Purang Abolmaesumi

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.

Hooman Vaseli, Ang Nan Gu, S. Neda Ahmadi Amiri et al.

Aortic stenosis (AS) is a common heart valve disease that requires accurate and timely diagnosis for appropriate treatment. Most current automatic AS severity detection methods rely on black-box models with a low level of trustworthiness, which hinders clinical adoption. To address this issue, we propose ProtoASNet, a prototypical network that directly detects AS from B-mode echocardiography videos, while making interpretable predictions based on the similarity between the input and learned spatio-temporal prototypes. This approach provides supporting evidence that is clinically relevant, as the prototypes typically highlight markers such as calcification and restricted movement of aortic valve leaflets. Moreover, ProtoASNet utilizes abstention loss to estimate aleatoric uncertainty by defining a set of prototypes that capture ambiguity and insufficient information in the observed data. This provides a reliable system that can detect and explain when it may fail. We evaluate ProtoASNet on a private dataset and the publicly available TMED-2 dataset, where it outperforms existing state-of-the-art methods with an accuracy of 80.0% and 79.7%, respectively. Furthermore, ProtoASNet provides interpretability and an uncertainty measure for each prediction, which can improve transparency and facilitate the interactive usage of deep networks to aid clinical decision-making. Our source code is available at: https://github.com/hooman007/ProtoASNet.

Mahdi Gilany, Paul Wilson, Andrea Perera-Ortega et al.

A large body of previous machine learning methods for ultrasound-based prostate cancer detection classify small regions of interest (ROIs) of ultrasound signals that lie within a larger needle trace corresponding to a prostate tissue biopsy (called biopsy core). These ROI-scale models suffer from weak labeling as histopathology results available for biopsy cores only approximate the distribution of cancer in the ROIs. ROI-scale models do not take advantage of contextual information that are normally considered by pathologists, i.e. they do not consider information about surrounding tissue and larger-scale trends when identifying cancer. We aim to improve cancer detection by taking a multi-scale, i.e. ROI-scale and biopsy core-scale, approach. Methods: Our multi-scale approach combines (i) an "ROI-scale" model trained using self-supervised learning to extract features from small ROIs and (ii) a "core-scale" transformer model that processes a collection of extracted features from multiple ROIs in the needle trace region to predict the tissue type of the corresponding core. Attention maps, as a byproduct, allow us to localize cancer at the ROI scale. We analyze this method using a dataset of micro-ultrasound acquired from 578 patients who underwent prostate biopsy, and compare our model to baseline models and other large-scale studies in the literature. Results and Conclusions: Our model shows consistent and substantial performance improvements compared to ROI-scale-only models. It achieves 80.3% AUROC, a statistically significant improvement over ROI-scale classification. We also compare our method to large studies on prostate cancer detection, using other imaging modalities. Our code is publicly available at www.github.com/med-i-lab/TRUSFormer

Fahimeh Fooladgar, Minh Nguyen Nhat To, Parvin Mousavi et al.

Deep neural networks have proven to be highly effective when large amounts of data with clean labels are available. However, their performance degrades when training data contains noisy labels, leading to poor generalization on the test set. Real-world datasets contain noisy label samples that either have similar visual semantics to other classes (in-distribution) or have no semantic relevance to any class (out-of-distribution) in the dataset. Most state-of-the-art methods leverage ID labeled noisy samples as unlabeled data for semi-supervised learning, but OOD labeled noisy samples cannot be used in this way because they do not belong to any class within the dataset. Hence, in this paper, we propose incorporating the information from all the training data by leveraging the benefits of self-supervised training. Our method aims to extract a meaningful and generalizable embedding space for each sample regardless of its label. Then, we employ a simple yet effective K-nearest neighbor method to remove portions of out-of-distribution samples. By discarding these samples, we propose an iterative "Manifold DivideMix" algorithm to find clean and noisy samples, and train our model in a semi-supervised way. In addition, we propose "MixEMatch", a new algorithm for the semi-supervised step that involves mixup augmentation at the input and final hidden representations of the model. This will extract better representations by interpolating both in the input and manifold spaces. Extensive experiments on multiple synthetic-noise image benchmarks and real-world web-crawled datasets demonstrate the effectiveness of our proposed framework. Code is available at https://github.com/Fahim-F/ManifoldDivideMix.

Masoud Mokhtari, Teresa Tsang, Purang Abolmaesumi et al.

Ejection fraction (EF) is a key indicator of cardiac function, allowing identification of patients prone to heart dysfunctions such as heart failure. EF is estimated from cardiac ultrasound videos known as echocardiograms (echo) by manually tracing the left ventricle and estimating its volume on certain frames. These estimations exhibit high inter-observer variability due to the manual process and varying video quality. Such sources of inaccuracy and the need for rapid assessment necessitate reliable and explainable machine learning techniques. In this work, we introduce EchoGNN, a model based on graph neural networks (GNNs) to estimate EF from echo videos. Our model first infers a latent echo-graph from the frames of one or multiple echo cine series. It then estimates weights over nodes and edges of this graph, indicating the importance of individual frames that aid EF estimation. A GNN regressor uses this weighted graph to predict EF. We show, qualitatively and quantitatively, that the learned graph weights provide explainability through identification of critical frames for EF estimation, which can be used to determine when human intervention is required. On EchoNet-Dynamic public EF dataset, EchoGNN achieves EF prediction performance that is on par with state of the art and provides explainability, which is crucial given the high inter-observer variability inherent in this task.

Masoud Mokhtari, Neda Ahmadi, Teresa S. M. Tsang et al.

Echocardiography (echo) is an ultrasound imaging modality that is widely used for various cardiovascular diagnosis tasks. Due to inter-observer variability in echo-based diagnosis, which arises from the variability in echo image acquisition and the interpretation of echo images based on clinical experience, vision-based machine learning (ML) methods have gained popularity to act as secondary layers of verification. For such safety-critical applications, it is essential for any proposed ML method to present a level of explainability along with good accuracy. In addition, such methods must be able to process several echo videos obtained from various heart views and the interactions among them to properly produce predictions for a variety of cardiovascular measurements or interpretation tasks. Prior work lacks explainability or is limited in scope by focusing on a single cardiovascular task. To remedy this, we propose a General, Echo-based, Multi-Level Transformer (GEMTrans) framework that provides explainability, while simultaneously enabling multi-video training where the inter-play among echo image patches in the same frame, all frames in the same video, and inter-video relationships are captured based on a downstream task. We show the flexibility of our framework by considering two critical tasks including ejection fraction (EF) and aortic stenosis (AS) severity detection. Our model achieves mean absolute errors of 4.15 and 4.84 for single and dual-video EF estimation and an accuracy of 96.5 % for AS detection, while providing informative task-specific attention maps and prototypical explainability.

