AI-powered computational approaches for enhanced skin lesion image analysis

Programme

Ernst Mach Mobility Grant worldwide

Project lead

Danube Private University (DPU, AT); Assoc.-Prof. Priv.-Doz. Amirreza Mahbod, PhD

Researchers involved at DPU

Univ.-Prof. Priv.-Doz. Dr. med. Ramona Woitek, PhD

Saeed Shakurimoghaddam, MSc (Incoming Visiting Researcher)

Abstract

Skin cancer is one of the most common cancer types worldwide, with a steadily increasing incidence rate. Automatic dermoscopic image analysis has gained significant research interest in recent years. While many computerized approaches have been proposed, deep learning–based methods using convolutional neural networks (CNNs) and Vision Transformers (ViTs) are among the most promising. More recently, the advent of medical and non-medical foundation models, as well as multimodal medical data analysis pipelines, has opened new research directions in medical image analysis.

In this project, we aim to enhance model performance for skin lesion image analysis, with a particular focus on segmentation and classification tasks. This will be achieved through the development of new deep learning–based models, as well as by adapting and fine-tuning medical and non-medical foundation models, including skin lesion–specific models such as PanDerm. In addition, we plan to integrate other medical data sources, such as clinical metadata, together with skin lesion images to further improve model performance.