Shape-based reconstruction of skin lesion for multispectral Nevoscope using genetic algorithm optimization

Multi-spectral optical Nevoscope is aimed at diagnosing malignant melanoma non-invasively by extracting its two key signatures: spatial distributions of melanin and blood. In this paper, we proposed a shape-based multi-constrained reconstruction algorithm. First, the malignant melanoma is broken into melanin and blood parts which are delineated by two cubic tensor-product B-spline surfaces. Second, the parameters are coded into a genetic algorithm in an effort to recover them through global optimization. Third, some reasonable constraints are incorporated into the genetic algorithm to further stabilize the solution. The algorithm is verified by the Monte Carlo simulation and the results suggest that it is a much better approach than the conventional ones which typically are involved in reconstructing voxel by voxel optical properties.