Automatic Diagnosis of Sarcoidosis Stage Using Machine Learning Algorithms

Abstract

Sarcoidosis is a rare multisystem disease primarily manifesting in the lungs and classified into four stages based on its impact on the lungs. Due to its rarity, sarcoidosis is often misdiagnosed. Data containing chest X-ray images were collected, preprocessed, and segmented to extract Regions of Interest (ROI). Relevant textural features were then extracted using Gray Level Co-occurrence Matrix techniques to represent lung tissue efficiently. The collected images were fed into classifiers, including "K Nearest Neighbors," "Support Vector Machine," and "Artificial Neural Network," for training. The classifiers were trained and tested on non-overlapping subsets of the datasets, achieving overall accuracies of 92.03%, 89.13%, and 81.16%, respectively, with the KNN classifier outperforming the others across all metrics.