Hamid Soltanian-Zadeh

Moein Heidari, Reza Rezaeian, Reza Azad et al.

Implicit Neural Representation (INR), leveraging a neural network to transform coordinate input into corresponding attributes, has recently driven significant advances in several vision-related domains. However, the performance of INR is heavily influenced by the choice of the nonlinear activation function used in its multilayer perceptron (MLP) architecture. To date, multiple nonlinearities have been investigated, but current INRs still face limitations in capturing high-frequency components and diverse signal types. We show that these challenges can be alleviated by introducing a novel approach in INR architecture. Specifically, we propose SL$^{2}$A-INR, a hybrid network that combines a single-layer learnable activation function with an MLP that uses traditional ReLU activations. Our method performs superior across diverse tasks, including image representation, 3D shape reconstruction, and novel view synthesis. Through comprehensive experiments, SL$^{2}$A-INR sets new benchmarks in accuracy, quality, and robustness for INR. Our Code is publicly available on~\href{https://github.com/Iceage7/SL2A-INR}{\textcolor{magenta}{GitHub}}.

Ehsan Karami, Hamid Soltanian-Zadeh

Knee osteoarthritis (KOA) is a common joint disease that causes pain and mobility issues. While MRI-based deep learning models have demonstrated superior performance in predicting total knee replacement (TKR) and disease progression, their generalizability remains challenging, particularly when applied to imaging data from different sources. In this study, we show that replacing batch normalization with instance normalization, using data augmentation, and applying contrastive loss improves generalization. For training and evaluation, we used MRI data from the Osteoarthritis Initiative (OAI) database, considering sagittal fat-suppressed intermediate-weighted turbo spin-echo (FS-IW-TSE) images as the source domain and sagittal fat-suppressed three-dimensional (3D) dual-echo in steady state (DESS) images as the target domain. The results demonstrated a statistically significant improvement in classification metrics across both domains by replacing batch normalization with instance normalization in the baseline model, generating augmented input views using the Global Intensity Non-linear (GIN) augmentation method, and incorporating a supervised contrastive loss alongside the classification loss to align representations of samples with the same label. The GIN method with contrastive loss performed better than all evaluated single-source domain generalization methods when using 3D instance normalization. Comparing GIN with and without contrastive loss (for both normalization types) showed that adding contrastive loss consistently led to better performance.

Ehsan Karami, Hamid Soltanian-Zadeh

Knee osteoarthritis (KOA) is a common joint disease that causes pain and mobility issues. While MRI-based deep learning models have demonstrated superior performance in predicting total knee replacement (TKR) and disease progression, their generalizability remains challenging, particularly when applied to imaging data from different sources. In this study, we show that replacing batch normalization with instance normalization, using data augmentation, and applying contrastive loss improves generalization. For training and evaluation, we used MRI data from the Osteoarthritis Initiative (OAI) database, considering sagittal fat-suppressed intermediate-weighted turbo spin-echo (FS-IW-TSE) images as the source domain and sagittal fat-suppressed three-dimensional (3D) dual-echo in steady state (DESS) images as the target domain. The results demonstrated a statistically significant improvement in classification metrics across both domains by replacing batch normalization with instance normalization in the baseline model, generating augmented input views using the Global Intensity Non-linear (GIN) augmentation method, and incorporating a supervised contrastive loss alongside the classification loss to align representations of samples with the same label. The GIN method with contrastive loss performed better than all evaluated single-source domain generalization methods when using 3D instance normalization. Comparing GIN with and without contrastive loss (for both normalization types) showed that adding contrastive loss consistently led to better performance.

Hamideh-Sadat Fazael-Ardakani, Hamid Soltanian-Zadeh

Knowledge graph embedding (KGE) methods aim to represent entities and relations in a continuous space while preserving their structural and semantic properties. Quaternion-based KGEs have demonstrated strong potential in capturing complex relational patterns. In this work, we propose QuatE-D, a novel quaternion-based model that employs a distance-based scoring function instead of traditional inner-product approaches. By leveraging Euclidean distance, QuatE-D enhances interpretability and provides a more flexible representation of relational structures. Experimental results demonstrate that QuatE-D achieves competitive performance while maintaining an efficient parameterization, particularly excelling in Mean Rank reduction. These findings highlight the effectiveness of distance-based scoring in quaternion embeddings, offering a promising direction for knowledge graph completion.

