Towards an Interpretable Classifier for Characterization of Endoscopic Mayo Scores in Ulcerative Colitis Using Raman Spectroscopy

Abstract

Ulcerative colitis (UC) is one of the main types of chronic inflammatory diseases that affect the bowel, but its pathogenesis is yet to be completely defined. Assessing the disease activity of UC is vital for developing a personalized treatment. Conventionally, the assessment of UC is performed by colonoscopy and histopathology. However, conventional methods fail to retain biomolecular information associated to the severity of UC and are solely based on morphological characteristics of the inflamed colon. Furthermore, assessing endoscopic disease severity is limited by the requirement for experienced human reviewers. Therefore, this work presents a nondestructive biospectroscopic technique, for example, Raman spectroscopy, for assessing endoscopic disease severity according to the four-level Mayo subscore. This contribution utilizes multidimensional Raman spectroscopic data to generate a predictive model for identifying colonic inflammation. The predictive modeling of the Raman spectroscopic data is performed using a one-dimensional deep convolutional neural network (1D-CNN). The classification results of 1D-CNN achieved a mean sensitivity of 78% and a mean specificity of 93% for the four Mayo endoscopic scores. Furthermore, the results of the 1D-CNN are interpreted by a first-order Taylor expansion, which extracts the Raman bands important for classification. Additionally, a regression model of the 1D-CNN model is constructed to study the extent of misclassification and border-line patients. The overall results of Raman spectroscopy with 1D-CNN as a classification and regression model show a good performance, and such a method can serve as a complementary method for UC analysis.

Fourier-Transform Infra-Red Microspectroscopy Can Accurately Diagnose Colitis and Assess Severity of Inflammation

“…Unlike infra-red spectroscopy, Raman spectroscopy uses laser energy to measure non-elastic light scattering to produce biochemical information. It has been shown to distinguish tissue [ 25 ] and plasma samples [ 26 ] from IBD and healthy controls, Crohn’s disease from ulcerative colitis [ 27 , 28 ], as well as active inflammation from mucosal healing based on varying profiles in lipids, phosphatidylcholines, myoglobin, and carotenoids [ 29 , 30 , 31 ]. Fibreoptic Raman probes have also been developed with the potential for application in point-of-care endoscopy, as the spectra acquired are not obscured by the presence of water in the colon [ 32 ].…”

Trends in artificial intelligence, machine learning, and chemometrics applied to chemical data

“…In their approach, the accuracy of Raman spectroscopy for bacterial species determination was improved and enabled rapid identification of Arcobacter . A further challenge in assessing endoscopic disease severity in Ulcerative colitis (UC) patients using Raman spectroscopy was explored by Kirchberger‐Tolstik et al 53 . In this study, the endoscopic disease severity evaluation was performed according to the four Mayo subscores.…”

Deep Learning for Raman Spectroscopy: A Review

“…Same as pre-processing, 1D CNNs also play a very important role in Raman spectral classification. For example, to distinguish human and animal blood, Dong et al, used a simplified network modified from LeNet-5 architecture with only two convolutional layers for feature extraction followed by one fully-connected layer for classification, which achieved an accuracy of 96.33% [37]; to detect prostate cancer, Lee et al, used another 1D CNN for Raman spectra from extracellular vesicles (EVs) [38]; to assess the disease activity of ulcerative colitis (UC), Kirchberger-Tolstik et al, used a 1D CNN as well and reached a mean sensitivity of 78% and a mean specificity of 93% for the four Mayo endoscopic scores [39]. Besides, an accuracy of 93% has been reached for classifying lymph node carcinoma of the prostate (LNCaP), prostate cancer cell line (PC3), and red blood cell (RBC) and platelet.…”

“…Examples of typical deep learning applications for Raman spectroscopy. Dong et al[37], Lee et al[38], Kirchberger-Tolstik et al[39], Maruthamuthu et al[40], Cheng et al[41], Fan et al[42], Fu et al[43], Houston et al[44], Ho et al[45], Ding et al[46], Chen et al[47], Saifuzzaman et al[48], Pan et al[49,50], Sohn et al[51], Yu et al[52], Thrift and Ragan[53], and Zhang et al[54] …”