This paper proposes a new approach to process spatially resolved bimodal spectroscopic data applied to mouse skin precancerous stages non-invasive diagnosis. In this field, the development of efficient methods of extraction of discriminant features followed by supervised classification step is of a crucial importance.Our idea is to exploit the spatial resolution dimensions (3 in the present study) to take advantage from the fact that each collecting fiber provides a complementary piece of information for the discrimination. The purpose of acquiring data at three different source-to-detector distances is to combine complementary spatially resolved information i.e. data from three sources on one single skin spot. Such spatial resolution allows to probe skin at three different mean depths in order to get knowledge from the various layers of skin dermis and epidermis.The first step of our method proposed here consists in applying (i) 2D discrete cosine transform to extract discriminant spectral features from autofluorescence excitation emission matrices and (ii) mutual information to select relevant features from diffuse reflectance spectra. In thesecond step, these feature sets, which capture information from each collecting fiber (among 3) per modality, is independently classified by one versus all decomposition involving support vector machines, creating a multiclassifier system. In the last step, classification results are fused using the first combination rule of belief function theory to produce one final classification. The proposed method improves overall classification accuracy over independent classifiers. The best recognition rate obtained is 86.1%.

EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Main/Merge

HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000492 | SxmlIndent | more

HfdSelect -h $EXPLOR_AREA/Data/Main/Merge/biblio.hfd -nk 000492 | SxmlIndent | more

{{Explor lien
   |wiki=    Wicri/Lorraine
   |area=    InforLorV4
   |flux=    Main
   |étape=   Merge
   |type=    RBID
   |clé=     Hal:hal-01246599
   |texte=   Multidimensional spectroscopic data fusion improves precancerous tissue discrimination based on spatially resolved autofluorescence and diffuse reflectance spectroscopy
}}