Skin cancer detection for Mohs-surgery

Project description

The previous works on the project were mostly focused on finding a way to make the AI learn based on bounding boxes as annotation. However, the amount of background clutter and other “noise” causes inaccuracies that may be avoided through the use of another annotation method. Therefore, the focus of the project has shifted to the following two research questions:

1. How can cancer tumors be detected using labels in the form of polygons? (continuation of current group)
2. How can feedback of the detections of the AI be applied to improve the model? (Active learning)

Context

The MohsA Project is a long-running project, which is a collaboration between Fontys and the dermatologists clinic MohsA. MohsA Skin Center treats various skin conditions. Named after "Mohs Academy," it works with University Hospitals to train medical professionals. The center employs dermatologists, nurses, skin therapists, and plastic surgeons who provide diverse skin treatments. Services include skin cancer care, general dermatology, eczema treatment, varicose vein procedures, and plastic surgery.

MohsA aims for short wait times and thorough patient care. Their team focuses on repairing or improving skin issues effectively. The collaboration between Fontys and MohsA concentrates on the cancer treatment service. More specifically, it aims at improving the cancer detection process in the current flow (see reference image below) by utilizing the technical capabilities an AI model can provide.

Results

The team was able to train couple of AI models, each with their own specifics towards the outputted predictions.

About the project group

The team is working on the project since September 2024. The work strategy is Agile.