Multi-modality point spectroscopy: Interests, development and application to in vivo diagnosis of bladder and skin cancers

Abstract

Combining autofluorescence (AF) and Diffuse Reflectance (DR) spectroscopies is supposed to improve diagnosis' accuracy of early stages of cancer (as well as precancerous stages) which is of great clinical importance. For the present study, we developed a bimodal instrumentation combining spatially resolved AF and DR spectroscopies, and evaluated its ability to distinguish between healthy, inflammatory and early stages of cancers in vivo. In order to get such tissue types, we used 2 animal models: a rat bladder orthotopic cancer model and a mice UV-irradiated skin model. The first study shows that combining AF and DR improves both sensitivity and specificity of the diagnosis compared to one modality used alone: Se = 67% (DR alone), 72% (AF alone) increases up to 78% when combining the two modalities. Preliminary results of the second study reveal that some spectroscopic criteria may help quantitative histological analysis in making the difference between acute and precancerous hyperplasia.

Habilitation à diriger des recherches

“…La dernière étape consiste à effectuer une classification supervisée (analyse qualitative) à partir du jeu de données spectrales le plus réduit (ou des CP) en référence à la classification obtenue par l'expertise histologique des lames des échantillons tissulaires biopsiés (propriété de référence). Nous avons évalué l'efficacité de différentes méthodes linéaires hiérarchiques (divisions successives du jeu pour aboutir à la formation d'agrégats -clusters-représentés sous forme de dendrogrammes par exemple) et non-hiérarchiques (méthodes itératives), plus ou moins sensibles au caractère gaussien des distributions de nos jeux de données : classification hiérarchique avec distance de Ward [CI22,CI27], méthode des k plus proches voisins (k -NN), Analyse Discriminante Linéaire (ADL) et Support Vector Machines (SVM) [CI23,RI18,RI16]. Les solutions méthodologiques développées (extraction et sélection de caractéristiques spectrales, réduction de dimension) et les résultats de classification obtenus dans le cadre de deux études in vivo menées sur modèles de cancers de vessie de rats et de pré-cancers de la peau de souris sont publiées respectivement dans [CI21, CI22, CI27] et [CI23,RI18,RI16].…”

Proposal for a Skin Layer-Wise Decomposition Model of Spatially-Resolved Diffuse Reflectance Spectra Based on Maximum Depth Photon Distributions: A Numerical Study

“…Indeed, photons detected at a short/large source-detector (SD) distance will be associated with shallower/deeper trajectories into the biological tissues under evaluation. The literature on this subject points to numerous applications of SR-DRS such as the measurement of blood glucose levels [5,6], hemoglobin oxygen saturation [7] and for pre-cancer and cancer diagnosis: lung [8], cervix [9], colon [10], bladder [11], brain [12], breast [13], oesophagus [14] or pancreas [15]. There are also a large number of applications of SR-DRS for the in vivo characterization of cutaneous pathological states, including melanoma [16][17][18] and carcinoma [19,20], and more generally for the characterization of skin layer optical properties [21,22].…”

“…In this context of skin optical biopsy, point spectroscopy devices exploiting lighttissue interactions have been designed and used in several clinical studies [11,23,24]. The estimation of diagnostically relevant optical parameters from in vivo SR-DR spectra, and especially their evolution around the lesion, can be linked to the tissue pathological state.…”