Mohammad Sakib (Graduate Student Member, IEEE) obtained his B.Sc. degree in Electrical and Electronic Engineering from the University of Liberal Arts Bangladesh (ULAB) in 2022. He is currently serving as a Senior Lecturer in the Department of Electronics and Communication Engineering at the Institute of Science Trade & Technology (ISTT). Additionally, he works as a Research Assistant (RA) in the Department of Electrical and Electronic Engineering at Independent University, Bangladesh (IUB). Previously, he served as a Teaching Assistant (TA) in the Department of EEE at ULAB and held the position of President of the ULAB Electronics and Robotics Club during his undergraduate studies.

Mr. Sakib is an active IEEE Student Member and currently serves as a Mentor for the IEEE ISTT Student Branch, particularly supporting the event management and logistics team. He has also served as a reviewer for three IEEE international conferences held in India, Turkey, and France.To date, he has published over 15 research papers in peer-reviewed journals and international conference proceedings. In recognition of his outstanding research contributions, he received a $2,500 USD Publication Award jointly from ULAB and IUB for the 2023–2025 academic period. His research interests include computer vision, wearable devices, biomedical signal and image processing, machine learning, deep learning, and the Internet of Things (IoT).

Beyond his academic and research pursuits, he enjoys exploring nature whether along coastal regions, in forests, or in mountainous landscapes. He has a keen interest in discovering diverse cuisines and local food cultures and is passionate about sports, particularly cricket and Football.

This paper presents CATF-Swin, a novel dual-branch CNN–Transformer framework for automated brain tumor diagnosis from MRI and CT images. The proposed model performs abnormality screening, tumor segmentation, and tumor classification in an end-to-end manner by combining EfficientNet-B4, Feature Pyramid Networks, dual Swin Transformers, and a bidirectional cross-attention fusion module. Evaluations on eight public datasets and a clinical dataset of 7,000 expert-annotated MRI images demonstrate excellent performance, achieving over 99% classification accuracy, 99% sensitivity and specificity, 98% IoU, and 99% ROC/AUC, while outperforming or matching existing state-of-the-art methods. 

This study presents an end-to-end dual-branch CNN–Transformer framework for automated breast tumor segmentation and classification using breast ultrasound (BUS) images. By combining EfficientNet-B4 and Vision Transformer (ViT) with multi-scale cross-attention feature fusion and multi-task learning, the proposed model effectively captures both local and global image features. Experimental results on eight public BUS datasets demonstrate outstanding performance, achieving an average Dice score of 99.46%, IoU of 99.31%, segmentation accuracy of 99.60%, and classification accuracy of over 99%. In addition, calibration analysis confirms that the model provides reliable and clinically consistent confidence estimates, making it a promising tool for accurate breast cancer diagnosis.