This results in divergent model convergence rates and performance, which may hinder their deployment in precision medicine. In real-world scenarios, client data is collected from different hospitals with extremely varying components (e.g., imaging modality, organ type, etc). Previous studies often overlooked the convoluted heterogeneity during the training stage where the target learning tasks vary across clients as well as the dataset type and their distributions. To address such limitations, we unprecedentedly introduce UniFed, a universal federated learning paradigm that aims to classify any disease from any imaging modality. UniFed also handles the issue of varying convergence times in the client-specific optimization based on the complexity of their learning tasks. Specifically, by dynamically adjusting both local and global models, UniFed considers the varying task complexities of clients and the server, enhancing its adaptability to real-world scenarios, thereby mitigating issues related to overtraining and excessive communication. Furthermore, our framework incorporates a sequential model transfer mechanism that takes into account the diverse tasks among hospitals and a dynamic task-complexity based ordering. We demonstrate the superiority of our framework in terms of accuracy, communication cost, and convergence time over relevant benchmarks in diagnosing retina, histopathology, and liver tumour diseases under federated learning. Our UniFed code is available at https://github.com/basiralab/UniFed.

Xuesong Wang, Lina Yao, Islem Rekik et al.

Contrastive self-supervised learning has recently benefited fMRI classification with inductive biases. Its weak label reliance prevents overfitting on small medical datasets and tackles the high intraclass variances. Nonetheless, existing contrastive methods generate resemblant pairs only on pixel-level features of 3D medical images, while the functional connectivity that reveals critical cognitive information is under-explored. Additionally, existing methods predict labels on individual contrastive representation without recognizing neighbouring information in the patient group, whereas interpatient contrast can act as a similarity measure suitable for population-based classification. We hereby proposed contrastive functional connectivity graph learning for population-based fMRI classification. Representations on the functional connectivity graphs are "repelled" for heterogeneous patient pairs meanwhile homogeneous pairs "attract" each other. Then a dynamic population graph that strengthens the connections between similar patients is updated for classification. Experiments on a multi-site dataset ADHD200 validate the superiority of the proposed method on various metrics. We initially visualize the population relationships and exploit potential subtypes.

Nada Chaari, Hatice Camgoz Akdag, Islem Rekik

One of the greatest scientific challenges in network neuroscience is to create a representative map of a population of heterogeneous brain networks, which acts as a connectional fingerprint. The connectional brain template (CBT), also named network atlas, presents a powerful tool for capturing the most representative and discriminative traits of a given population while preserving its topological patterns. The idea of a CBT is to integrate a population of heterogeneous brain connectivity networks, derived from different neuroimaging modalities or brain views (e.g., structural and functional), into a unified holistic representation. Here we review current state-of-the-art methods designed to estimate well-centered and representative CBT for populations of single-view and multi-view brain networks. We start by reviewing each CBT learning method, then we introduce the evaluation measures to compare CBT representativeness of populations generated by single-view and multigraph integration methods, separately, based on the following criteria: centeredness, biomarker-reproducibility, node-level similarity, global-level similarity, and distance-based similarity. We demonstrate that the deep graph normalizer (DGN) method significantly outperforms other multi-graph and all single-view integration methods for estimating CBTs using a variety of healthy and disordered datasets in terms of centeredness, reproducibility (i.e., graph-derived biomarkers reproducibility that disentangle the typical from the atypical connectivity variability), and preserving the topological traits at both local and global graph-levels.

Ece Cinar, Sinem Elif Haseki, Alaa Bessadok et al.

The connectional brain template (CBT) captures the shared traits across all individuals of a given population of brain connectomes, thereby acting as a fingerprint. Estimating a CBT from a population where brain graphs are derived from diverse neuroimaging modalities (e.g., functional and structural) and at different resolutions (i.e., number of nodes) remains a formidable challenge to solve. Such network integration task allows for learning a rich and universal representation of the brain connectivity across varying modalities and resolutions. The resulting CBT can be substantially used to generate entirely new multimodal brain connectomes, which can boost the learning of the downs-stream tasks such as brain state classification. Here, we propose the Multimodal Multiresolution Brain Graph Integrator Network (i.e., M2GraphIntegrator), the first multimodal multiresolution graph integration framework that maps a given connectomic population into a well centered CBT. M2GraphIntegrator first unifies brain graph resolutions by utilizing resolution-specific graph autoencoders. Next, it integrates the resulting fixed-size brain graphs into a universal CBT lying at the center of its population. To preserve the population diversity, we further design a novel clustering-based training sample selection strategy which leverages the most heterogeneous training samples. To ensure the biological soundness of the learned CBT, we propose a topological loss that minimizes the topological gap between the ground-truth brain graphs and the learned CBT. Our experiments show that from a single CBT, one can generate realistic connectomic datasets including brain graphs of varying resolutions and modalities. We further demonstrate that our framework significantly outperforms benchmarks in reconstruction quality, augmentation task, centeredness and topological soundness.

Islem Rekik

Perceptions of intelligence shape how learners evaluate and rely on artificial intelligence (AI) systems. Despite rapid advances in AI capabilities, the impact of sustained exposure to these tools on students' valuation of human intelligence (HI) relative to AI remains underexplored. This paper presents a longitudinal analysis of classroom poll responses collected between 2020 and 2026 in AI-focused undergraduate and MSc courses in computer science. Data from 471 students across technical courses (such as Machine Learning and Deep Graph Learning) and design-oriented courses (such as Design Thinking for AI) reveal four recurring phases: hype, distrust, trust, and dependency. While early responses in 2020 slightly favored AI, a consistent shift toward HI emerged from 2024 onward across all MSc cohorts. By 2026, preference for HI reached 65 percent in a technical course (a 12 percentage-point increase from 2025) and 90 percent in a design-oriented course (a 36 percentage-point increase). These findings suggest a gradual reappraisal of human intelligence as AI becomes a routine tool, with implications for learner autonomy and epistemic agency. We conclude by reflecting on this cognitive shift from favoring artificial intelligence toward prioritizing human intelligence.

Oben Özgür, Arwa Rekik, Islem Rekik

The challenges of collecting medical data on neurological disorder diagnosis problems paved the way for learning methods with scarce number of samples. Due to this reason, one-shot learning still remains one of the most challenging and trending concepts of deep learning as it proposes to simulate the human-like learning approach in classification problems. Previous studies have focused on generating more accurate fingerprints of the population using graph neural networks (GNNs) with connectomic brain graph data. Thereby, generated population fingerprints named connectional brain template (CBTs) enabled detecting discriminative bio-markers of the population on classification tasks. However, the reverse problem of data augmentation from single graph data representing brain connectivity has never been tackled before. In this paper, we propose an augmentation pipeline in order to provide improved metrics on our binary classification problem. Divergently from the previous studies, we examine augmentation from a single population template by utilizing graph-based generative adversarial network (gGAN) architecture for a classification problem. We benchmarked our proposed solution on AD/LMCI dataset consisting of brain connectomes with Alzheimer's Disease (AD) and Late Mild Cognitive Impairment (LMCI). In order to evaluate our model's generalizability, we used cross-validation strategy and randomly sampled the folds multiple times. Our results on classification not only provided better accuracy when augmented data generated from one sample is introduced, but yields more balanced results on other metrics as well.

Doğa Türkseven, Islem Rekik, Christoph von Tycowicz et al.

