Machine Learning and Sampling Techniques to Enhance Radiological Diagnosis of Cerebral Tuberculosis

Abstract

Cerebral tuberculosis (TB) is one of the neurological manifestations of tuberculosis infections responsible for devastating sequelae and mortality. It is a challenge to diagnose as it mimics other infectious and neoplastic pathologies of the brain. There is a need for a rapid and accurate diagnostic approaches, in order to prevent the dismal outcomes arising as a result of delayed or incorrect diagnosis. This paper aims to develop a classifier to diagnose cerebral TB using various radiological features present on Magnetic Resonance Imaging (MRI) of the brain with the help of Machine Learning (ML). Cases of TB and non-TB conditions (including meningiomas, gliomas, fungal and bacterial brain infection) presenting to Aga Khan University Hospital, Karachi, Pakistan, were included and divided into training and test datasets. Features were selected using correlation, and besides age and gender, included multiple radiological features recorded from MRI of the brain. After the application of Synthetic Minority Over-sampling Technique (SMOTE), SMOTE-Tomek Links, Edited Nearest Neighbor (ENN) SMOTE-ENN, and Adaptive Synthetic (ADASYN) techniques for balancing the datasets, classifier accuracy was tested using two models: logistic regression and random forest. Highest accuracy (90.9%) was achieved using logistic regression along with SMOTE+TOMEK with 95.4% Area under the Curve while obtaining a F1 score of 92.8%