Skin Cancer Screening

TNT members involved in this project:
Prof. Dr.-Ing. Bodo Rosenhahn
Show all

Each year more than 200.000 people develop skin cancer in germany alone1. Therefore an affordable and reliable screening process becomes all the more important. At present detection and diagnosis of skin lesions is handled mostly manually by experienced dermatologists using a plethora of simple criteria, like the ABCDE principle2.

Our goal is to develop an automatic process for diagnostic support of skin lesions based on close-up views of the patient in question. The process should perform three main tasks:

  • Preprocessing and removal of irrelevant information(e.g. body hair)
  • Detection and segmentation of candidate regions
  • Provide diagnostic support for the located regions

Our framework incorporates different techniques from the computer-vision and machine-learning community and our own custom algorithms:

  • Object removal by texture synthesis
  • Supervised classification based on the boosting framework
  • Automatic evaluation of diagnostic indices based on color, texture, shape and structural features
  • ...