Navapat Nananukul, Hamid Soltanian-zadeh, Mohammad Rostami

Automatic semantic segmentation of magnetic resonance imaging (MRI) images using deep neural networks greatly assists in evaluating and planning treatments for various clinical applications. However, training these models is conditioned on the availability of abundant annotated data to implement the end-to-end supervised learning procedure. Even if we annotate enough data, MRI images display considerable variability due to factors such as differences in patients, MRI scanners, and imaging protocols. This variability necessitates retraining neural networks for each specific application domain, which, in turn, requires manual annotation by expert radiologists for all new domains. To relax the need for persistent data annotation, we develop a method for unsupervised federated domain adaptation using multiple annotated source domains. Our approach enables the transfer of knowledge from several annotated source domains to adapt a model for effective use in an unannotated target domain. Initially, we ensure that the target domain data shares similar representations with each source domain in a latent embedding space, modeled as the output of a deep encoder, by minimizing the pair-wise distances of the distributions for the target domain and the source domains. We then employ an ensemble approach to leverage the knowledge obtained from all domains. We provide theoretical analysis and perform experiments on the MICCAI 2016 multi-site dataset to demonstrate our method is effective.

Saman Sotoudeh-Paima, Ata Jodeiri, Fedra Hajizadeh et al.

Age-related macular degeneration (AMD) is the most common cause of blindness in developed countries, especially in people over 60 years of age. The workload of specialists and the healthcare system in this field has increased in recent years mainly due to the prevalence of population aging worldwide and the chronic nature of AMD. Recent developments in deep learning have provided a unique opportunity to develop fully automated diagnosis frameworks. Considering the presence of AMD-related retinal pathologies in varying sizes in OCT images, our objective was to propose a multi-scale convolutional neural network (CNN) capable of distinguishing pathologies using receptive fields with various sizes. The multi-scale CNN was designed based on the feature pyramid network (FPN) structure and was used to diagnose normal and two common clinical characteristics of dry and wet AMD, namely drusen and choroidal neovascularization (CNV). The proposed method was evaluated on a national dataset gathered at Noor Eye Hospital (NEH) and the UCSD public dataset. Experimental results show the superior performance of our proposed multi-scale structure over several well-known OCT classification frameworks. This feature combination strategy has proved to be effective on all tested backbone models, with improvements ranging from 0.4% to 3.3%. In addition, gradual learning has proven to improve performance in two consecutive stages. In the first stage, the performance was boosted from 87.2%+-2.5% to 92.0%+-1.6% using pre-trained ImageNet weights. In the second stage, another performance boost from 92.0%+-1.6% to 93.4%+-1.4% was observed due to fine-tuning the previous model on the UCSD dataset. Lastly, generating heatmaps provided additional proof for the effectiveness of our multi-scale structure, enabling the detection of retinal pathologies appearing in different sizes.

Haikal Abdulah, Benjamin Huber, Sinan Lal et al.

CXR-Net is a two-module Artificial Intelligence pipeline for the quick detection of SARS-CoV-2 from chest X-rays (CXRs). Module 1 was trained on a public dataset of 6395 CXRs with radiologist annotated lung contours to generate masks of the lungs that overlap the heart and large vasa. Module 2 is a hybrid convnet in which the first convolutional layer with learned coefficients is replaced by a layer with fixed coefficients provided by the Wavelet Scattering Transform (WST). Module 2 takes as inputs the patients CXRs and corresponding lung masks calculated by Module 1, and produces as outputs a class assignment (Covid vs. non-Covid) and high resolution heat maps that identify the SARS associated lung regions. Module 2 was trained on a dataset of CXRs from non-Covid and RT-PCR confirmed Covid patients acquired at the Henry Ford Health System (HFHS) Hospital in Detroit. All non-Covid CXRs were from pre-Covid era (2018-2019), and included images from both normal lungs and lungs affected by non-Covid pathologies. Training and test sets consisted of 2265 CXRs (1417 Covid negative, 848 Covid positive), and 1532 CXRs (945 Covid negative, 587 Covid positive), respectively. Six distinct cross-validation models, each trained on 1887 images and validated against 378 images, were combined into an ensemble model that was used to classify the CXR images of the test set with resulting Accuracy = 0.789, Precision = 0.739, Recall = 0.693, F1 score = 0.715, ROC(AUC) = 0.852.

Seyed Ziae Mousavi Mojab, Seyedmohammad Shams, Hamid Soltanian-Zadeh et al.

This paper proposes a novel evolutionary algorithm called Epistocracy which incorporates human socio-political behavior and intelligence to solve complex optimization problems. The inspiration of the Epistocracy algorithm originates from a political regime where educated people have more voting power than the uneducated or less educated. The algorithm is a self-adaptive, and multi-population optimizer in which the evolution process takes place in parallel for many populations led by a council of leaders. To avoid stagnation in poor local optima and to prevent a premature convergence, the algorithm employs multiple mechanisms such as dynamic and adaptive leadership based on gravitational force, dynamic population allocation and diversification, variance-based step-size determination, and regression-based leadership adjustment. The algorithm uses a stratified sampling method called Latin Hypercube Sampling (LHS) to distribute the initial population more evenly for exploration of the search space and exploitation of the accumulated knowledge. To investigate the performance and evaluate the reliability of the algorithm, we have used a set of multimodal benchmark functions, and then applied the algorithm to the MNIST dataset to further verify the accuracy, scalability, and robustness of the algorithm. Experimental results show that the Epistocracy algorithm outperforms the tested state-of-the-art evolutionary and swarm intelligence algorithms in terms of performance, precision, and convergence.