Predicting the future development of an anatomical shape from a single baseline observation is a challenging task. But it can be essential for clinical decision-making. Research has shown that it should be tackled in curved shape spaces, as (e.g., disease-related) shape changes frequently expose nonlinear characteristics. We thus propose a novel prediction method that encodes the whole shape in a Riemannian shape space. It then learns a simple prediction technique founded on hierarchical statistical modeling of longitudinal training data. When applied to predict the future development of the shape of the right hippocampus under Alzheimer's disease and to human body motion, it outperforms deep learning-supported variants as well as state-of-the-art.

K. Mancini, I. Rekik

Graph Neural Networks (GNNs) have proven effective in various medical imaging applications, such as automated disease diagnosis. However, due to the local neighborhood aggregation paradigm in message passing which characterizes these models, they inherently suffer from two fundamental limitations: first, indistinguishable node embeddings due to heterophilic node aggregation (known as over-smoothing), and second, impaired message passing due to aggregation through graph bottlenecks (known as over-squashing). These challenges hinder the model expressiveness and prevent us from using deeper models to capture long-range node dependencies within the graph. Popular solutions in the literature are either too expensive to process large graphs due to high time complexity or do not generalize across all graph topologies. To address these limitations, we propose DuoGNN, a scalable and generalizable architecture which leverages topology to decouple homophilic and heterophilic edges and capture both short-range and long-range interactions. Our three core contributions introduce (i) a topological edge-filtering algorithm which extracts homophilic interactions and enables the model to generalize well for any graph topology, (ii) a heterophilic graph condensation technique which extracts heterophilic interactions and ensures scalability, and (iii) a dual homophilic and heterophilic aggregation pipeline which prevents over-smoothing and over-squashing during the message passing. We benchmark our model on medical and non-medical node classification datasets and compare it with its variants, showing consistent improvements across all tasks. Our DuoGNN code is available at https://github.com/basiralab/DuoGNN.

Kevin Mancini, Islem Rekik

Graph Neural Networks (GNNs) are popular deep learning models designed to process graph-structured data through recursive neighborhood aggregations in the message passing process. When applied to semi-supervised node classification, the message-passing enables GNNs to understand short-range spatial interactions, but also causes them to suffer from over-smoothing and over-squashing. These challenges hinder model expressiveness and prevent the use of deeper models to capture long-range node interactions (LRIs) within the graph. Popular solutions for LRIs detection are either too expensive to process large graphs due to high time complexity or fail to generalize across diverse graph structures. To address these limitations, we propose a mechanism called \emph{information flow control}, which leverages a novel connectivity measure, called \emph{information flow score}, to address over-smoothing and over-squashing with linear computational overhead, supported by theoretical evidence. Finally, to prove the efficacy of our methodology we design DeltaGNN, the first scalable and generalizable approach for detecting long-range and short-range interactions. We benchmark our model across 10 real-world datasets, including graphs with varying sizes, topologies, densities, and homophilic ratios, showing superior performance with limited computational complexity. The implementation of the proposed methods are publicly available at https://github.com/basiralab/DeltaGNN.

Prajit Sengupta, Islem Rekik

Medical image classification requires not only high predictive performance but also interpretability to ensure clinical trust and adoption. Graph Neural Networks (GNNs) offer a powerful framework for modeling relational structures within datasets; however, standard GNNs often operate as black boxes, limiting transparency and usability, particularly in clinical settings. In this work, we present an interpretable graph-based learning framework named FireGNN that integrates trainable fuzzy rules into GNNs for medical image classification. These rules embed topological descriptors - node degree, clustering coefficient, and label agreement - using learnable thresholds and sharpness parameters to enable intrinsic symbolic reasoning. Additionally, we explore auxiliary self-supervised tasks (e.g., homophily prediction, similarity entropy) as a benchmark to evaluate the contribution of topological learning. Our fuzzy-rule-enhanced model achieves strong performance across five MedMNIST benchmarks and the synthetic dataset MorphoMNIST, while also generating interpretable rule-based explanations. To our knowledge, this is the first integration of trainable fuzzy rules within a GNN. Source Code: https://github.com/basiralab/FireGNN

Ramona Ghilea, Islem Rekik

Learning from limited data has been extensively studied in machine learning, considering that deep neural networks achieve optimal performance when trained using a large amount of samples. Although various strategies have been proposed for centralized training, the topic of federated learning with small datasets remains largely unexplored. Moreover, in realistic scenarios, such as settings where medical institutions are involved, the number of participating clients is also constrained. In this work, we propose a novel federated learning framework, named RepTreeFL. At the core of the solution is the concept of a replica, where we replicate each participating client by copying its model architecture and perturbing its local data distribution. Our approach enables learning from limited data and a small number of clients by aggregating a larger number of models with diverse data distributions. Furthermore, we leverage the hierarchical structure of the client network (both original and virtual), alongside the model diversity across replicas, and introduce a diversity-based tree aggregation, where replicas are combined in a tree-like manner and the aggregation weights are dynamically updated based on the model discrepancy. We evaluated our method on two tasks and two types of data, graph generation and image classification (binary and multi-class), with both homogeneous and heterogeneous model architectures. Experimental results demonstrate the effectiveness and outperformance of RepTreeFL in settings where both data and clients are limited. Our code is available at https://github.com/basiralab/RepTreeFL.

Christopher Adnel, Islem Rekik

Graph Neural Network (GNN) ushered in a new era of machine learning with interconnected datasets. While traditional neural networks can only be trained on independent samples, GNN allows for the inclusion of inter-sample interactions in the training process. This gain, however, incurs additional memory cost, rendering most GNNs unscalable for real-world applications involving vast and complicated networks with tens of millions of nodes (e.g., social circles, web graphs, and brain graphs). This means that storing the graph in the main memory can be difficult, let alone training the GNN model with significantly less GPU memory. While much of the recent literature has focused on either mini-batching GNN methods or quantization, graph reduction methods remain largely scarce. Furthermore, present graph reduction approaches have several drawbacks. First, most graph reduction focuses only on the inference stage (e.g., condensation and distillation) and requires full graph GNN training, which does not reduce training memory footprint. Second, many methods focus solely on the graph's structural aspect, ignoring the initial population feature-label distribution, resulting in a skewed post-reduction label distribution. Here, we propose a Feature-Label COnstrained graph Net collapse, FALCON, to address these limitations. Our three core contributions lie in (i) designing FALCON, a topology-aware graph reduction technique that preserves feature-label distribution; (ii) implementation of FALCON with other memory reduction methods (i.e., mini-batched GNN and quantization) for further memory reduction; (iii) extensive benchmarking and ablation studies against SOTA methods to evaluate FALCON memory reduction. Our extensive results show that FALCON can significantly collapse various public datasets while achieving equal prediction quality across GNN models. Code: https://github.com/basiralab/FALCON

Umut Guvercin, Mohammed Amine Gharsallaoui, Islem Rekik

Few-shot learning presents a challenging paradigm for training discriminative models on a few training samples representing the target classes to discriminate. However, classification methods based on deep learning are ill-suited for such learning as they need large amounts of training data --let alone one-shot learning. Recently, graph neural networks (GNNs) have been introduced to the field of network neuroscience, where the brain connectivity is encoded in a graph. However, with scarce neuroimaging datasets particularly for rare diseases and low-resource clinical facilities, such data-devouring architectures might fail in learning the target task. In this paper, we take a very different approach in training GNNs, where we aim to learn with one sample and achieve the best performance --a formidable challenge to tackle. Specifically, we present the first one-shot paradigm where a GNN is trained on a single population-driven template --namely a connectional brain template (CBT). A CBT is a compact representation of a population of brain graphs capturing the unique connectivity patterns shared across individuals. It is analogous to brain image atlases for neuroimaging datasets. Using a one-representative CBT as a training sample, we alleviate the training load of GNN models while boosting their performance across a variety of classification and regression tasks. We demonstrate that our method significantly outperformed benchmark one-shot learning methods with downstream classification and time-dependent brain graph data forecasting tasks while competing with the train-on-all conventional training strategy. Our source code can be found at https://github.com/basiralab/one-representative-shot-learning.

