# Class-balanced dermoscopic lesion segmentation using MoG-LISA and optimized Swin-UNet via the GM-FDE framework

## TL;DR

This paper introduces a new framework combining MoG-LISA and CB-SwinGMO to improve skin lesion segmentation accuracy and address class imbalance in dermatological images.

## Contribution

A unified framework using morphology-guided augmentation and evolutionary optimization for precise and efficient lesion segmentation.

## Key findings

- The proposed method achieved a DSC of 93.8% and IoU of 91.2% on the SIIM-ISIC dataset.
- Boundary accuracy reached 92.7% with a 4.3-pixel reduction in Hausdorff distance.
- The method maintains accuracy without increasing computational cost during inference.

## Abstract

Automatic skin lesion segmentation is one of the key pivotal tasks in dermatological image processing, with important consequences in early melanoma diagnosis and treatment planning. Nevertheless, issues of extreme class imbalance, morphological variability, and poor boundary delineation still exist in current deep learning-based approaches. This study proposes a unified framework that combines morphology-guided latent interpolation and synthesis for lesion augmentation (MoG-LISA) and CB-SwinGMO (Class-Balanced Swin-UNet optimization using geometric mean-driven feedback evolutionary framework) to address these challenges. MoG-LISA generates high-fidelity synthetic samples in a morphology-aware latent space, enriching underrepresented lesion classes such as melanoma and vascular anomalies. Meanwhile, CB-SwinGMO employs multi-objective evolutionary optimization to adapt Swin-UNet parameters for improved generalization and precise boundary detection. Quantitative results highlight the superior performance of our approach, achieving a dice similarity coefficient (DSC) of 93.8%, IoU of 91.2%, boundary accuracy of 92.7%, and Hausdorff distance reduction up to 4.3 pixels on the SIIM-ISIC dataset.

•Morphology-guided latent augmentation addresses class imbalance in lesion segmentation•CB-SwinGMO improves boundary accuracy using feedback-driven evolutionary optimization•Hybrid deep features enhance dermoscopic lesion representation•Accurate segmentation achieved without increasing inference-time computational cost

Morphology-guided latent augmentation addresses class imbalance in lesion segmentation

CB-SwinGMO improves boundary accuracy using feedback-driven evolutionary optimization

Hybrid deep features enhance dermoscopic lesion representation

Accurate segmentation achieved without increasing inference-time computational cost

Health sciences; Medicine; Dermatology; Health technology