Haikal Abdulah, Benjamin Huber, Sinan Lal et al.

Deep Neural Networks (DNN) are widely used to carry out segmentation tasks in biomedical images. Most DNNs developed for this purpose are based on some variation of the encoder-decoder U-Net architecture. Here we show that Res-CR-Net, a new type of fully convolutional neural network, which was originally developed for the semantic segmentation of microscopy images, and which does not adopt a U-Net architecture, is very effective at segmenting the lung fields in chest X-rays from either healthy patients or patients with a variety of lung pathologies.

Erfan Darzi, Armin Mohammadie-Zand, Hamid Soltanian-Zadeh

Heart rate and blink duration are two vital physiological signals which give information about cardiac activity and consciousness. Monitoring these two signals is crucial for various applications such as driver drowsiness detection. As there are several problems posed by the conventional systems to be used for continuous, long-term monitoring, a remote blink and ECG monitoring system can be used as an alternative. For estimating the blink duration, two strategies are used. In the first approach, pictures of open and closed eyes are fed into an Artificial Neural Network (ANN) to decide whether the eyes are open or close. In the second approach, they are classified and labeled using Linear Discriminant Analysis (LDA). The labeled images are then be used to determine the blink duration. For heart rate variability, two strategies are used to evaluate the passing blood volume: Independent Component Analysis (ICA); and a chrominance based method. Eye recognition yielded 78-92% accuracy in classifying open/closed eyes with ANN and 71-91% accuracy with LDA. Heart rate evaluations had a mean loss of around 16 Beats Per Minute (BPM) for the ICA strategy and 13 BPM for the chrominance based technique.

Mohammad Hadi Mehdizavareh, Sobhan Hemati, Hamid Soltanian-Zadeh

Recently, brain-computer interface (BCI) systems developed based on steady-state visual evoked potential (SSVEP) have attracted much attention due to their high information transfer rate (ITR) and increasing number of targets. However, SSVEP-based methods can be improved in terms of their accuracy and target detection time. We propose a new method based on canonical correlation analysis (CCA) to integrate subject-specific models and subject-independent information and enhance BCI performance. We propose to use training data of other subjects to optimize hyperparameters for CCA-based model of a specific subject. An ensemble version of the proposed method is also developed for a fair comparison with ensemble task-related component analysis (TRCA). The proposed method is compared with TRCA and extended CCA methods. A publicly available, 35-subject SSVEP benchmark dataset is used for comparison studies and performance is quantified by classification accuracy and ITR. The ITR of the proposed method is higher than those of TRCA and extended CCA. The proposed method outperforms extended CCA in all conditions and TRCA for time windows greater than 0.3 s. The proposed method also outperforms TRCA when there are limited training blocks and electrodes. This study illustrates that adding subject-independent information to subject-specific models can improve performance of SSVEP-based BCIs.

Mohammad-Parsa Hosseini, Hamid Soltanian-Zadeh, Kost Elisevich et al.

Developing a Brain-Computer Interface~(BCI) for seizure prediction can help epileptic patients have a better quality of life. However, there are many difficulties and challenges in developing such a system as a real-life support for patients. Because of the nonstationary nature of EEG signals, normal and seizure patterns vary across different patients. Thus, finding a group of manually extracted features for the prediction task is not practical. Moreover, when using implanted electrodes for brain recording massive amounts of data are produced. This big data calls for the need for safe storage and high computational resources for real-time processing. To address these challenges, a cloud-based BCI system for the analysis of this big EEG data is presented. First, a dimensionality-reduction technique is developed to increase classification accuracy as well as to decrease the communication bandwidth and computation time. Second, following a deep-learning approach, a stacked autoencoder is trained in two steps for unsupervised feature extraction and classification. Third, a cloud-computing solution is proposed for real-time analysis of big EEG data. The results on a benchmark clinical dataset illustrate the superiority of the proposed patient-specific BCI as an alternative method and its expected usefulness in real-life support of epilepsy patients.

Mohammad-Parsa Hosseini, Mohammad-Reza Nazem-Zadeh, Dario Pompili et al.

The hippocampus is a seminal structure in the most common surgically-treated form of epilepsy. Accurate segmentation of the hippocampus aids in establishing asymmetry regarding size and signal characteristics in order to disclose the likely site of epileptogenicity. With sufficient refinement, it may ultimately aid in the avoidance of invasive monitoring with its expense and risk for the patient. To this end, a reliable and consistent method for segmentation of the hippocampus from magnetic resonance imaging (MRI) is needed. In this work, we present a systematic and statistical analysis approach for evaluation of automated segmentation methods in order to establish one that reliably approximates the results achieved by manual tracing of the hippocampus.