Paul F. R. Wilson, Mahdi Gilany, Amoon Jamzad et al.

Deep learning-based analysis of high-frequency, high-resolution micro-ultrasound data shows great promise for prostate cancer detection. Previous approaches to analysis of ultrasound data largely follow a supervised learning paradigm. Ground truth labels for ultrasound images used for training deep networks often include coarse annotations generated from the histopathological analysis of tissue samples obtained via biopsy. This creates inherent limitations on the availability and quality of labeled data, posing major challenges to the success of supervised learning methods. On the other hand, unlabeled prostate ultrasound data are more abundant. In this work, we successfully apply self-supervised representation learning to micro-ultrasound data. Using ultrasound data from 1028 biopsy cores of 391 subjects obtained in two clinical centres, we demonstrate that feature representations learnt with this method can be used to classify cancer from non-cancer tissue, obtaining an AUROC score of 91% on an independent test set. To the best of our knowledge, this is the first successful end-to-end self-supervised learning approach for prostate cancer detection using ultrasound data. Our method outperforms baseline supervised learning approaches, generalizes well between different data centers, and scale well in performance as more unlabeled data are added, making it a promising approach for future research using large volumes of unlabeled data.

Mahdi Gilany, Paul Wilson, Amoon Jamzad et al.

MOTIVATION: Detection of prostate cancer during transrectal ultrasound-guided biopsy is challenging. The highly heterogeneous appearance of cancer, presence of ultrasound artefacts, and noise all contribute to these difficulties. Recent advancements in high-frequency ultrasound imaging - micro-ultrasound - have drastically increased the capability of tissue imaging at high resolution. Our aim is to investigate the development of a robust deep learning model specifically for micro-ultrasound-guided prostate cancer biopsy. For the model to be clinically adopted, a key challenge is to design a solution that can confidently identify the cancer, while learning from coarse histopathology measurements of biopsy samples that introduce weak labels. METHODS: We use a dataset of micro-ultrasound images acquired from 194 patients, who underwent prostate biopsy. We train a deep model using a co-teaching paradigm to handle noise in labels, together with an evidential deep learning method for uncertainty estimation. We evaluate the performance of our model using the clinically relevant metric of accuracy vs. confidence. RESULTS: Our model achieves a well-calibrated estimation of predictive uncertainty with area under the curve of 88$\%$. The use of co-teaching and evidential deep learning in combination yields significantly better uncertainty estimation than either alone. We also provide a detailed comparison against state-of-the-art in uncertainty estimation.

Sana Ayromlou, Purang Abolmaesumi, Teresa Tsang et al.

Standard deep learning-based classification approaches require collecting all samples from all classes in advance and are trained offline. This paradigm may not be practical in real-world clinical applications, where new classes are incrementally introduced through the addition of new data. Class incremental learning is a strategy allowing learning from such data. However, a major challenge is catastrophic forgetting, i.e., performance degradation on previous classes when adapting a trained model to new data. Prior methodologies to alleviate this challenge save a portion of training data require perpetual storage of such data that may introduce privacy issues. Here, we propose a novel data-free class incremental learning framework that first synthesizes data from the model trained on previous classes to generate a \ours. Subsequently, it updates the model by combining the synthesized data with new class data. Furthermore, we incorporate a cosine normalized Cross-entropy loss to mitigate the adverse effects of the imbalance, a margin loss to increase separation among previous classes and new ones, and an intra-domain contrastive loss to generalize the model trained on the synthesized data to real data. We compare our proposed framework with state-of-the-art methods in class incremental learning, where we demonstrate improvement in accuracy for the classification of 11,062 echocardiography cine series of patients.

Masoud Mokhtari, Mobina Mahdavi, Hooman Vaseli et al.

The functional assessment of the left ventricle chamber of the heart requires detecting four landmark locations and measuring the internal dimension of the left ventricle and the approximate mass of the surrounding muscle. The key challenge of automating this task with machine learning is the sparsity of clinical labels, i.e., only a few landmark pixels in a high-dimensional image are annotated, leading many prior works to heavily rely on isotropic label smoothing. However, such a label smoothing strategy ignores the anatomical information of the image and induces some bias. To address this challenge, we introduce an echocardiogram-based, hierarchical graph neural network (GNN) for left ventricle landmark detection (EchoGLAD). Our main contributions are: 1) a hierarchical graph representation learning framework for multi-resolution landmark detection via GNNs; 2) induced hierarchical supervision at different levels of granularity using a multi-level loss. We evaluate our model on a public and a private dataset under the in-distribution (ID) and out-of-distribution (OOD) settings. For the ID setting, we achieve the state-of-the-art mean absolute errors (MAEs) of 1.46 mm and 1.86 mm on the two datasets. Our model also shows better OOD generalization than prior works with a testing MAE of 4.3 mm.

Bahar Khodabakhshian, Nima Hashemi, Armin Saadat et al.

Purpose: Myocardium segmentation in echocardiography videos is a challenging task due to low contrast, noise, and anatomical variability. Traditional deep learning models either process frames independently, ignoring temporal information, or rely on memory-based feature propagation, which accumulates error over time. Methods: We propose Point-Seg, a transformer-based segmentation framework that integrates point tracking as a temporal cue to ensure stable and consistent segmentation of myocardium across frames. Our method leverages a point-tracking module trained on a synthetic echocardiography dataset to track key anatomical landmarks across video sequences. These tracked trajectories provide an explicit motion-aware signal that guides segmentation, reducing drift and eliminating the need for memory-based feature accumulation. Additionally, we incorporate a temporal smoothing loss to further enhance temporal consistency across frames. Results: We evaluate our approach on both public and private echocardiography datasets. Experimental results demonstrate that Point-Seg has statistically similar accuracy in terms of Dice to state-of-the-art segmentation models in high quality echo data, while it achieves better segmentation accuracy in lower quality echo with improved temporal stability. Furthermore, Point-Seg has the key advantage of pixel-level myocardium motion information as opposed to other segmentation methods. Such information is essential in the computation of other downstream tasks such as myocardial strain measurement and regional wall motion abnormality detection. Conclusion: Point-Seg demonstrates that point tracking can serve as an effective temporal cue for consistent video segmentation, offering a reliable and generalizable approach for myocardium segmentation in echocardiography videos. The code is available at https://github.com/DeepRCL/PointSeg.

