Nail apparatus melanoma (NAM) is a malign melanoma with relatively high incidence among Japanese compared with melanomas at other sites. The prognosis is poor because of the difficulty of diagnosis. In this study, we developed an algorithm to distinguish malign melanomas from benign melanomas based on images, captured uninvasively by dermoscopic examination. This algorithm extracts common waveform patterns in images of melanonychia striata by use of a genetic algorithm (GA), and carries out diagnosis based on common patterns by using linear discriminant analysis. In the present study, images were captured from 28 cases of NAM, including 9 malign melanomas and 19 benign melanomas by dermoscopy. The algorithm was applied to NAM these NAM images, and the accuracy of diagnosis was found to be 80.0%. Moreover, we compared the common patterns extracted from malign and benign melanomas and confirmed that the common waveform pattern extracted from malign melanomas shows marked variations compared with that from benign melanomas.