Alaa Bessadok, Ahmed Nebli, Mohamed Ali Mahjoub et al.

Charting the baby connectome evolution trajectory during the first year after birth plays a vital role in understanding dynamic connectivity development of baby brains. Such analysis requires acquisition of longitudinal connectomic datasets. However, both neonatal and postnatal scans are rarely acquired due to various difficulties. A small body of works has focused on predicting baby brain evolution trajectory from a neonatal brain connectome derived from a single modality. Although promising, large training datasets are essential to boost model learning and to generalize to a multi-trajectory prediction from different modalities (i.e., functional and morphological connectomes). Here, we unprecedentedly explore the question: Can we design a few-shot learning-based framework for predicting brain graph trajectories across different modalities? To this aim, we propose a Graph Multi-Trajectory Evolution Network (GmTE-Net), which adopts a teacher-student paradigm where the teacher network learns on pure neonatal brain graphs and the student network learns on simulated brain graphs given a set of different timepoints. To the best of our knowledge, this is the first teacher-student architecture tailored for brain graph multi-trajectory growth prediction that is based on few-shot learning and generalized to graph neural networks (GNNs). To boost the performance of the student network, we introduce a local topology-aware distillation loss that forces the predicted graph topology of the student network to be consistent with the teacher network. Experimental results demonstrate substantial performance gains over benchmark methods. Hence, our GmTE-Net can be leveraged to predict atypical brain connectivity trajectory evolution across various modalities. Our code is available at https: //github.com/basiralab/GmTE-Net.

Mohammed Amine Gharsallaoui, Islem Rekik

Graph neural networks (GNNs) have witnessed an unprecedented proliferation in tackling several problems in computer vision, computer-aided diagnosis, and related fields. While prior studies have focused on boosting the model accuracy, quantifying the reproducibility of the most discriminative features identified by GNNs is still an intact problem that yields concerns about their reliability in clinical applications in particular. Specifically, the reproducibility of biological markers across clinical datasets and distribution shifts across classes (e.g., healthy and disordered brains) is of paramount importance in revealing the underpinning mechanisms of diseases as well as propelling the development of personalized treatment. Motivated by these issues, we propose, for the first time, reproducibility-based GNN selection (RG-Select), a framework for GNN reproducibility assessment via the quantification of the most discriminative features (i.e., biomarkers) shared between different models. To ascertain the soundness of our framework, the reproducibility assessment embraces variations of different factors such as training strategies and data perturbations. Despite these challenges, our framework successfully yielded replicable conclusions across different training strategies and various clinical datasets. Our findings could thus pave the way for the development of biomarker trustworthiness and reliability assessment methods for computer-aided diagnosis and prognosis tasks. RG-Select code is available on GitHub at https://github.com/basiralab/RG-Select.

Alaa Bessadok, Mohamed Ali Mahjoub, Islem Rekik

Noninvasive medical neuroimaging has yielded many discoveries about the brain connectivity. Several substantial techniques mapping morphological, structural and functional brain connectivities were developed to create a comprehensive road map of neuronal activities in the human brain -namely brain graph. Relying on its non-Euclidean data type, graph neural network (GNN) provides a clever way of learning the deep graph structure and it is rapidly becoming the state-of-the-art leading to enhanced performance in various network neuroscience tasks. Here we review current GNN-based methods, highlighting the ways that they have been used in several applications related to brain graphs such as missing brain graph synthesis and disease classification. We conclude by charting a path toward a better application of GNN models in network neuroscience field for neurological disorder diagnosis and population graph integration. The list of papers cited in our work is available at https://github.com/basiralab/GNNs-in-Network-Neuroscience.

Megi Isallari, Islem Rekik

Brain image analysis has advanced substantially in recent years with the proliferation of neuroimaging datasets acquired at different resolutions. While research on brain image super-resolution has undergone a rapid development in the recent years, brain graph super-resolution is still poorly investigated because of the complex nature of non-Euclidean graph data. In this paper, we propose the first-ever deep graph super-resolution (GSR) framework that attempts to automatically generate high-resolution (HR) brain graphs with N' nodes (i.e., anatomical regions of interest (ROIs)) from low-resolution (LR) graphs with N nodes where N < N'. First, we formalize our GSR problem as a node feature embedding learning task. Once the HR nodes' embeddings are learned, the pairwise connectivity strength between brain ROIs can be derived through an aggregation rule based on a novel Graph U-Net architecture. While typically the Graph U-Net is a node-focused architecture where graph embedding depends mainly on node attributes, we propose a graph-focused architecture where the node feature embedding is based on the graph topology. Second, inspired by graph spectral theory, we break the symmetry of the U-Net architecture by super-resolving the low-resolution brain graph structure and node content with a GSR layer and two graph convolutional network layers to further learn the node embeddings in the HR graph. Third, to handle the domain shift between the ground-truth and the predicted HR brain graphs, we incorporate adversarial regularization to align their respective distributions. Our proposed AGSR-Net framework outperformed its variants for predicting high-resolution functional brain graphs from low-resolution ones. Our AGSR-Net code is available on GitHub at https://github.com/basiralab/AGSR-Net.

Islem Rekik, Mustafa Burak Gurbuz

With the recent technological advances, biological datasets, often represented by networks (i.e., graphs) of interacting entities, proliferate with unprecedented complexity and heterogeneity. Although modern network science opens new frontiers of analyzing connectivity patterns in such datasets, we still lack data-driven methods for extracting an integral connectional fingerprint of a multi-view graph population, let alone disentangling the typical from the atypical variations across the population samples. We present the multi-view graph normalizer network (MGN-Net; https://github.com/basiralab/MGN-Net), a graph neural network based method to normalize and integrate a set of multi-view biological networks into a single connectional template that is centered, representative, and topologically sound. We demonstrate the use of MGN-Net by discovering the connectional fingerprints of healthy and neurologically disordered brain network populations including Alzheimer's disease and Autism spectrum disorder patients. Additionally, by comparing the learned templates of healthy and disordered populations, we show that MGN-Net significantly outperforms conventional network integration methods across extensive experiments in terms of producing the most centered templates, recapitulating unique traits of populations, and preserving the complex topology of biological networks. Our evaluations showed that MGN-Net is powerfully generic and easily adaptable in design to different graph-based problems such as identification of relevant connections, normalization and integration.