Yeganeh Ghamary, Victoria Wu, Hooman Vaseli et al.

Ejection fraction (EF) is a crucial metric for assessing cardiac function and diagnosing conditions such as heart failure. Traditionally, EF estimation requires manual tracing and domain expertise, making the process time-consuming and subject to interobserver variability. Most current deep learning methods for EF prediction are black-box models with limited transparency, which reduces clinical trust. Some post-hoc explainability methods have been proposed to interpret the decision-making process after the prediction is made. However, these explanations do not guide the model's internal reasoning and therefore offer limited reliability in clinical applications. To address this, we introduce ProtoEFNet, a novel video-based prototype learning model for continuous EF regression. The model learns dynamic spatiotemporal prototypes that capture clinically meaningful cardiac motion patterns. Additionally, the proposed Prototype Angular Separation (PAS) loss enforces discriminative representations across the continuous EF spectrum. Our experiments on the EchonetDynamic dataset show that ProtoEFNet can achieve accuracy on par with its non-interpretable counterpart while providing clinically relevant insight. The ablation study shows that the proposed loss boosts performance with a 2% increase in F1 score from 77.67$\pm$2.68 to 79.64$\pm$2.10. Our source code is available at: https://github.com/DeepRCL/ProtoEF

Sarah Nassar, Nooshin Maghsoodi, Sophia Mannina et al.

Objective: Atrial fibrillation (AF) is the most common cardiac arrhythmia experienced by intensive care unit (ICU) patients and can cause adverse health effects. In this study, we publish a labelled ICU dataset and benchmarks for AF detection. Methods: We compared machine learning models across three data-driven artificial intelligence (AI) approaches: feature-based classifiers, deep learning (DL), and ECG foundation models (FMs). This comparison addresses a critical gap in the literature and aims to pinpoint which AI approach is best for accurate AF detection. Electrocardiograms (ECGs) from a Canadian ICU and the 2021 PhysioNet/Computing in Cardiology Challenge were used to conduct the experiments. Multiple training configurations were tested, ranging from zero-shot inference to transfer learning. Results: On average and across both datasets, ECG FMs performed best, followed by DL, then feature-based classifiers. The model that achieved the top F1 score on our ICU test set was ECG-FM through a transfer learning strategy (F1=0.89). Conclusion: This study demonstrates promising potential for using AI to build an automatic patient monitoring system. Significance: By publishing our labelled ICU dataset (LinkToBeAdded) and performance benchmarks, this work enables the research community to continue advancing the state-of-the-art in AF detection in the ICU.

Mahdi Gilany, Mohamed Harmanani, Paul Wilson et al.

High resolution micro-ultrasound has demonstrated promise in real-time prostate cancer detection, with deep learning becoming a prominent tool for learning complex tissue properties reflected on ultrasound. However, a significant roadblock to real-world deployment remains, which prior works often overlook: model performance suffers when applied to data from different clinical centers due to variations in data distribution. This distribution shift significantly impacts the model's robustness, posing major challenge to clinical deployment. Domain adaptation and specifically its test-time adaption (TTA) variant offer a promising solution to address this challenge. In a setting designed to reflect real-world conditions, we compare existing methods to state-of-the-art TTA approaches adopted for cancer detection, demonstrating the lack of robustness to distribution shifts in the former. We then propose Diverse Ensemble Entropy Minimization (DEnEM), questioning the effectiveness of current TTA methods on ultrasound data. We show that these methods, although outperforming baselines, are suboptimal due to relying on neural networks output probabilities, which could be uncalibrated, or relying on data augmentation, which is not straightforward to define on ultrasound data. Our results show a significant improvement of $5\%$ to $7\%$ in AUROC over the existing methods and $3\%$ to $5\%$ over TTA methods, demonstrating the advantage of DEnEM in addressing distribution shift. \keywords{Ultrasound Imaging \and Prostate Cancer \and Computer-aided Diagnosis \and Distribution Shift Robustness \and Test-time Adaptation.}

Mohammad Mahdi Abootorabi, Parvin Mousavi, Purang Abolmaesumi et al.

Deep networks that rely on prototypes-interpretable representations that can be related to the model input-have gained significant attention for balancing high accuracy with inherent interpretability, which makes them suitable for critical domains such as healthcare. However, these models are limited by their reliance on training data, which hampers their robustness to distribution shifts. While test-time adaptation (TTA) improves the robustness of deep networks by updating parameters and statistics, the prototypes of interpretable models have not been explored for this purpose. We introduce ProtoTTA, a general framework for prototypical models that leverages intermediate prototype signals rather than relying solely on model outputs. ProtoTTA minimizes the entropy of the prototype-similarity distribution to encourage more confident and prototype-specific activations on shifted data. To maintain stability, we employ geometric filtering to restrict updates to samples with reliable prototype activations, regularized by prototype-importance weights and model-confidence scores. Experiments across four prototypical backbones on four diverse benchmarks spanning fine-grained vision, histopathology, and NLP demonstrate that ProtoTTA improves robustness over standard output entropy minimization while restoring correct semantic focus in prototype activations. We also introduce novel interpretability metrics and a vision-language model (VLM) evaluation framework to explain TTA dynamics, confirming ProtoTTA restores human-aligned semantic focus and correlates reliably with VLM-rated reasoning quality. Code is available at: https://github.com/DeepRCL/ProtoTTA.

Ang Nan Gu, Michael Tsang, Hooman Vaseli et al.

The fundamental problem with ultrasound-guided diagnosis is that the acquired images are often 2-D cross-sections of a 3-D anatomy, potentially missing important anatomical details. This limitation leads to challenges in ultrasound echocardiography, such as poor visualization of heart valves or foreshortening of ventricles. Clinicians must interpret these images with inherent uncertainty, a nuance absent in machine learning's one-hot labels. We propose Re-Training for Uncertainty (RT4U), a data-centric method to introduce uncertainty to weakly informative inputs in the training set. This simple approach can be incorporated to existing state-of-the-art aortic stenosis classification methods to further improve their accuracy. When combined with conformal prediction techniques, RT4U can yield adaptively sized prediction sets which are guaranteed to contain the ground truth class to a high accuracy. We validate the effectiveness of RT4U on three diverse datasets: a public (TMED-2) and a private AS dataset, along with a CIFAR-10-derived toy dataset. Results show improvement on all the datasets.

