University of Houston

More than 1 million skin cancers are detected in the U.S. each year, of which melanoma is the deadliest causing an estimated 8,110 deaths. Melanoma is a malignant tumor of melanocytes, which are found predominantly in the skin but also in the bowel and in the eye. However, if detected at an early stage, skin cancer has a very high cure rate, and it requires rather simple and economical treatment.

Commercially available imaging devices for melanoma detection such as SolarScan and MelaFind have helped to increase accuracy of practitioners. But these devices lack portability because of their size and can only be used inside the hospital environment. The objective of this research is to develop image processing and pattern recognition algorithms for classification of melanoma based on well known diagnostic criteria for smart handheld devices. We have implemented these algorithms for melanoma detection into a library that can run on smart embedded devices which have limited memory and computation power.