Ahmed Nebli, Ugur Ali Kaplan, Islem Rekik

Learning how to predict the brain connectome (i.e. graph) development and aging is of paramount importance for charting the future of within-disorder and cross-disorder landscape of brain dysconnectivity evolution. Indeed, predicting the longitudinal (i.e., time-dependent ) brain dysconnectivity as it emerges and evolves over time from a single timepoint can help design personalized treatments for disordered patients in a very early stage. Despite its significance, evolution models of the brain graph are largely overlooked in the literature. Here, we propose EvoGraphNet, the first end-to-end geometric deep learning-powered graph-generative adversarial network (gGAN) for predicting time-dependent brain graph evolution from a single timepoint. Our EvoGraphNet architecture cascades a set of time-dependent gGANs, where each gGAN communicates its predicted brain graphs at a particular timepoint to train the next gGAN in the cascade at follow-up timepoint. Therefore, we obtain each next predicted timepoint by setting the output of each generator as the input of its successor which enables us to predict a given number of timepoints using only one single timepoint in an end- to-end fashion. At each timepoint, to better align the distribution of the predicted brain graphs with that of the ground-truth graphs, we further integrate an auxiliary Kullback-Leibler divergence loss function. To capture time-dependency between two consecutive observations, we impose an l1 loss to minimize the sparse distance between two serialized brain graphs. A series of benchmarks against variants and ablated versions of our EvoGraphNet showed that we can achieve the lowest brain graph evolution prediction error using a single baseline timepoint. Our EvoGraphNet code is available at http://github.com/basiralab/EvoGraphNet.

Zeynep Gurler, Ahmed Nebli, Islem Rekik

Foreseeing the brain evolution as a complex highly inter-connected system, widely modeled as a graph, is crucial for mapping dynamic interactions between different anatomical regions of interest (ROIs) in health and disease. Interestingly, brain graph evolution models remain almost absent in the literature. Here we design an adversarial brain network normalizer for representing each brain network as a transformation of a fixed centered population-driven connectional template. Such graph normalization with respect to a fixed reference paves the way for reliably identifying the most similar training samples (i.e., brain graphs) to the testing sample at baseline timepoint. The testing evolution trajectory will be then spanned by the selected training graphs and their corresponding evolution trajectories. We base our prediction framework on geometric deep learning which naturally operates on graphs and nicely preserves their topological properties. Specifically, we propose the first graph-based Generative Adversarial Network (gGAN) that not only learns how to normalize brain graphs with respect to a fixed connectional brain template (CBT) (i.e., a brain template that selectively captures the most common features across a brain population) but also learns a high-order representation of the brain graphs also called embeddings. We use these embeddings to compute the similarity between training and testing subjects which allows us to pick the closest training subjects at baseline timepoint to predict the evolution of the testing brain graph over time. A series of benchmarks against several comparison methods showed that our proposed method achieved the lowest brain disease evolution prediction error using a single baseline timepoint. Our gGAN code is available at http://github.com/basiralab/gGAN.

Mert Lostar, Islem Rekik

-Background. Network neuroscience examines the brain as a complex system represented by a network (or connectome), providing deeper insights into the brain morphology and function, allowing the identification of atypical brain connectivity alterations, which can be used as diagnostic markers of neurological disorders. -Existing Methods. Graph embedding methods which map data samples (e.g., brain networks) into a low dimensional space have been widely used to explore the relationship between samples for classification or prediction tasks. However, the majority of these works are based on modeling the pair-wise relationships between samples, failing to capture their higher-order relationships. -New Method. In this paper, inspired by the nascent field of geometric deep learning, we propose Hypergraph U-Net (HUNet), a novel data embedding framework leveraging the hypergraph structure to learn low-dimensional embeddings of data samples while capturing their high-order relationships. Specifically, we generalize the U-Net architecture, naturally operating on graphs, to hypergraphs by improving local feature aggregation and preserving the high-order relationships present in the data. -Results. We tested our method on small-scale and large-scale heterogeneous brain connectomic datasets including morphological and functional brain networks of autistic and demented patients, respectively. -Conclusion. Our HUNet outperformed state-of-the-art geometric graph and hypergraph data embedding techniques with a gain of 4-14% in classification accuracy, demonstrating both scalability and generalizability. HUNet code is available at https://github.com/basiralab/HUNet.

Markus D. Schirmer, Archana Venkataraman, Islem Rekik et al.

Large, open-source consortium datasets have spurred the development of new and increasingly powerful machine learning approaches in brain connectomics. However, one key question remains: are we capturing biologically relevant and generalizable information about the brain, or are we simply overfitting to the data? To answer this, we organized a scientific challenge, the Connectomics in NeuroImaging Transfer Learning Challenge (CNI-TLC), held in conjunction with MICCAI 2019. CNI-TLC included two classification tasks: (1) diagnosis of Attention-Deficit/Hyperactivity Disorder (ADHD) within a pre-adolescent cohort; and (2) transference of the ADHD model to a related cohort of Autism Spectrum Disorder (ASD) patients with an ADHD comorbidity. In total, 240 resting-state fMRI time series averaged according to three standard parcellation atlases, along with clinical diagnosis, were released for training and validation (120 neurotypical controls and 120 ADHD). We also provided demographic information of age, sex, IQ, and handedness. A second set of 100 subjects (50 neurotypical controls, 25 ADHD, and 25 ASD with ADHD comorbidity) was used for testing. Models were submitted in a standardized format as Docker images through ChRIS, an open-source image analysis platform. Utilizing an inclusive approach, we ranked the methods based on 16 different metrics. The final rank was calculated using the rank product for each participant across all measures. Furthermore, we assessed the calibration curves of each method. Five participants submitted their model for evaluation, with one outperforming all other methods in both ADHD and ASD classification. However, further improvements are needed to reach the clinical translation of functional connectomics. We are keeping the CNI-TLC open as a publicly available resource for developing and validating new classification methodologies in the field of connectomics.

Pragya Singh, Islem Rekik

Graph super-resolution, the task of inferring high-resolution (HR) graphs from low-resolution (LR) counterparts, is an underexplored yet crucial research direction that circumvents the need for costly data acquisition. This makes it especially desirable for resource-constrained fields such as the medical domain. While recent GNN-based approaches show promise, they suffer from two key limitations: (1) matrix-based node super-resolution that disregards graph structure and lacks permutation invariance; and (2) reliance on node representations to infer edge weights, which limits scalability and expressivity. In this work, we propose two GNN-agnostic frameworks to address these issues. First, Bi-SR introduces a bipartite graph connecting LR and HR nodes to enable structure-aware node super-resolution that preserves topology and permutation invariance. Second, DEFEND learns edge representations by mapping HR edges to nodes of a dual graph, allowing edge inference via standard node-based GNNs. We evaluate both frameworks on a real-world brain connectome dataset, where they achieve state-of-the-art performance across seven topological measures. To support generalization, we introduce twelve new simulated datasets that capture diverse topologies and LR-HR relationships. These enable comprehensive benchmarking of graph super-resolution methods.

Yicheng Leng, Syed Muhammad Anwar, Islem Rekik et al.

Functional Magnetic Resonance Imaging (fMRI) provides useful insights into the brain function both during task or rest. Representing fMRI data using correlation matrices is found to be a reliable method of analyzing the inherent connectivity of the brain in the resting and active states. Graph Neural Networks (GNNs) have been widely used for brain network analysis due to their inherent explainability capability. In this work, we introduce a novel framework using contrastive self-supervised learning graph transformers, incorporating a brain network transformer encoder with random graph alterations. The proposed network leverages both contrastive learning and graph alterations to effectively train the graph transformer for autism detection. Our approach, tested on Autism Brain Imaging Data Exchange (ABIDE) data, demonstrates superior autism detection, achieving an AUROC of 82.6 and an accuracy of 74%, surpassing current state-of-the-art methods.