Emma Willis, Tarek Elghareb, Paul F. R. Wilson et al.

Purpose: Non-invasive grading of prostate cancer (PCa) from micro-ultrasound (micro-US) could expedite triage and guide biopsies toward the most aggressive regions, yet current models struggle to infer tissue micro-structure at coarse imaging resolutions. Methods: We introduce an unpaired histopathology knowledge-distillation strategy that trains a micro-US encoder to emulate the embedding distribution of a pretrained histopathology foundation model, conditioned on International Society of Urological Pathology (ISUP) grades. Training requires no patient-level pairing or image registration, and histopathology inputs are not used at inference. Results: Compared to the current state of the art, our approach increases sensitivity to clinically significant PCa (csPCa) at 60% specificity by 3.5% and improves overall sensitivity at 60% specificity by 1.2%. Conclusion: By enabling earlier and more dependable cancer risk stratification solely from imaging, our method advances clinical feasibility. Source code will be publicly released upon publication.

Paul F. R. Wilson, Mohamed Harmanani, Minh Nguyen Nhat To et al.

Purpose: Medical foundation models (FMs) offer a path to build high-performance diagnostic systems. However, their application to prostate cancer (PCa) detection from micro-ultrasound (μUS) remains untested in clinical settings. We present ProstNFound+, an adaptation of FMs for PCa detection from μUS, along with its first prospective validation. Methods: ProstNFound+ incorporates a medical FM, adapter tuning, and a custom prompt encoder that embeds PCa-specific clinical biomarkers. The model generates a cancer heatmap and a risk score for clinically significant PCa. Following training on multi-center retrospective data, the model is prospectively evaluated on data acquired five years later from a new clinical site. Model predictions are benchmarked against standard clinical scoring protocols (PRI-MUS and PI-RADS). Results: ProstNFound+ shows strong generalization to the prospective data, with no performance degradation compared to retrospective evaluation. It aligns closely with clinical scores and produces interpretable heatmaps consistent with biopsy-confirmed lesions. Conclusion: The results highlight its potential for clinical deployment, offering a scalable and interpretable alternative to expert-driven protocols.

Armin Saadat, Nima Hashemi, Hooman Vaseli et al.

Aortic stenosis (AS) is a life-threatening condition caused by a narrowing of the aortic valve, leading to impaired blood flow. Despite its high prevalence, access to echocardiography (echo), the gold-standard diagnostic tool, is often limited due to resource constraints, particularly in rural and underserved areas. Point-of-care ultrasound (POCUS) offers a more accessible alternative but is restricted by operator expertise and the challenge of selecting the most relevant imaging views. To address this, we propose a reinforcement learning (RL)-driven active video acquisition framework that dynamically selects each patient's most informative echo videos. Unlike traditional methods that rely on a fixed set of videos, our approach continuously evaluates whether additional imaging is needed, optimizing both accuracy and efficiency. Tested on data from 2,572 patients, our method achieves 80.6% classification accuracy while using only 47% of the echo videos compared to a full acquisition. These results demonstrate the potential of active feature acquisition to enhance AS diagnosis, making echocardiographic assessments more efficient, scalable, and personalized. Our source code is available at: https://github.com/Armin-Saadat/PRECISE-AS.

Minh Nguyen Nhat To, Paul F RWilson, Viet Nguyen et al.

The subpopulationtion shift, characterized by a disparity in subpopulation distributibetween theween the training and target datasets, can significantly degrade the performance of machine learning models. Current solutions to subpopulation shift involve modifying empirical risk minimization with re-weighting strategies to improve generalization. This strategy relies on assumptions about the number and nature of subpopulations and annotations on group membership, which are unavailable for many real-world datasets. Instead, we propose using an ensemble of diverse classifiers to adaptively capture risk associated with subpopulations. Given a feature extractor network, we replace its standard linear classification layer with a mixture of prototypical classifiers, where each member is trained to classify the data while focusing on different features and samples from other members. In empirical evaluation on nine real-world datasets, covering diverse domains and kinds of subpopulation shift, our method of Diverse Prototypical Ensembles (DPEs) often outperforms the prior state-of-the-art in worst-group accuracy. The code is available at https://github.com/minhto2802/dpe4subpop

Fatemeh Taheri Dezaki, Christina Luong, Tom Ginsberg et al.

In echocardiography (echo), an electrocardiogram (ECG) is conventionally used to temporally align different cardiac views for assessing critical measurements. However, in emergencies or point-of-care situations, acquiring an ECG is often not an option, hence motivating the need for alternative temporal synchronization methods. Here, we propose Echo-SyncNet, a self-supervised learning framework to synchronize various cross-sectional 2D echo series without any external input. The proposed framework takes advantage of both intra-view and inter-view self supervisions. The former relies on spatiotemporal patterns found between the frames of a single echo cine and the latter on the interdependencies between multiple cines. The combined supervisions are used to learn a feature-rich embedding space where multiple echo cines can be temporally synchronized. We evaluate the framework with multiple experiments: 1) Using data from 998 patients, Echo-SyncNet shows promising results for synchronizing Apical 2 chamber and Apical 4 chamber cardiac views; 2) Using data from 3070 patients, our experiments reveal that the learned representations of Echo-SyncNet outperform a supervised deep learning method that is optimized for automatic detection of fine-grained cardiac phase; 3) We show the usefulness of the learned representations in a one-shot learning scenario of cardiac keyframe detection. Without any fine-tuning, keyframes in 1188 validation patient studies are identified by synchronizing them with only one labeled reference study. We do not make any prior assumption about what specific cardiac views are used for training and show that Echo-SyncNet can accurately generalize to views not present in its training set. Project repository: github.com/fatemehtd/Echo-SyncNet.

Atrin Arya, Sana Ayromlou, Armin Saadat et al.

Standard deep learning-based classification approaches may not always be practical in real-world clinical applications, as they require a centralized collection of all samples. Federated learning (FL) provides a paradigm that can learn from distributed datasets across clients without requiring them to share data, which can help mitigate privacy and data ownership issues. In FL, sub-optimal convergence caused by data heterogeneity is common among data from different health centers due to the variety in data collection protocols and patient demographics across centers. Through experimentation in this study, we show that data heterogeneity leads to the phenomenon of catastrophic forgetting during local training. We propose FedImpres which alleviates catastrophic forgetting by restoring synthetic data that represents the global information as federated impression. To achieve this, we distill the global model resulting from each communication round. Subsequently, we use the synthetic data alongside the local data to enhance the generalization of local training. Extensive experiments show that the proposed method achieves state-of-the-art performance on both the BloodMNIST and Retina datasets, which contain label imbalance and domain shift, with an improvement in classification accuracy of up to 20%.

