Feature-based differentiation of malignant melanomas, lesions and healthy skin in multiphoton tomography skin images

Malignant  melanoma  is  a  very  aggressive  tumour  with  the  ability  to  metastasize  at  an  early  stage.  Therefore,  early    detection    is    of    great    importance.    Multiphoton    tomography is a new non-invasive examination method in the clinical diagnosis of skin alterations that can be used for such early diagnosis.In   this   paper,   a   method   for   automated   evaluation   of   multiphoton images of the skin is presented.The  following  features  at  the  cellular  and  subcellular  level  were extracted to differentiate between malignant melanomas, lesions,  and  healthy  skin:  cell  symmetry,  cell  distance,  cell  density,  cell  and  nucleus  contrast,  nucleus  cell  ratio,  and  homogeneity of cytoplasm. The extracted features formed the basis for the subsequent classification. Two feature sets wereused.  The  first  feature  set  included  all  the  above-mentioned  features, while the second feature set included the significantly different  features  between  the three  classes  resulting  from  a  multivariate   analysis   of   variance.   The   classification   wasperformed   by  a   Support   Vector   Machine,   the   k-Nearest   Neighbour algorithm, and Ensemble Learning.The best classification results were obtained with the Support Vector Machine using the first feature set with an accuracy of 52 %  and  79.6 %  for  malignant  melanoma  and  healthy  skin,  respectively.Despite the small number of subjects investigated our results indicate that the proposed automatic method can differentiate malignant  melanoma,  lesions,  and  healthy  skin.  For  future  clinical application, an extended study with more multiphoton images is needed.