Additional consideration of image features suitable for melanoma diagnosis

Abstract

Melanoma is a type of skin cancer. The diagnosis of melanoma depends on the subjectivity of a doctor. Therefore, a quantitative diagnosis method is required. We have developed a melanoma diagnosis system using a hyperspectral imager (HSI). However, HSI tends to be larger and more expensive than a commercially available digital camera. By selecting some wavelengths suitable for melanoma diagnosis from among the wavelengths measured by the HSI, we think it will result in simplification and cost reduction of HSI. A spectral angle map for the three wavelengths is calculated from data measured by HSI. Then, entropy and some image features are calculated from a spectral angle map. Analysis of a diagnostic performance from the entropy and image features showed that a diagnostic rate of 77.8% was achieved with a combination of 500 nm, 620 nm, and 710 nm.