Mohamed Harmanani, Amoon Jamzad, Minh Nguyen Nhat To et al.

Prostate cancer (PCa) detection using deep learning (DL) models has shown potential for enhancing real-time guidance during biopsies. However, prostate ultrasound images lack pixel-level cancer annotations, introducing label noise. Current approaches often focus on limited regions of interest (ROIs), disregarding anatomical context necessary for accurate diagnosis. Foundation models can overcome this limitation by analyzing entire images to capture global spatial relationships; however, they still encounter challenges stemming from the weak labels associated with coarse pathology annotations in ultrasound data. We introduce Cinepro, a novel framework that strengthens foundation models' ability to localize PCa in ultrasound cineloops. Cinepro adapts robust training by integrating the proportion of cancer tissue reported by pathology in a biopsy core into its loss function to address label noise, providing a more nuanced supervision. Additionally, it leverages temporal data across multiple frames to apply robust augmentations, enhancing the model's ability to learn stable cancer-related features. Cinepro demonstrates superior performance on a multi-center prostate ultrasound dataset, achieving an AUROC of 77.1% and a balanced accuracy of 83.8%, surpassing current benchmarks. These findings underscore Cinepro's promise in advancing foundation models for weakly labeled ultrasound data.

Mohamed Harmanani, Paul F. R. Wilson, Fahimeh Fooladgar et al.

PURPOSE: Deep learning methods for classifying prostate cancer (PCa) in ultrasound images typically employ convolutional networks (CNNs) to detect cancer in small regions of interest (ROI) along a needle trace region. However, this approach suffers from weak labelling, since the ground-truth histopathology labels do not describe the properties of individual ROIs. Recently, multi-scale approaches have sought to mitigate this issue by combining the context awareness of transformers with a CNN feature extractor to detect cancer from multiple ROIs using multiple-instance learning (MIL). In this work, we present a detailed study of several image transformer architectures for both ROI-scale and multi-scale classification, and a comparison of the performance of CNNs and transformers for ultrasound-based prostate cancer classification. We also design a novel multi-objective learning strategy that combines both ROI and core predictions to further mitigate label noise. METHODS: We evaluate 3 image transformers on ROI-scale cancer classification, then use the strongest model to tune a multi-scale classifier with MIL. We train our MIL models using our novel multi-objective learning strategy and compare our results to existing baselines. RESULTS: We find that for both ROI-scale and multi-scale PCa detection, image transformer backbones lag behind their CNN counterparts. This deficit in performance is even more noticeable for larger models. When using multi-objective learning, we can improve performance of MIL, with a 77.9% AUROC, a sensitivity of 75.9%, and a specificity of 66.3%. CONCLUSION: Convolutional networks are better suited for modelling sparse datasets of prostate ultrasounds, producing more robust features than transformers in PCa detection. Multi-scale methods remain the best architecture for this task, with multi-objective learning presenting an effective way to improve performance.

Matin Daghyani, Lyuyang Wang, Nima Hashemi et al.

Purpose: Echocardiographic interpretation requires video-level reasoning and guideline-based measurement analysis, which current deep learning models for cardiac ultrasound do not support. We present EchoAgent, a framework that enables structured, interpretable automation for this domain. Methods: EchoAgent orchestrates specialized vision tools under Large Language Model (LLM) control to perform temporal localization, spatial measurement, and clinical interpretation. A key contribution is a measurement-feasibility prediction model that determines whether anatomical structures are reliably measurable in each frame, enabling autonomous tool selection. We curated a benchmark of diverse, clinically validated video-query pairs for evaluation. Results: EchoAgent achieves accurate, interpretable results despite added complexity of spatiotemporal video analysis. Outputs are grounded in visual evidence and clinical guidelines, supporting transparency and traceability. Conclusion: This work demonstrates the feasibility of agentic, guideline-aligned reasoning for echocardiographic video analysis, enabled by task-specific tools and full video-level automation. EchoAgent sets a new direction for trustworthy AI in cardiac ultrasound.

Jiayun Luo, Wan-Cyuan Fan, Lyuyang Wang et al.

Large Vision Language Models (LVLMs) have recently emerged as powerful architectures capable of understanding and reasoning over both visual and textual information. These models typically rely on two key components: a Vision Transformer (ViT) and a Large Language Model (LLM). ViT encodes visual content into a sequence of image tokens and serves as the perceptual front-end -- the eyes of the model. In contrast, the LLM interprets these tokens to perform high-level reasoning, generates responses, and functions as the cognitive core -- the brain of the model. However, it remains unclear which visual tokens contribute most significantly to understanding and reasoning, and how effectively these signals are propagated from ViT to the LLM. While most existing works have focused on identifying attention sinks, low-semantic tokens receiving disproportionately high attention, within the LLM, we shift the focus to the vision encoder by identifying a class of high-norm visual tokens from ViT, referred to as ViT attention sinks -- a problem that has been rarely studied but is indeed very important for LVLMs. Our findings show that these ViT sinks encapsulate high-level semantic concepts from images, allowing the LLM to perform more effective understanding and reasoning. Despite their importance, these sink tokens are often overlooked in existing LVLM architectures. To explore their contribution, we present both qualitative and quantitative analyses of the information embedded in these sink tokens. We also propose both training-free and training-based approaches to better leverage how this information is interpreted by the LLM, and to what extent. By explicitly utilizing these tokens, we demonstrate substantial improvements across a range of LVLMs and visual reasoning tasks, highlighting the untapped potential of ViT attention sinks in enhancing visual reasoning.

Diane Kim, Minh Nguyen Nhat To, Sherif Abdalla et al.

