A SURVEY ON DIAGNOSIS OF SKIN LESIONS USING DEEP LEARNING

Read full text Summarise with VIRAT
Cited by: 0
References: 9

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Medical image segmentation has witnessed remarkable progress with the development of lightweight and enhanced deep learning architectures that balance computational efficiency with segmentation accuracy.This paper presents a comprehensive review of four state-of-the-art approaches published in 2025, focusing on lightweight and enhanced architectures for diverse medical imaging applications.The reviewed methods include DAMMD-Net, a lightweight DeepLabV3+-based network for skin lesion segmentation with attention mechanisms; DE-UNeXt, a dual-encoder UNeXt architecture with feature compensation for intracranial hemorrhage segmentation; along with IS-UNeXt and MQAUNet from previous analyses.We systematically analyze architectural innovations, methodological contributions, experimental results across benchmark datasets (ISIC2018, PH2, HBU CH, BCIHM, BHSD), and computational efficiency metrics.Our review identifies emerging paradigms including dual-encoder structures, multi-scale feature fusion, attention mechanisms for local feature enhancement, and hybrid CNN-Transformer designs that achieve superior segmentation performance with minimal parameter overhead.The comparative analysis reveals that lightweight architectures can achieve state-of-the-art performance while maintaining real-time inference capabilities, making them suitable for clinical deployment in resource-constrained environments.

Fields of Study & Subjects

Medical Image Segmentation Techniques Cutaneous Melanoma Detection and Management Advanced Neural Network Applications

Publication details

Journal / source
International Research Journal of Modernization in Engineering Technology and Science
Publication date
28 March 2026
Language
EN