20 May 2024 in Cluster Computing
The early and accurate diagnosis of brain tumors is critical for effective treatment planning, with Magnetic Resonance Imaging (MRI) serving as a key tool in the non-invasive examination of such conditions. Despite the advancements in Computer-Aided Diagnosis (CADx) systems powered by deep learning, the challenge of accurately classifying brain tumors from MRI scans persists due to the high variability of tumor appearances and the subtlety of early-stage manifestations. This work introduces a novel adaptation of the EfficientNetv2 architecture, enhanced with Global Attention Mechanism (GAM) and Efficient Channel Attention (ECA), aimed at overcoming these hurdles. This enhancement not only amplifies the model’s ability to focus on salient features within complex MRI images but also significantly improves the classification accuracy of brain tumors. Our approach distinguishes itself by meticulously integrating attention mechanisms that systematically enhance feature extraction, thereby achieving superior performance in detecting a broad spectrum of brain tumors. Demonstrated through extensive experiments on a large public dataset, our model achieves an exceptional high-test accuracy of 99.76%, setting a new benchmark in MRI-based brain tumor classification. Moreover, the incorporation of Grad-CAM visualization techniques sheds light on the model’s decision-making process, offering transparent and interpretable insights that are invaluable for clinical assessment. By addressing the limitations inherent in previous models, this study not only advances the field of medical imaging analysis but also highlights the pivotal role of attention mechanisms in enhancing the interpretability and accuracy of deep learning models for brain tumor diagnosis. This research sets the stage for advanced CADx systems, enhancing patient care and treatment outcomes.
11 January 2025 in Neuroscience
31 March 2025 in Chaos Theory and Applications
Ischemic stroke, a widespread neurological condition with a substantial mortality rate, necessitates accurate delineation of affected regions to enable proper evaluation of patient outcomes. However, such precision is complicated by factors like variable lesion sizes, noise interference, and the overlapping intensity characteristics of different tissue structures. This research addresses these issues by focusing on the segmentation of Diffusion Weighted Imaging (DWI) scans from the ISLES 2022 dataset and conducting a comparative assessment of three advanced deep learning models: the U-Net framework, its U-Net++ extension, and the Attention U-Net. Applying consistent evaluation criteria specifically, Intersection over Union (IoU), Dice Similarity Coefficient (DSC), and recall the Attention U-Net emerged as the superior choice, establishing record high values for IoU (0.8223) and DSC (0.9021). Although U-Net achieved commendable recall, its performance lagged behind that of U-Net++ in other critical measures. These findings underscore the value of integrating attention mechanisms to achieve more precise segmentation. Moreover, they highlight that the Attention U-Net model is a reliable candidate for medical imaging tasks where both accuracy and efficiency hold paramount importance, while U Net and U Net++ may still prove suitable in certain niche scenarios.
30 October 2025 in Journal of Integrative Neuroscience
Background: Accurate and timely segmentation of ischemic stroke lesions from diffusion-weighted imaging (DWI) is crucial for diagnosis and treatment planning. Manual segmentation is labor-intensive, time-consuming, and prone to inter-observer variability. This study aims to develop and validate a novel deep learning framework that overcomes the common trade-off between high segmentation accuracy and the computational efficiency required for practical clinical use. Methods: We developed FasterNet and Attention-Gated UNet (FA-UNet), a hybrid U-Net-based architecture. The model’s design features two key innovations: a computationally efficient FasterNet block at the bottleneck to capture global lesion context and multi-scale attention gates (MSAGs) on the skip connections to adaptively refine features and suppress noise. The model was trained and validated on the public Ischemic Stroke Lesion Segmentation (ISLES) 2022 dataset (n = 250 patients) and its performance was assessed on an independent, private test set of 600 DWI scans from 80 patients. FA-UNet’s performance was benchmarked against several state-of-the-art U-Net variants using the Dice coefficient, Intersection over Union (IoU), sensitivity, and precision as primary outcome measures. Results: On the independent test set (n = 80), the proposed FA-UNet model achieved a Dice coefficient of 0.8676 and an IoU of 0.7584. This performance surpassed all benchmarked architectures, including U-Net, U-Net3plus, and CMU-Net. Compared with the next best performing model, this represents a relative improvement of approximately 1.64% in the Dice score and 1.42% in IoU. Conclusion: The FA-UNet architecture establishes a new state-of-the-art performance benchmark for automated ischemic stroke segmentation. By effectively balancing high accuracy with computational efficiency, it offers a robust, reliable, and clinically viable tool.
18 January 2026 in Journal of Soft Computing and Decision Analytics
Rapid, accurate detection of acute ischemic stroke (AIS) is vital for patient outcomes. This study evaluates YOLOv10 variants using Diffusion-Weighted Imaging (DWI) data from the ISLES 2022 dataset to identify a model that optimally balances diagnostic accuracy, speed, and computational efficiency for clinical deployment. Methods: YOLOv10 variants were trained on 1652 preprocessed DWI images using transfer learning and data augmentation. Performance was assessed using precision, recall, mAP50, and inference time to evaluate each model comprehensively. Results: The YOLOv10l variant achieved the best performance with 0.933 precision, 0.783 recall, and 0.887 mAP50. YOLOv10n was fastest (0.8 ms) but less accurate, highlighting a trade-off between speed and precision. A key challenge across models was the reliable detection of smaller, less distinct stroke lesions. These findings underscore the importance of balancing accuracy and efficiency. YOLOv10l is optimal for high-precision hospital diagnostics, whereas YOLOv10n suits real-time, resource-limited settings. Future work must focus on improving the detection of small lesions to enhance clinical utility and ensure comprehensive diagnostic support. YOLOv10 is a powerful tool for automated stroke detection, with the YOLOv10l model providing a balanced solution for clinical use. Integrating such models can significantly enhance diagnostic workflows, aid clinical decisions, and ultimately improve patient outcomes in stroke care.