GS-TransUNet: Integrated 2D Gaussian Splatting and Transformer UNet for Accurate Skin Lesion Analysis

TL;DR

GS-TransUNet is a novel integrated deep learning model combining Gaussian splatting and Transformer UNet for accurate, efficient skin lesion segmentation and classification, outperforming existing models on standard datasets.

Contribution

The paper introduces GS-TransUNet, a new unified model that combines Gaussian splatting with Transformer UNet for simultaneous skin lesion segmentation and classification.

Findings

Outperforms state-of-the-art models on ISIC-2017 and PH2 datasets.

Achieves higher accuracy in skin lesion segmentation and classification.

Demonstrates the effectiveness of integrated multi-task learning in medical imaging.

Abstract

We can achieve fast and consistent early skin cancer detection with recent developments in computer vision and deep learning techniques. However, the existing skin lesion segmentation and classification prediction models run independently, thus missing potential efficiencies from their integrated execution. To unify skin lesion analysis, our paper presents the Gaussian Splatting - Transformer UNet (GS-TransUNet), a novel approach that synergistically combines 2D Gaussian splatting with the Transformer UNet architecture for automated skin cancer diagnosis. Our unified deep learning model efficiently delivers dual-function skin lesion classification and segmentation for clinical diagnosis. Evaluated on ISIC-2017 and PH2 datasets, our network demonstrates superior performance compared to existing state-of-the-art models across multiple metrics through 5-fold cross-validation. Our findings illustrate significant advancements in the precision of segmentation and classification. This integration sets new benchmarks in the field and highlights the potential for further research into multi-task medical image analysis methodologies, promising enhancements in automated diagnostic systems.