Pragya Singh, Islem Rekik

Brain graph super-resolution (SR) is an under-explored yet highly relevant task in network neuroscience. It circumvents the need for costly and time-consuming medical imaging data collection, preparation, and processing. Current SR methods leverage graph neural networks (GNNs) thanks to their ability to natively handle graph-structured datasets. However, most GNNs perform node feature learning, which presents two significant limitations: (1) they require computationally expensive methods to learn complex node features capable of inferring connectivity strength or edge features, which do not scale to larger graphs; and (2) computations in the node space fail to adequately capture higher-order brain topologies such as cliques and hubs. However, numerous studies have shown that brain graph topology is crucial in identifying the onset and presence of various neurodegenerative disorders like Alzheimer and Parkinson. Motivated by these challenges and applications, we propose our STP-GSR framework. It is the first graph SR architecture to perform representation learning in higher-order topological space. Specifically, using the primal-dual graph formulation from graph theory, we develop an efficient mapping from the edge space of our low-resolution (LR) brain graphs to the node space of a high-resolution (HR) dual graph. This approach ensures that node-level computations on this dual graph correspond naturally to edge-level learning on our HR brain graphs, thereby enforcing strong topological consistency within our framework. Additionally, our framework is GNN layer agnostic and can easily learn from smaller, scalable GNNs, reducing computational requirements. We comprehensively benchmark our framework across seven key topological measures and observe that it significantly outperforms the previous state-of-the-art methods and baselines.

Mayssa Soussia, Yijun Lin, Mohamed Ali Mahjoub et al.

Generative learning has advanced network neuroscience, enabling tasks like graph super-resolution, temporal graph prediction, and multimodal brain graph fusion. However, current methods, mainly based on graph neural networks (GNNs), focus solely on structural and topological properties, neglecting cognitive traits. To address this, we introduce the first cognified generative model, CogGNN, which endows GNNs with cognitive capabilities (e.g., visual memory) to generate brain networks that preserve cognitive features. While broadly applicable, we present CogGNN, a specific variant designed to integrate visual input, a key factor in brain functions like pattern recognition and memory recall. As a proof of concept, we use our model to learn connectional brain templates (CBTs), population-level fingerprints from multi-view brain networks. Unlike prior work that overlooks cognitive properties, CogGNN generates CBTs that are both cognitively and structurally meaningful. Our contributions are: (i) a novel cognition-aware generative model with a visual-memory-based loss; (ii) a CBT-learning framework with a co-optimization strategy to yield well-centered, discriminative, cognitively enhanced templates. Extensive experiments show that CogGNN outperforms state-of-the-art methods, establishing a strong foundation for cognitively grounded brain network modeling.

Geng Chen, Qingyue Wang, Islem Rekik

A connectional brain template (CBT) is a holistic representation of a population of multi-view brain connectivity graphs, encoding shared patterns and normalizing typical variations across individuals. The federation of CBT learning allows for an inclusive estimation of the representative center of multi-domain brain connectivity datasets in a fully data-preserving manner. However, existing methods overlook the non-independent and identically distributed (non-IDD) issue stemming from multidomain brain connectivity heterogeneity, in which data domains are drawn from different hospitals and imaging modalities. To overcome this limitation, we unprecedentedly propose a metadata-driven federated learning framework, called MetaFedCBT, for cross-domain CBT learning. Given the data drawn from a specific domain (i.e., hospital), our model aims to learn metadata in a fully supervised manner by introducing a local client-based regressor network. The generated meta-data is forced to meet the statistical attributes (e.g., mean) of other domains, while preserving their privacy. Our supervised meta-data generation approach boosts the unsupervised learning of a more centered, representative, and holistic CBT of a particular brain state across diverse domains. As the federated learning progresses over multiple rounds, the learned metadata and associated generated connectivities are continuously updated to better approximate the target domain information. MetaFedCBT overcomes the non-IID issue of existing methods by generating informative brain connectivities for privacy-preserving holistic CBT learning with guidance using metadata. Extensive experiments on multi-view morphological brain networks of normal and patient subjects demonstrate that our MetaFedCBT is a superior federated CBT learning model and significantly advances the state-of-the-art performance.

Kelei He, Chen Gan, Zhuoyuan Li et al.

Transformers have dominated the field of natural language processing, and recently impacted the computer vision area. In the field of medical image analysis, Transformers have also been successfully applied to full-stack clinical applications, including image synthesis/reconstruction, registration, segmentation, detection, and diagnosis. Our paper aims to promote awareness and application of Transformers in the field of medical image analysis. Specifically, we first overview the core concepts of the attention mechanism built into Transformers and other basic components. Second, we review various Transformer architectures tailored for medical image applications and discuss their limitations. Within this review, we investigate key challenges revolving around the use of Transformers in different learning paradigms, improving the model efficiency, and their coupling with other techniques. We hope this review can give a comprehensive picture of Transformers to the readers in the field of medical image analysis.

Oytun Demirbilek, Islem Rekik

Learning how to estimate a connectional brain template(CBT) from a population of brain multigraphs, where each graph (e.g., functional) quantifies a particular relationship between pairs of brain regions of interest (ROIs), allows to pin down the unique connectivity patterns shared across individuals. Specifically, a CBT is viewed as an integral representation of a set of highly heterogeneous graphs and ideally meeting the centeredness (i.e., minimum distance to all graphs in the population) and discriminativeness (i.e., distinguishes the healthy from the disordered population) criteria. So far, existing works have been limited to only integrating and fusing a population of brain multigraphs acquired at a single timepoint. In this paper, we unprecedentedly tackle the question: Given a baseline multigraph population, can we learn how to integrate and forecast its CBT representations at follow-up timepoints? Addressing such question is of paramount in predicting common alternations across healthy and disordered populations. To fill this gap, we propose Recurrent Multigraph Integrator Network (ReMI-Net), the first graph recurrent neural network which infers the baseline CBT of an input population t1 and predicts its longitudinal evolution over time (ti > t1). Our ReMI-Net is composed of recurrent neural blocks with graph convolutional layers using a cross-node message passing to first learn hidden-states embeddings of each CBT node (i.e., brain region of interest) and then predict its evolution at the consecutive timepoint. Moreover, we design a novel time-dependent loss to regularize the CBT evolution trajectory over time and further introduce a cyclic recursion and learnable normalization layer to generate well-centered CBTs from time-dependent hidden-state embeddings. Finally, we derive the CBT adjacency matrix from the learned hidden state graph representation.

Islem Mhiri, Mohamed Ali Mahjoub, Islem Rekik

Synthesizing multimodality medical data provides complementary knowledge and helps doctors make precise clinical decisions. Although promising, existing multimodal brain graph synthesis frameworks have several limitations. First, they mainly tackle only one problem (intra- or inter-modality), limiting their generalizability to synthesizing inter- and intra-modality simultaneously. Second, while few techniques work on super-resolving low-resolution brain graphs within a single modality (i.e., intra), inter-modality graph super-resolution remains unexplored though this would avoid the need for costly data collection and processing. More importantly, both target and source domains might have different distributions, which causes a domain fracture between them. To fill these gaps, we propose a multi-resolution StairwayGraphNet (SG-Net) framework to jointly infer a target graph modality based on a given modality and super-resolve brain graphs in both inter and intra domains. Our SG-Net is grounded in three main contributions: (i) predicting a target graph from a source one based on a novel graph generative adversarial network in both inter (e.g., morphological-functional) and intra (e.g., functional-functional) domains, (ii) generating high-resolution brain graphs without resorting to the time consuming and expensive MRI processing steps, and (iii) enforcing the source distribution to match that of the ground truth graphs using an inter-modality aligner to relax the loss function to optimize. Moreover, we design a new Ground Truth-Preserving loss function to guide both generators in learning the topological structure of ground truth brain graphs more accurately. Our comprehensive experiments on predicting target brain graphs from source graphs using a multi-resolution stairway showed the outperformance of our method in comparison with its variants and state-of-the-art method.