Coronary angiography remains the gold standard for diagnosing Acute Coronary Syndrome (ACS). However, its resource-intensive and invasive nature can expose patients to procedural risks and diagnostic delays, leading to postponed treatment initiation. In this work, we introduce TREAT-Net, a multimodal deep learning framework for ACS treatment prediction that leverages non-invasive modalities, including echocardiography videos and structured clinical records. TREAT-Net integrates tabular-guided cross-attention to enhance video interpretation, along with a late fusion mechanism to align predictions across modalities. Trained on a dataset of over 9000 ACS cases, the model outperforms unimodal and non-fused baselines, achieving a balanced accuracy of 67.6% and an AUROC of 71.1%. Cross-modality agreement analysis demonstrates 88.6% accuracy for intervention prediction. These findings highlight the potential of TREAT-Net as a non-invasive tool for timely and accurate patient triage, particularly in underserved populations with limited access to coronary angiography.

Ang Nan Gu, Michael Tsang, Hooman Vaseli et al.

Computer-aided diagnosis systems must make critical decisions from medical images that are often noisy, ambiguous, or conflicting, yet today's models are trained on overly simplistic labels that ignore diagnostic uncertainty. One-hot labels erase inter-rater variability and force models to make overconfident predictions, especially when faced with incomplete or artifact-laden inputs. We address this gap by introducing a novel framework that brings uncertainty back into the label space. Our method leverages neural network training dynamics (NNTD) to assess the inherent difficulty of each training sample. By aggregating and calibrating model predictions during training, we generate uncertainty-aware pseudo-labels that reflect the ambiguity encountered during learning. This label augmentation approach is architecture-agnostic and can be applied to any supervised learning pipeline to enhance uncertainty estimation and robustness. We validate our approach on a challenging echocardiography classification benchmark, demonstrating superior performance over specialized baselines in calibration, selective classification, and multi-view fusion.

Nooshin Maghsoodi, Sarah Nassar, Paul F R Wilson et al.

Objective: Electrocardiograms (ECGs) play a crucial role in diagnosing heart conditions; however, the effectiveness of artificial intelligence (AI)-based ECG analysis is often hindered by the limited availability of labeled data. Self-supervised learning (SSL) can address this by leveraging large-scale unlabeled data. We introduce PhysioCLR (Physiology-aware Contrastive Learning Representation for ECG), a physiology-aware contrastive learning framework that incorporates domain-specific priors to enhance the generalizability and clinical relevance of ECG-based arrhythmia classification. Methods: During pretraining, PhysioCLR learns to bring together embeddings of samples that share similar clinically relevant features while pushing apart those that are dissimilar. Unlike existing methods, our method integrates ECG physiological similarity cues into contrastive learning, promoting the learning of clinically meaningful representations. Additionally, we introduce ECG- specific augmentations that preserve the ECG category post augmentation and propose a hybrid loss function to further refine the quality of learned representations. Results: We evaluate PhysioCLR on two public ECG datasets, Chapman and Georgia, for multilabel ECG diagnoses, as well as a private ICU dataset labeled for binary classification. Across the Chapman, Georgia, and private cohorts, PhysioCLR boosts the mean AUROC by 12% relative to the strongest baseline, underscoring its robust cross-dataset generalization. Conclusion: By embedding physiological knowledge into contrastive learning, PhysioCLR enables the model to learn clinically meaningful and transferable ECG eatures. Significance: PhysioCLR demonstrates the potential of physiology-informed SSL to offer a promising path toward more effective and label-efficient ECG diagnostics.

Nima Kondori, Hanwen Liang, Hooman Vaseli et al.

Synthetic data generation represents a significant advancement in boosting the performance of machine learning (ML) models, particularly in fields where data acquisition is challenging, such as echocardiography. The acquisition and labeling of echocardiograms (echo) for heart assessment, crucial in point-of-care ultrasound (POCUS) settings, often encounter limitations due to the restricted number of echo views available, typically captured by operators with varying levels of experience. This study proposes a novel approach for enhancing clinical diagnosis accuracy by synthetically generating echo views. These views are conditioned on existing, real views of the heart, focusing specifically on the estimation of ejection fraction (EF), a critical parameter traditionally measured from biplane apical views. By integrating a conditional generative model, we demonstrate an improvement in EF estimation accuracy, providing a comparative analysis with traditional methods. Preliminary results indicate that our synthetic echoes, when used to augment existing datasets, not only enhance EF estimation but also show potential in advancing the development of more robust, accurate, and clinically relevant ML models. This approach is anticipated to catalyze further research in synthetic data applications, paving the way for innovative solutions in medical imaging diagnostics.

Mohamed Harmanani, Paul F. R. Wilson, Minh Nguyen Nhat To et al.

While deep learning methods have shown great promise in improving the effectiveness of prostate cancer (PCa) diagnosis by detecting suspicious lesions from trans-rectal ultrasound (TRUS), they must overcome multiple simultaneous challenges. There is high heterogeneity in tissue appearance, significant class imbalance in favor of benign examples, and scarcity in the number and quality of ground truth annotations available to train models. Failure to address even a single one of these problems can result in unacceptable clinical outcomes.We propose TRUSWorthy, a carefully designed, tuned, and integrated system for reliable PCa detection. Our pipeline integrates self-supervised learning, multiple-instance learning aggregation using transformers, random-undersampled boosting and ensembling: these address label scarcity, weak labels, class imbalance, and overconfidence, respectively. We train and rigorously evaluate our method using a large, multi-center dataset of micro-ultrasound data. Our method outperforms previous state-of-the-art deep learning methods in terms of accuracy and uncertainty calibration, with AUROC and balanced accuracy scores of 79.9% and 71.5%, respectively. On the top 20% of predictions with the highest confidence, we can achieve a balanced accuracy of up to 91%. The success of TRUSWorthy demonstrates the potential of integrated deep learning solutions to meet clinical needs in a highly challenging deployment setting, and is a significant step towards creating a trustworthy system for computer-assisted PCa diagnosis.

Sana Ayromlou, Teresa Tsang, Purang Abolmaesumi et al.

In real-world clinical settings, traditional deep learning-based classification methods struggle with diagnosing newly introduced disease types because they require samples from all disease classes for offline training. Class incremental learning offers a promising solution by adapting a deep network trained on specific disease classes to handle new diseases. However, catastrophic forgetting occurs, decreasing the performance of earlier classes when adapting the model to new data. Prior proposed methodologies to overcome this require perpetual storage of previous samples, posing potential practical concerns regarding privacy and storage regulations in healthcare. To this end, we propose a novel data-free class incremental learning framework that utilizes data synthesis on learned classes instead of data storage from previous classes. Our key contributions include acquiring synthetic data known as Continual Class-Specific Impression (CCSI) for previously inaccessible trained classes and presenting a methodology to effectively utilize this data for updating networks when introducing new classes. We obtain CCSI by employing data inversion over gradients of the trained classification model on previous classes starting from the mean image of each class inspired by common landmarks shared among medical images and utilizing continual normalization layers statistics as a regularizer in this pixel-wise optimization process. Subsequently, we update the network by combining the synthesized data with new class data and incorporate several losses, including an intra-domain contrastive loss to generalize the deep network trained on the synthesized data to real data, a margin loss to increase separation among previous classes and new ones, and a cosine-normalized cross-entropy loss to alleviate the adverse effects of imbalanced distributions in training data.

