Dealing with the wave: Automating skin cancer assessment

D. Geijs

  • Promotor: G. Litjens and J. van der Laak
  • Copromotor: A. Amir
  • Graduation year: 2026
  • Radboud University

Abstract

With the growing incidence of BCC and a shortage of pathologists worldwide, there is a clear need for reproducible and efficient diagnostic tools. This thesis aims to develop AI systems that can be integrated into clinical practice, rather than remaining at the research stage. A key challenge in this field is the limited availability of high-quality datasets. To address this, we create a large, publicly available dermatopathology dataset to support the development and validation of clinically relevant AI systems. Using this dataset, we focus on two main diagnostic tasks in BCC: tumor detection and subtype classification. These represent essential components of clinical assessment. To show their practical value, we further explore the use of these algorithms in Mohs micrographic surgery, assessing whether models trained on routine sections can be fine-tuned for frozen sections to support dermatologists during intraoperative decision-making.