Basar Demir, Alaa Bessadok, Islem Rekik

Accurate and automated super-resolution image synthesis is highly desired since it has the great potential to circumvent the need for acquiring high-cost medical scans and a time-consuming preprocessing pipeline of neuroimaging data. However, existing deep learning frameworks are solely designed to predict high-resolution (HR) image from a low-resolution (LR) one, which limits their generalization ability to brain graphs (i.e., connectomes). A small body of works has focused on superresolving brain graphs where the goal is to predict a HR graph from a single LR graph. Although promising, existing works mainly focus on superresolving graphs belonging to the same domain (e.g., functional), overlooking the domain fracture existing between multimodal brain data distributions (e.g., morphological and structural). To this aim, we propose a novel inter-domain adaptation framework namely, Learn to SuperResolve Brain Graphs with Knowledge Distillation Network (L2S-KDnet), which adopts a teacher-student paradigm to superresolve brain graphs. Our teacher network is a graph encoder-decoder that firstly learns the LR brain graph embeddings, and secondly learns how to align the resulting latent representations to the HR ground truth data distribution using an adversarial regularization. Ultimately, it decodes the HR graphs from the aligned embeddings. Next, our student network learns the knowledge of the aligned brain graphs as well as the topological structure of the predicted HR graphs transferred from the teacher. We further leverage the decoder of the teacher to optimize the student network. L2S-KDnet presents the first TS architecture tailored for brain graph super-resolution synthesis that is based on inter-domain alignment. Our experimental results demonstrate substantial performance gains over benchmark methods.

Alpay Tekin, Ahmed Nebli, Islem Rekik

Several brain disorders can be detected by observing alterations in the brain's structural and functional connectivities. Neurological findings suggest that early diagnosis of brain disorders, such as mild cognitive impairment (MCI), can prevent and even reverse its development into Alzheimer's disease (AD). In this context, recent studies aimed to predict the evolution of brain connectivities over time by proposing machine learning models that work on brain images. However, such an approach is costly and time-consuming. Here, we propose to use brain connectivities as a more efficient alternative for time-dependent brain disorder diagnosis by regarding the brain as instead a large interconnected graph characterizing the interconnectivity scheme between several brain regions. We term our proposed method Recurrent Brain Graph Mapper (RBGM), a novel efficient edge-based recurrent graph neural network that predicts the time-dependent evaluation trajectory of a brain graph from a single baseline. Our RBGM contains a set of recurrent neural network-inspired mappers for each time point, where each mapper aims to project the ground-truth brain graph onto its next time point. We leverage the teacher forcing method to boost training and improve the evolved brain graph quality. To maintain the topological consistency between the predicted brain graphs and their corresponding ground-truth brain graphs at each time point, we further integrate a topological loss. We also use l1 loss to capture time-dependency and minimize the distance between the brain graph at consecutive time points for regularization. Benchmarks against several variants of RBGM and state-of-the-art methods prove that we can achieve the same accuracy in predicting brain graph evolution more efficiently, paving the way for novel graph neural network architecture and a highly efficient training scheme.

Şeymanur Aktı, Doğay Kamar, Özgür Anıl Özlü et al.

Predicting the evolution of the brain network, also called connectome, by foreseeing changes in the connectivity weights linking pairs of anatomical regions makes it possible to spot connectivity-related neurological disorders in earlier stages and detect the development of potential connectomic anomalies. Remarkably, such a challenging prediction problem remains least explored in the predictive connectomics literature. It is a known fact that machine learning (ML) methods have proven their predictive abilities in a wide variety of computer vision problems. However, ML techniques specifically tailored for the prediction of brain connectivity evolution trajectory from a single timepoint are almost absent. To fill this gap, we organized a Kaggle competition where 20 competing teams designed advanced machine learning pipelines for predicting the brain connectivity evolution from a single timepoint. The competing teams developed their ML pipelines with a combination of data pre-processing, dimensionality reduction, and learning methods. Utilizing an inclusive evaluation approach, we ranked the methods based on two complementary evaluation metrics (mean absolute error (MAE) and Pearson Correlation Coefficient (PCC)) and their performances using different training and testing data perturbation strategies (single random split and cross-validation). The final rank was calculated using the rank product for each competing team across all evaluation measures and validation strategies. In support of open science, the developed 20 ML pipelines along with the connectomic dataset are made available on GitHub. The outcomes of this competition are anticipated to lead to the further development of predictive models that can foresee the evolution of brain connectivity over time, as well as other types of networks (e.g., genetic networks).

Islem Mhiri, Ahmed Nebli, Mohamed Ali Mahjoub et al.

Brain graph synthesis marked a new era for predicting a target brain graph from a source one without incurring the high acquisition cost and processing time of neuroimaging data. However, existing multi-modal graph synthesis frameworks have several limitations. First, they mainly focus on generating graphs from the same domain (intra-modality), overlooking the rich multimodal representations of brain connectivity (inter-modality). Second, they can only handle isomorphic graph generation tasks, limiting their generalizability to synthesizing target graphs with a different node size and topological structure from those of the source one. More importantly, both target and source domains might have different distributions, which causes a domain fracture between them (i.e., distribution misalignment). To address such challenges, we propose an inter-modality aligner of non-isomorphic graphs (IMANGraphNet) framework to infer a target graph modality based on a given modality. Our three core contributions lie in (i) predicting a target graph (e.g., functional) from a source graph (e.g., morphological) based on a novel graph generative adversarial network (gGAN); (ii) using non-isomorphic graphs for both source and target domains with a different number of nodes, edges and structure; and (iii) enforcing the predicted target distribution to match that of the ground truth graphs using a graph autoencoder to relax the designed loss oprimization. To handle the unstable behavior of gGAN, we design a new Ground Truth-Preserving (GT-P) loss function to guide the generator in learning the topological structure of ground truth brain graphs. Our comprehensive experiments on predicting functional from morphological graphs demonstrate the outperformance of IMANGraphNet in comparison with its variants. This can be further leveraged for integrative and holistic brain mapping in health and disease.

Martin Hanik, Mehmet Arif Demirtaş, Mohammed Amine Gharsallaoui et al.

Analyzing the relation between intelligence and neural activity is of the utmost importance in understanding the working principles of the human brain in health and disease. In existing literature, functional brain connectomes have been used successfully to predict cognitive measures such as intelligence quotient (IQ) scores in both healthy and disordered cohorts using machine learning models. However, existing methods resort to flattening the brain connectome (i.e., graph) through vectorization which overlooks its topological properties. To address this limitation and inspired from the emerging graph neural networks (GNNs), we design a novel regression GNN model (namely RegGNN) for predicting IQ scores from brain connectivity. On top of that, we introduce a novel, fully modular sample selection method to select the best samples to learn from for our target prediction task. However, since such deep learning architectures are computationally expensive to train, we further propose a \emph{learning-based sample selection} method that learns how to choose the training samples with the highest expected predictive power on unseen samples. For this, we capitalize on the fact that connectomes (i.e., their adjacency matrices) lie in the symmetric positive definite (SPD) matrix cone. Our results on full-scale and verbal IQ prediction outperforms comparison methods in autism spectrum disorder cohorts and achieves a competitive performance for neurotypical subjects using 3-fold cross-validation. Furthermore, we show that our sample selection approach generalizes to other learning-based methods, which shows its usefulness beyond our GNN architecture.

