Integrated image analysis system for the early recognition of malignant melanoma

The project aims at the creation of a computer image analysis work station for clinical application. The workstation will receive Epiluminescence Microscopy (ELM) digital images of pigmented skin lesions and information pertaining to a patient. This input will be converted into a proposed diagnosis of a lesion as malignant, benign or dysplastic, with an associated error probability. The connection between camera and computer will initially remain in an open loop so that the camera will produce images with pre-set parameters for the illumination. Should it prove useful the loop will be closed to control the imaging parameters from the computer software. The system will be subject to a large scale real-life test using up to 3,000 actual pigmented skin lesions during the course of the project. Annually, the Department of Dermatology at the University of Vienna Medical School (DDW) processes more than 10,000 requests from patients to study pigmented skin lesions. 15 % of these will be selected each year and studied by the computer-based system. This will then be the source material for a thorough assessment of the performance of the computerized approach. It will also be the basis for improvements of this approach, fine-tuning the imaging and image analysis parameters. These tests will initially be performed by data collection in the field at DDW, the analysis will be performed in the laboratory at Institute for Computer Graphics (ICG) at TU Graz. During the second year the tests will be performed entirely in the clinical environment at DDUV, with the support of the non-medical partners. For imaging technology this will be the Forschungszentrum Seibersdorf, for computer issues this is ICG.