Towards a Visual Distinction of Benign and Tumorous Wound Surface in Epidermolysis Bullosa
摘要
It is investigated if a purely vision-based approach might be able to conduct a discrimination of benign and tumorous Epidermolysis bullosa wound surfaces. Although being based on a critically small dataset, best results obtained are above 0.9 in specificity and precision/recall for the two class discrimination. Results encourage further work in this direction, in particular to continue training data collection to further enhance classification.