For dermatology start-up View Derma we developed a mole inspection camera (multispectral imaging solution). This camera takes multispectral images of skin lesions (10 different wavelengths) that are then classified as malignant or benign.
To develop this classification algorithm we collaborated with the Eindhoven University of Technology. During an internship project, several neural networks were designed and compared. A skin image dataset with annotations was acquired in a clinical study. The networks were trained with different sets of chromophore maps that were computed from the multispectral image stacks. The details and results of this study can be found in the following whitepaper.