Mohammad H. Jafari, Christina Luong, Michael Tsang et al.

This paper presents U-LanD, a framework for joint detection of key frames and landmarks in videos. We tackle a specifically challenging problem, where training labels are noisy and highly sparse. U-LanD builds upon a pivotal observation: a deep Bayesian landmark detector solely trained on key video frames, has significantly lower predictive uncertainty on those frames vs. other frames in videos. We use this observation as an unsupervised signal to automatically recognize key frames on which we detect landmarks. As a test-bed for our framework, we use ultrasound imaging videos of the heart, where sparse and noisy clinical labels are only available for a single frame in each video. Using data from 4,493 patients, we demonstrate that U-LanD can exceedingly outperform the state-of-the-art non-Bayesian counterpart by a noticeable absolute margin of 42% in R2 score, with almost no overhead imposed on the model size. Our approach is generic and can be potentially applied to other challenging data with noisy and sparse training labels.

Jianzhe Lin, Ghazal Sahebzamani, Christina Luong et al.

Localization of anatomical landmarks to perform two-dimensional measurements in echocardiography is part of routine clinical workflow in cardiac disease diagnosis. Automatic localization of those landmarks is highly desirable to improve workflow and reduce interobserver variability. Training a machine learning framework to perform such localization is hindered given the sparse nature of gold standard labels; only few percent of cardiac cine series frames are normally manually labeled for clinical use. In this paper, we propose a new end-to-end reciprocal detection and tracking model that is specifically designed to handle the sparse nature of echocardiography labels. The model is trained using few annotated frames across the entire cardiac cine sequence to generate consistent detection and tracking of landmarks, and an adversarial training for the model is proposed to take advantage of these annotated frames. The superiority of the proposed reciprocal model is demonstrated using a series of experiments.

Alireza Mehrtash, Purang Abolmaesumi, Polina Golland et al.

Ensembling is now recognized as an effective approach for increasing the predictive performance and calibration of deep networks. We introduce a new approach, Parameter Ensembling by Perturbation (PEP), that constructs an ensemble of parameter values as random perturbations of the optimal parameter set from training by a Gaussian with a single variance parameter. The variance is chosen to maximize the log-likelihood of the ensemble average ($\mathbb{L}$) on the validation data set. Empirically, and perhaps surprisingly, $\mathbb{L}$ has a well-defined maximum as the variance grows from zero (which corresponds to the baseline model). Conveniently, calibration level of predictions also tends to grow favorably until the peak of $\mathbb{L}$ is reached. In most experiments, PEP provides a small improvement in performance, and, in some cases, a substantial improvement in empirical calibration. We show that this "PEP effect" (the gain in log-likelihood) is related to the mean curvature of the likelihood function and the empirical Fisher information. Experiments on ImageNet pre-trained networks including ResNet, DenseNet, and Inception showed improved calibration and likelihood. We further observed a mild improvement in classification accuracy on these networks. Experiments on classification benchmarks such as MNIST and CIFAR-10 showed improved calibration and likelihood, as well as the relationship between the PEP effect and overfitting; this demonstrates that PEP can be used to probe the level of overfitting that occurred during training. In general, no special training procedure or network architecture is needed, and in the case of pre-trained networks, no additional training is needed.

Amir H. Abdi, Purang Abolmaesumi, Sidney Fels

Unsupervised learning of disentangled representations is an open problem in machine learning. The Disentanglement-PyTorch library is developed to facilitate research, implementation, and testing of new variational algorithms. In this modular library, neural architectures, dimensionality of the latent space, and the training algorithms are fully decoupled, allowing for independent and consistent experiments across variational methods. The library handles the training scheduling, logging, and visualizations of reconstructions and latent space traversals. It also evaluates the encodings based on various disentanglement metrics. The library, so far, includes implementations of the following unsupervised algorithms VAE, Beta-VAE, Factor-VAE, DIP-I-VAE, DIP-II-VAE, Info-VAE, and Beta-TCVAE, as well as conditional approaches such as CVAE and IFCVAE. The library is compatible with the Disentanglement Challenge of NeurIPS 2019, hosted on AICrowd, and achieved the 3rd rank in both the first and second stages of the challenge.

Amir H. Abdi, Mohammad H. Jafari, Sidney Fels et al.

Transthoracic echo is one of the most common means of cardiac studies in the clinical routines. During the echo exam, the sonographer captures a set of standard cross sections (echo views) of the heart. Each 2D echo view cuts through the 3D cardiac geometry via a unique plane. Consequently, different views share some limited information. In this work, we investigate the feasibility of generating a 2D echo view using another view based on adversarial generative models. The objective optimized to train the view-conversion model is based on the ideas introduced by LSGAN, PatchGAN and Conditional GAN (cGAN). The size and length of the left ventricle in the generated target echo view is compared against that of the target ground-truth to assess the validity of the echo view conversion. Results show that there is a correlation of 0.50 between the LV areas and 0.49 between the LV lengths of the generated target frames and the real target frames.

Alireza Mehrtash, William M. Wells, Clare M. Tempany et al.

Fully convolutional neural networks (FCNs), and in particular U-Nets, have achieved state-of-the-art results in semantic segmentation for numerous medical imaging applications. Moreover, batch normalization and Dice loss have been used successfully to stabilize and accelerate training. However, these networks are poorly calibrated i.e. they tend to produce overconfident predictions both in correct and erroneous classifications, making them unreliable and hard to interpret. In this paper, we study predictive uncertainty estimation in FCNs for medical image segmentation. We make the following contributions: 1) We systematically compare cross entropy loss with Dice loss in terms of segmentation quality and uncertainty estimation of FCNs; 2) We propose model ensembling for confidence calibration of the FCNs trained with batch normalization and Dice loss; 3) We assess the ability of calibrated FCNs to predict segmentation quality of structures and detect out-of-distribution test examples. We conduct extensive experiments across three medical image segmentation applications of the brain, the heart, and the prostate to evaluate our contributions. The results of this study offer considerable insight into the predictive uncertainty estimation and out-of-distribution detection in medical image segmentation and provide practical recipes for confidence calibration. Moreover, we consistently demonstrate that model ensembling improves confidence calibration.

