A Deep Learning Approach for Salivary Gland Echographic Image Analysis in Sjögren's Syndrome Detection

Abstract

The purpose of this study is to propose a deep learning (DL) approach for the analysis of CT images aiming to assist in the diagnosis of Sjogren's syndrome (SS), a chronic autoimmune disease characterized by exocrine gland inflammation. A dataset consisting of 225 CT images obtained from 115 patients diagnosed with Sjögren's syndrome was employed for this study. To enhance the model's generalizability and robustness, the images underwent a preprocessing and augmentation pipeline. Then a DL model was trained, based on a convolutional neural network (CNN) architecture in order to classify the images whether they correspond to SS or healthy patients. The proposed model achieved an overall accuracy of 99.0%, with a sensitivity of 98.57% and a specificity of 99.45%.