Alaa Bessadok, Mohamed Ali Mahjoub, Islem Rekik

Brain graphs (i.e, connectomes) constructed from medical scans such as magnetic resonance imaging (MRI) have become increasingly important tools to characterize the abnormal changes in the human brain. Due to the high acquisition cost and processing time of multimodal MRI, existing deep learning frameworks based on Generative Adversarial Network (GAN) focused on predicting the missing multimodal medical images from a few existing modalities. While brain graphs help better understand how a particular disorder can change the connectional facets of the brain, synthesizing a target brain multigraph (i.e, multiple brain graphs) from a single source brain graph is strikingly lacking. Additionally, existing graph generation works mainly learn one model for each target domain which limits their scalability in jointly predicting multiple target domains. Besides, while they consider the global topological scale of a graph (i.e., graph connectivity structure), they overlook the local topology at the node scale (e.g., how central a node is in the graph). To address these limitations, we introduce topology-aware graph GAN architecture (topoGAN), which jointly predicts multiple brain graphs from a single brain graph while preserving the topological structure of each target graph. Its three key innovations are: (i) designing a novel graph adversarial auto-encoder for predicting multiple brain graphs from a single one, (ii) clustering the encoded source graphs in order to handle the mode collapse issue of GAN and proposing a cluster-specific decoder, (iii) introducing a topological loss to force the prediction of topologically sound target brain graphs. The experimental results using five target domains demonstrated the outperformance of our method in brain multigraph prediction from a single graph in comparison with baseline approaches.

Alin Banka, Inis Buzi, Islem Rekik

Graph embedding is a powerful method to represent graph neurological data (e.g., brain connectomes) in a low dimensional space for brain connectivity mapping, prediction and classification. However, existing embedding algorithms have two major limitations. First, they primarily focus on preserving one-to-one topological relationships between nodes (i.e., regions of interest (ROIs) in a connectome), but they have mostly ignored many-to-many relationships (i.e., set to set), which can be captured using a hyperconnectome structure. Second, existing graph embedding techniques cannot be easily adapted to multi-view graph data with heterogeneous distributions. In this paper, while cross-pollinating adversarial deep learning with hypergraph theory, we aim to jointly learn deep latent embeddings of subject0specific multi-view brain graphs to eventually disentangle different brain states. First, we propose a new simple strategy to build a hyperconnectome for each brain view based on nearest neighbour algorithm to preserve the connectivities across pairs of ROIs. Second, we design a hyperconnectome autoencoder (HCAE) framework which operates directly on the multi-view hyperconnectomes based on hypergraph convolutional layers to better capture the many-to-many relationships between brain regions (i.e., nodes). For each subject, we further regularize the hypergraph autoencoding by adversarial regularization to align the distribution of the learned hyperconnectome embeddings with that of the input hyperconnectomes. We formalize our hyperconnectome embedding within a geometric deep learning framework to optimize for a given subject, thereby designing an individual-based learning framework. Our experiments showed that the learned embeddings by HCAE yield to better results for brain state classification compared with other deep graph embedding methods methods.

Mustafa Saglam, Islem Rekik

The individual brain can be viewed as a highly-complex multigraph (i.e. a set of graphs also called connectomes), where each graph represents a unique connectional view of pairwise brain region (node) relationships such as function or morphology. Due to its multifold complexity, understanding how brain disorders alter not only a single view of the brain graph, but its multigraph representation at the individual and population scales, remains one of the most challenging obstacles to profiling brain connectivity for ultimately disentangling a wide spectrum of brain states (e.g., healthy vs. disordered). In this work, while cross-pollinating the fields of spectral graph theory and diffusion models, we unprecedentedly propose an eigen-based cross-diffusion strategy for multigraph brain integration, comparison, and profiling. Specifically, we first devise a brain multigraph fusion model guided by eigenvector centrality to rely on most central nodes in the cross-diffusion process. Next, since the graph spectrum encodes its shape (or geometry) as if one can hear the shape of the graph, for the first time, we profile the fused multigraphs at several diffusion timescales by extracting the compact heat-trace signatures of their corresponding Laplacian matrices. Here, we reveal for the first time autistic and healthy profiles of morphological brain multigraphs, derived from T1-w magnetic resonance imaging (MRI), and demonstrate their discriminability in boosting the classification of unseen samples in comparison with state-of-the-art methods. This study presents the first step towards hearing the shape of the brain multigraph that can be leveraged for profiling and disentangling comorbid neurological disorders, thereby advancing precision medicine.

Ahmed Nebli, Islem Rekik

Brain connectivity networks, derived from magnetic resonance imaging (MRI), non-invasively quantify the relationship in function, structure, and morphology between two brain regions of interest (ROIs) and give insights into gender-related connectional differences. However, to the best of our knowledge, studies on gender differences in brain connectivity were limited to investigating pairwise (i.e., low-order) relationship ROIs, overlooking the complex high-order interconnectedness of the brain as a network. To address this limitation, brain multiplexes have been introduced to model the relationship between at least two different brain networks. However, this inhibits their application to datasets with single brain networks such as functional networks. To fill this gap, we propose the first work on predicting brain multiplexes from a source network to investigate gender differences. Recently, generative adversarial networks (GANs) submerged the field of medical data synthesis. However, although conventional GANs work well on images, they cannot handle brain networks due to their non-Euclidean topological structure. Differently, in this paper, we tap into the nascent field of geometric-GANs (G-GAN) to design a deep multiplex prediction architecture comprising (i) a geometric source to target network translator mimicking a U-Net architecture with skip connections and (ii) a conditional discriminator which classifies predicted target intra-layers by conditioning on the multiplex source intra-layers. Such architecture simultaneously learns the latent source network representation and the deep non-linear mapping from the source to target multiplex intra-layers. Our experiments on a large dataset demonstrated that predicted multiplexes significantly boost gender classification accuracy compared with source networks and identifies both low and high-order gender-specific multiplex connections.

Ahmet Serkan Goktas, Alaa Bessadok, Islem Rekik

While existing predictive frameworks are able to handle Euclidean structured data (i.e, brain images), they might fail to generalize to geometric non-Euclidean data such as brain networks. Besides, these are rooted the sample selection step in using Euclidean or learned similarity measure between vectorized training and testing brain networks. Such sample connectomic representation might include irrelevant and redundant features that could mislead the training sample selection step. Undoubtedly, this fails to exploit and preserve the topology of the brain connectome. To overcome this major drawback, we propose Residual Embedding Similarity-Based Network selection (RESNets) for predicting brain network evolution trajectory from a single timepoint. RESNets first learns a compact geometric embedding of each training and testing sample using adversarial connectome embedding network. This nicely reduces the high-dimensionality of brain networks while preserving their topological properties via graph convolutional networks. Next, to compute the similarity between subjects, we introduce the concept of a connectional brain template (CBT), a fixed network reference, where we further represent each training and testing network as a deviation from the reference CBT in the embedding space. As such, we select the most similar training subjects to the testing subject at baseline by comparing their learned residual embeddings with respect to the pre-defined CBT. Once the best training samples are selected at baseline, we simply average their corresponding brain networks at follow-up timepoints to predict the evolution trajectory of the testing network. Our experiments on both healthy and disordered brain networks demonstrate the success of our proposed method in comparison to RESNets ablated versions and traditional approaches.