Amir H. Abdi, Purang Abolmaesumi, Sidney Fels

Disentangled encoding is an important step towards a better representation learning. However, despite the numerous efforts, there still is no clear winner that captures the independent features of the data in an unsupervised fashion. In this work we empirically evaluate the performance of six unsupervised disentanglement approaches on the mpi3d toy dataset curated and released for the NeurIPS 2019 Disentanglement Challenge. The methods investigated in this work are Beta-VAE, Factor-VAE, DIP-I-VAE, DIP-II-VAE, Info-VAE, and Beta-TCVAE. The capacities of all models were progressively increased throughout the training and the hyper-parameters were kept intact across experiments. The methods were evaluated based on five disentanglement metrics, namely, DCI, Factor-VAE, IRS, MIG, and SAP-Score. Within the limitations of this study, the Beta-TCVAE approach was found to outperform its alternatives with respect to the normalized sum of metrics. However, a qualitative study of the encoded latents reveal that there is not a consistent correlation between the reported metrics and the disentanglement potential of the model.

Amir H. Abdi, Teresa Tsang, Purang Abolmaesumi

Echocardiography (echo) is a common means of evaluating cardiac conditions. Due to the label scarcity, semi-supervised paradigms in automated echo analysis are getting traction. One of the most sought-after problems in echo is the segmentation of cardiac structures (e.g. chambers). Accordingly, we propose an echocardiogram generation approach using generative adversarial networks with a conditional patch-based discriminator. In this work, we validate the feasibility of GAN-enhanced echo generation with different conditions (segmentation masks), namely, the left ventricle, ventricular myocardium, and atrium. Results show that the proposed adversarial algorithm can generate high-quality echo frames whose cardiac structures match the given segmentation masks. This method is expected to facilitate the training of other machine learning models in a semi-supervised fashion as suggested in similar researches.

Zhibin Liao, Hany Girgis, Amir Abdi et al.

Uncertainty of labels in clinical data resulting from intra-observer variability can have direct impact on the reliability of assessments made by deep neural networks. In this paper, we propose a method for modelling such uncertainty in the context of 2D echocardiography (echo), which is a routine procedure for detecting cardiovascular disease at point-of-care. Echo imaging quality and acquisition time is highly dependent on the operator's experience level. Recent developments have shown the possibility of automating echo image quality quantification by mapping an expert's assessment of quality to the echo image via deep learning techniques. Nevertheless, the observer variability in the expert's assessment can impact the quality quantification accuracy. Here, we aim to model the intra-observer variability in echo quality assessment as an aleatoric uncertainty modelling regression problem with the introduction of a novel method that handles the regression problem with categorical labels. A key feature of our design is that only a single forward pass is sufficient to estimate the level of uncertainty for the network output. Compared to the $0.11 \pm 0.09$ absolute error (in a scale from 0 to 1) archived by the conventional regression method, the proposed method brings the error down to $0.09 \pm 0.08$, where the improvement is statistically significant and equivalents to $5.7\%$ test accuracy improvement. The simplicity of the proposed approach means that it could be generalized to other applications of deep learning in medical imaging, where there is often uncertainty in clinical labels.

Amir H. Abdi, Mehran Pesteie, Eitan Prisman et al.

The premorbid geometry of the mandible is of significant relevance in jaw reconstructive surgeries and occasionally unknown to the surgical team. In this paper, an optimization framework is introduced to train deep models for completion (reconstruction) of the missing segments of the bone based on the remaining healthy structure. To leverage the contextual information of the surroundings of the dissected region, the voxel-weighted Dice loss is introduced. To address the non-deterministic nature of the shape completion problem, we leverage a weighted multi-target probabilistic solution which is an extension to the conditional variational autoencoder (CVAE). This approach considers multiple targets as acceptable reconstructions, each weighted according to their conformity with the original shape. We quantify the performance gain of the proposed method against similar algorithms, including CVAE, where we report statistically significant improvements in both deterministic and probabilistic paradigms. The probabilistic model is also evaluated on its ability to generate anatomically relevant variations for the missing bone. As a unique aspect of this work, the model is tested on real surgical cases where the clinical relevancy of its reconstructions and their compliance with surgeon's virtual plan are demonstrated as necessary steps towards clinical adoption.

Mehran Pesteie, Purang Abolmaesumi, Robert Rohling

We introduce a new unsupervised representation learning and visualization using deep convolutional networks and self organizing maps called Deep Neural Maps (DNM). DNM jointly learns an embedding of the input data and a mapping from the embedding space to a two-dimensional lattice. We compare visualizations of DNM with those of t-SNE and LLE on the MNIST and COIL-20 data sets. Our experiments show that the DNM can learn efficient representations of the input data, which reflects characteristics of each class. This is shown via back-projecting the neurons of the map on the data space.

Mohsen Ghafoorian, Alireza Mehrtash, Tina Kapur et al.

Magnetic Resonance Imaging (MRI) is widely used in routine clinical diagnosis and treatment. However, variations in MRI acquisition protocols result in different appearances of normal and diseased tissue in the images. Convolutional neural networks (CNNs), which have shown to be successful in many medical image analysis tasks, are typically sensitive to the variations in imaging protocols. Therefore, in many cases, networks trained on data acquired with one MRI protocol, do not perform satisfactorily on data acquired with different protocols. This limits the use of models trained with large annotated legacy datasets on a new dataset with a different domain which is often a recurring situation in clinical settings. In this study, we aim to answer the following central questions regarding domain adaptation in medical image analysis: Given a fitted legacy model, 1) How much data from the new domain is required for a decent adaptation of the original network?; and, 2) What portion of the pre-trained model parameters should be retrained given a certain number of the new domain training samples? To address these questions, we conducted extensive experiments in white matter hyperintensity segmentation task. We trained a CNN on legacy MR images of brain and evaluated the performance of the domain-adapted network on the same task with images from a different domain. We then compared the performance of the model to the surrogate scenarios where either the same trained network is used or a new network is trained from scratch on the new dataset.The domain-adapted network tuned only by two training examples achieved a Dice score of 0.63 substantially outperforming a similar network trained on the same set of examples from scratch.