SCP: Skin Classifcation Project

Funded by the BMG from 2019 to 2023
Project management: Achim Hekler

Malignant melanoma incidence has increased significantly in the last few decades. If a melanoma has metastasized, it is usually no longer curable. The detection of melanomas in their early stages is therefore critical to their curability. If, however, a harmless benign mole is diagnosed (and possibly treated) as malignant melanoma, this may also lead to - sometimes severe - psychological and physical stress for those affected and to avoidable costs for the health care system.

One approach to improved early skin cancer detection, which we already pursued in the first phase of the Skin Classification Project (SCP), is the evaluation of clinical and / or dermatoscopic images of suspicious skin areas using artificial intelligence. For example, a neural network trained by our working group achieved significantly better diagnostic results in a melanoma classification task in an experimental test environment based on individual image analysis than German dermatologists.

Based on our results, we would now like to establish the basis for a diagnostic assistance system for skin lesions based on such an algorithm that can be tested in a large-scale clinical study for its real benefit in dermatological practice. For this purpose, image data of skin lesions and associated metadata are collected in a database in cooperation with 8 German dermatological clinics. These images wil be used to optimize the algorithms we developed previously. In parallel, we test strategies to achieve universal applicability in different clinics / with different image recording systems and develop workflows for the actual practical application in everyday clinical practice.

to top
powered by webEdition CMS