Automated analysis of small intestinal lamina propria to distinguish normal, Celiac Disease, and Non-Celiac Duodenitis biopsy images

• DOI: 10.1016/j.cmpb.2022.107320
• Publication Date: 2022-12-19T16:17:14Z
• Authors: Oliver Faust; Simona De Michele; Joel EW Koh; V Jahmunah; Oh Shu Lih; Aditya P Kamath; Prabal Datta Barua; Edward J. Ciaccio; Suzanne K. Lewis; Peter H. Green; Govind Bhagat; U. Rajendra Acharya
• Keywords: Duodenitis; Lamina propria; Enteroscopy

Celiac Disease (CD) is characterized by gluten intolerance in genetically predisposed individuals. High disease prevalence, absence of a cure, and low diagnosis rates make this public health problem. The CD predominantly relies on recognizing characteristic mucosal alterations the small intestine, such as villous atrophy, crypt hyperplasia, intraepithelial lymphocytosis. However, these changes are not entirely specific to overlap with Non-Celiac Duodenitis (NCD) due various etiologies. We investigated whether Artificial Intelligence (AI) models could assist distinguishing normal, CD, NCD (and unaffected individuals) based characteristics intestinal lamina propria (LP). Our method was developed using dataset comprising high magnification biopsy images duodenal LP compartment patients different clinical stages those NCD, individuals lacking an inflammatory disorder (controls). A pre-processing step used standardize enhance acquired images. For normal controls versus use case, Support Vector Machine (SVM) achieved Accuracy (ACC) 98.53%. second we ability classification algorithm differentiate between NCD. In SVM linear kernel outperformed all tested classifiers achieving 98.55% ACC. To best our knowledge, first study that documents automated differentiation These findings stepping stone toward image analysis can significantly benefit healthcare providers.