Highlighting peritumoral areas in human skin cancer biopsies by infrared micro-spectroscopy

Background Fourier transform mid-infrared (FT–IR) microspectroscopy is a label-free optical method based on the interaction between an incident light beam and matter. This vibrational spectroscopy permits to probe the biochemical composition of the analyzed sample and thus gives information about the structure of this sample. Associated with an imaging system, FT–IR microspectroscopy of human tissues can be used as a very sensitive, non-destructive and non-subjective tool for the detection and localization of tumoral areas independently of visual morphology. Thus, FT–IR microimaging has demonstrated potential to provide clinically relevant diagnostic information in oncology [1-3]. The biochemical changes related to carcinogenesis between cancerous and surrounding tissue areas are subtle. As a consequence, IR hyperspectral images need to be processed by powerful digital signal processing and pattern recognition methods in order to highlight these changes [4,5]. To this end, an innovative fuzzy C-means (FCM) clustering-based algorithm was proposed in [6]. The real advantage of FCM is that it introduces the notion of nuance into the clustering of IR image pixels. Consequently, FCM allows considering the progressive transition between non-cancerous tissues and cancer lesions and reveals every nuance of intratumoral heterogeneity [6]. Moreover, the FCM-based algorithm proposed in [6] is fully automatic, i.e. the optimal clustering parameters such as the number of clusters are automatically determined. The main drawbacks with this algorithm are that it is very time consuming and that the transition areas can be difficultly seen. In this work, we thus propose solutions to these problems. Material and methods IR spectral images were acquired on 8 biopsies of formalin-fixed paraffin-embedded human skin carcinomas, squamous cell carcinomas (SCC, n=3) and basal cell carcinomas (BCC, n=5). The samples were selected by the pathologists from the tumor bank of the Pathology Department of the University Hospital of Reims (France). From samples, 10-micron thick slices were cut and mounted on a calcium fluoride (CaF2) (Crystran, Dorset, UK) window for FT–IR imaging without any particular preparation, especially no chemical dewaxing. First adjacent slices (5-μm thick) to those used for FT–IR analysis were stained with hematoxylin and eosin (H&E) for conventional histology. From these slices, the cancer outlines defined by the pathologists were drawn on the photomicrographs. FT–IR hyperspectral images were recorded with a Spectrum Spotlight 300 FT–IR imaging system coupled to a Spectrum one FT–IR spectrometer (Perkin Elmer Life Sciences, France), with a spatial resolution of 6.25 μm and a spectral resolution of 4 cm. Each spectral image, covering a substantial part of the biopsy, consisted of about 30000 spectra. The samples being analyzed without previous chemical dewaxing, the recorded FT–IR hyperspectral image were digitally corrected from paraffin spectral contribution thanks to an automated pre-processing method based on extended multiplicative signal correction (EMSC) [7]. Only the spectral variability of the molecular composition of the tissue is thus retained in the data sets. In order to highlight the different biological structures of the samples from the weak inter-spectra differences, the EMSC-based pre-processed IR spectra were analysed by an upgraded version of the FCM clustering-based algorithm proposed in [6]. The innovations mainly consist in the breaking of the FCM algorithm as soon as the estimated clusters present some uninteresting characteristics and in the limitation of the number of computed FCM. * Correspondence: [email protected] MéDIAN, CNRS FRE 3481 MEDyC, SFR Cap-Santé, Université de Reims Champagne-Ardenne, 51 rue Cognacq-Jay, 51096 Reims, France Full list of author information is available at the end of the article Sebiskveradze et al. Diagnostic Pathology 2013, 8(Suppl 1):S33 http://www.diagnosticpathology.org/content/8/S1/S33