Fourier-transform infrared imaging and clustering: toward an automated histology of normal colon

Congrès sous l’égide de la Société Française de Génie Biologique et Médical (SFGBM).; National audience; Fourier-transform infrared (FTIR) imaging is currently used as a non-destructive and label free method for analyzing biological specimens. Combined with unsupervised clustering method, this biophotonic approach allows to perform a spectral histopathology of human tissues. However, this method requires the subjective choice of the number of clusters. To overcome this problem, we developed a hierarchical double application of 9 cluster validity indices (CVIs) using K-Means clustering. Applying this approach to FTIR images of normal human colon tissue samples, PBM and SI reveal to be the most efficient indices in retreving the main structures of colon histology. These results suggest that the hierarchical double CVI application is thus a promising method for an automated spectral histology.