Multispectral and autofluorescence, human and mouse

Published: 02-03-2021| Version 1 | DOI: 10.17632/bcbdcn3t6c.1
Thomas Bazin


Gastric inflammation is a major risk factor for gastric cancer. Current endoscopic methods are not capable to efficiently detect and characterize gastric inflammation, leading to a sub-optimal patients’ care. New non-invasive methods are needed. Reflectance mucosal light analysis is of particular interest in this indication. The aim of our study was to analyze the reflectance light and specific autofluorescence signals, both in humans and in a mouse model of gastritis. We recruited patients undergoing gastroendoscopic procedure during which reflectance was analysed with a multispectral camera. In parallel, the gastritis mouse model of Helicobacter pylori infection was used to investigate reflectance from ex vivo gastric samples using a spectrometer. In both cases, autofluorescence signals were measured using a confocal microscope. In gastritis patients, reflectance modifications were significant in near-infrared spectrum, with a decrease between 610 and 725 nm and an increase between 750 and 840 nm. Autofluorescence was also modified, showing variations around 550 nm of emission. In H. pylori infected mice developing gastric inflammatory lesions, we observed significant reflectance modifications 18 months after infection, with increased intensity between 617 and 672 nm. Autofluorescence was significantly modified after 1, 3 and 6 months around 550 and 630 nm. Both in human and in mouse, these reflectance data can be considered as biomarkers and accurately predicted inflammatory state.


Steps to reproduce

Multispectral human : Each gastroendoscopic video, acquired during endoscopy procedure, was divided in images NIR and VIS with a resolution of 512x256 for VIS sensor and 409x216 for NIR sensor. Each image was filtered by a 7x7 low-pass filter to eliminate noise and we then selected the pixels arranged on a regular grid by taking 1 pixel out of 7 horizontally and vertically. In order to avoid under- or over-exposure, only pixels with intensity values between 128 and 512 have been kept. We applied a L1 type normalization for each pixel: all the values were expressed as a ratio of the total intensity per pixel. We reduced data dimensionality using a principal component analysis and then used a simple SVM classifier with a linear kernel. We have chosen to use a linear kernel to avoid the risk of overfitting. In order to assess the diagnosis properties of multispectral data for gastritis, we used a “leave-one-out” method: we trained the SVM classifier with the multispectral data of all but one patient and tested diagnostic properties of the test on this “left out” patient. We repeated this test construction as many times as they were patients. This method constructed artificially a validation cohort from our data, the test being slightly different each time. We described this specific classification pipeline in a previous study (5). Diagnostic properties of all the tests were aggregated on a confusion matrix and represented on ROC curve. Autofluorescence human : For each acquisition we obtained a matrix of 211 intensities values for 211 excitation/emission couples. For each sample the median acquisition number was six. The matrices for one sample were averaged to obtain one matrix for one sample. For one time-point we normalized the matrices so that the sum of the intensities of each matrix was equal. Multispectral mouse : Depending on the size of the samples, between 4 and 8 spectra acquisitions were obtained per sample, with a median of six acquisitions. For one timepoint we normalized the intensities so that the sum of the intensities of each mean acquisition was equal. Autofluorescence mouse : AF was measured as described above for human, using 6 µm-thick dewaxed sections of paraffin blocs from stomach biopsies collection. Statistical analysis : In order to compare in one hand the differences on the reflectance spectra and in another hand on AF spectra between the two groups: gastritis patients vs controls for human and infected vs non-infected for mice, we performed parametric Student t-test after confirmation of data normal distribution using Shapiro test. We applied Benjamini-Hochberg correction on the normalized spectra from spectrometer and from confocal microscope; α risk was set at 0.05. Of note, we corrected AF p-values for the number of couple excitation-emission that varied of more than 5% between gastritis group and controls. Raw data are available as supplementary data.