Megi Isallari, Islem Rekik

Catchy but rigorous deep learning architectures were tailored for image super-resolution (SR), however, these fail to generalize to non-Euclidean data such as brain connectomes. Specifically, building generative models for super-resolving a low-resolution (LR) brain connectome at a higher resolution (HR) (i.e., adding new graph nodes/edges) remains unexplored although this would circumvent the need for costly data collection and manual labelling of anatomical brain regions (i.e. parcellation). To fill this gap, we introduce GSR-Net (Graph Super-Resolution Network), the first super-resolution framework operating on graph-structured data that generates high-resolution brain graphs from low-resolution graphs. First, we adopt a U-Net like architecture based on graph convolution, pooling and unpooling operations specific to non-Euclidean data. However, unlike conventional U-Nets where graph nodes represent samples and node features are mapped to a low-dimensional space (encoding and decoding node attributes or sample features), our GSR-Net operates directly on a single connectome: a fully connected graph where conventionally, a node denotes a brain region, nodes have no features, and edge weights denote brain connectivity strength between two regions of interest (ROIs). In the absence of original node features, we initially assign identity feature vectors to each brain ROI (node) and then leverage the learned local receptive fields to learn node feature representations. Second, inspired by spectral theory, we break the symmetry of the U-Net architecture by topping it up with a graph super-resolution (GSR) layer and two graph convolutional network layers to predict a HR graph while preserving the characteristics of the LR input. Our proposed GSR-Net framework outperformed its variants for predicting high-resolution brain functional connectomes from low-resolution connectomes.

Alaa Bessadok, Mohamed Ali Mahjoub, Islem Rekik

Several works based on Generative Adversarial Networks (GAN) have been recently proposed to predict a set of medical images from a single modality (e.g, FLAIR MRI from T1 MRI). However, such frameworks are primarily designed to operate on images, limiting their generalizability to non-Euclidean geometric data such as brain graphs. While a growing number of connectomic studies has demonstrated the promise of including brain graphs for diagnosing neurological disorders, no geometric deep learning work was designed for multiple target brain graphs prediction from a source brain graph. Despite the momentum the field of graph generation has gained in the last two years, existing works have two critical drawbacks. First, the bulk of such works aims to learn one model for each target domain to generate from a source domain. Thus, they have a limited scalability in jointly predicting multiple target domains. Second, they merely consider the global topological scale of a graph (i.e., graph connectivity structure) and overlook the local topology at the node scale of a graph (e.g., how central a node is in the graph). To meet these challenges, we introduce MultiGraphGAN architecture, which not only predicts multiple brain graphs from a single brain graph but also preserves the topological structure of each target graph to predict. Its three core contributions lie in: (i) designing a graph adversarial auto-encoder for jointly predicting brain graphs from a single one, (ii) handling the mode collapse problem of GAN by clustering the encoded source graphs and proposing a cluster-specific decoder, (iii) introducing a topological loss to force the reconstruction of topologically sound target brain graphs. Our MultiGraphGAN significantly outperformed its variants thereby showing its great potential in multi-view brain graph generation from a single graph.

Islem Mhiri, Mohamed Ali Mahjoub, Islem Rekik

Estimating a representative and discriminative brain network atlas (BNA) is a nascent research field in mapping a population of brain networks in health and disease. Although limited, existing BNA estimation methods have several limitations. First, they primarily rely on a similarity network diffusion and fusion technique, which only considers node degree as a topological measure in the cross-network diffusion process, thereby overlooking rich topological measures of the brain network (e.g., centrality). Second, both diffusion and fusion techniques are implemented in fully unsupervised manner, which might decrease the discriminative power of the estimated BNAs. To fill these gaps, we propose a supervised multi-topology network cross-diffusion (SM-netFusion) framework for estimating a BNA satisfying : (i) well-representativeness (captures shared traits across subjects), (ii) well-centeredness (optimally close to all subjects), and (iii) high discriminativeness (can easily and efficiently identify discriminative brain connections that distinguish between two populations). For a specific class, given the cluster labels of the training data, we learn a weighted combination of the topological diffusion kernels derived from degree, closeness and eigenvector centrality measures in a supervised manner. Specifically, we learn the cross-diffusion process by normalizing the training brain networks using the learned diffusion kernels. Our SM-netFusion produces the most centered and representative template in comparison with its variants and state-of-the-art methods and further boosted the classification of autistic subjects by 5-15%. SM-netFusion presents the first work for supervised network cross-diffusion based on graph topological measures, which can be further leveraged to design an efficient graph feature selection method for training predictive learners in network neuroscience.

Ismail Bilgen, Goktug Guvercin, Islem Rekik

Autism spectrum disorder (ASD) affects the brain connectivity at different levels. Nonetheless, non-invasively distinguishing such effects using magnetic resonance imaging (MRI) remains very challenging to machine learning diagnostic frameworks due to ASD heterogeneity. So far, existing network neuroscience works mainly focused on functional (derived from functional MRI) and structural (derived from diffusion MRI) brain connectivity, which might not capture relational morphological changes between brain regions. Indeed, machine learning (ML) studies for ASD diagnosis using morphological brain networks derived from conventional T1-weighted MRI are very scarce. To fill this gap, we leverage crowdsourcing by organizing a Kaggle competition to build a pool of machine learning pipelines for neurological disorder diagnosis with application to ASD diagnosis using cortical morphological networks derived from T1-weighted MRI. During the competition, participants were provided with a training dataset and only allowed to check their performance on a public test data. The final evaluation was performed on both public and hidden test datasets based on accuracy, sensitivity, and specificity metrics. Teams were ranked using each performance metric separately and the final ranking was determined based on the mean of all rankings. The first-ranked team achieved 70% accuracy, 72.5% sensitivity, and 67.5% specificity, while the second-ranked team achieved 63.8%, 62.5%, 65% respectively. Leveraging participants to design ML diagnostic methods within a competitive machine learning setting has allowed the exploration and benchmarking of wide spectrum of ML methods for ASD diagnosis using cortical morphological networks.

Can Gafuroglu, Islem Rekik

Patients initially diagnosed with early mild cognitive impairment (eMCI) are known to be a clinically heterogeneous group with very subtle patterns of brain atrophy. To examine the boarders between normal controls (NC) and eMCI, Magnetic Resonance Imaging (MRI) was extensively used as a non-invasive imaging modality to pin-down subtle changes in brain images of MCI patients. However, eMCI research remains limited by the number of available MRI acquisition timepoints. Ideally, one would learn how to diagnose MCI patients in an early stage from MRI data acquired at a single timepoint, while leveraging 'non-existing' follow-up observations. To this aim, we propose novel supervised and unsupervised frameworks that learn how to jointly predict and label the evolution trajectory of intensity patches, each seeded at a specific brain landmark, from a baseline intensity patch. Specifically, both strategies aim to identify the best training atlas patches at baseline timepoint to predict and classify the evolution trajectory of a given testing baseline patch. The supervised technique learns how to select the best atlas patches by training bidirectional mappings from the space of pairwise patch similarities to their corresponding prediction errors -when one patch was used to predict the other. On the other hand, the unsupervised technique learns a manifold of baseline atlas and testing patches using multiple kernels to well capture patch distributions at multiple scales. Once the best baseline atlas patches are selected, we retrieve their evolution trajectories and average them to predict the evolution trajectory of the testing baseline patch. Next, we input the predicted trajectories to an ensemble of linear classifiers, each trained at a specific landmark. Our classification accuracy increased by up to 10% points in comparison to single timepoint-based classification methods.