Deep learning architectures for medical image segmentation: an organized analysis of CNN-based models and uses

Cherifa Abdellaoui (University of Boumerdes)
Samia Belkacem (University of Boumerdes)
Noureddine Messaoudi (University of Boumerdes)

Bulletin of Electrical Engineering and Informatics
Vol 15, No 1: February 2026

Numerous techniques, especially those based on deep learning (DL), have been developed and applied to a wide range of tasks, including image recognition, classification, object detection, and image segmentation, as a result of extensive research in the field of image processing. Image processing has become crucial in the medical field, with segmentation emerging as a crucial method for organ identification, disease detection, and abnormality analysis in medical images. Convolutional neural networks (CNNs), one of the many approaches, have recently demonstrated great promise in resolving intricate problems associated with medical image analysis because of their capacity to automatically learn hierarchical features. In this review, we discuss recent developments in deep CNNs for medical image segmentation. The architectures and features of the most popular CNN-based models are examined, along with the different publicly accessible medical imaging datasets that are used in studies and the evaluation metrics that are frequently used to gauge segmentation performance and accuracy, also the advantages and disadvantages of each one. In addition, we look at comparative research and the shortcomings of existing approaches, offering suggestions for future lines of clinical relevance.