Hyperspectral imaging and spectral classification for assisting in vivo diagnosis of melanoma precursors: preliminary results obtained from mice

Dysplastic nevi are at high risk of converting to melanoma. Early detection of signs of skin dysplasias at their curable stage, is crucial. We present an integrated system combining a Hyperspectral Camera system (HyCam), with spectral classification algorithms, developed for the in vivo and objective discrimination between non dysplastic, dysplastic nevi and melanomas. A melanoma animal model was developed in order to monitor the process of melanoma development. Reference spectra were collected from normal skin areas, which were compared with the spectra obtained from the lesions using the Spectral Angle Mapper (SAM) algorithm. A pseoudocolor map was created with different colors representing various degrees of similarity between test and reference spectra. The system's predictions ware validated with biopsy/histology. Both sensitivity and specificity were found to be 77%, highlighting the potential of the method to improve diagnostic and screening accuracies.