Proposing a two-step decision support system for differential diagnosis of tuberculosis from pneumonia

Abstract

Mycobacterium tuberculosis (TB) is an infectious bacterial disease presenting similar symptoms to pneumonia. Therefore, differentiating between TB and pneumonia can be challenging for physicians and lead to delays in diagnosis and treatment. Early diagnosis of TB in particular, is critical in preventing community spread. The purpose of this study is to propose a method of differential diagnosis of TB from pneumonia using low-cost features. A two-step decision support system called Pneumonia-Tuberculosis Diagnosis Support System (PTBDSS) is proposed for differential diagnosis of TB from pneumonia based on stacked ensemble classifiers. The first step of the proposed model aims to identify an early diagnosis based on low-cost features, including demographic characteristics and patient symptoms. The second step of the proposed model confirms a diagnosis based on meta features extracted in the first step, laboratory tests, and chest radiography reports. The meta feature is a vector of length five, and each number in that vector comes from the vote of one classifier. This retrospective study considers 199 medical records of patients admitted to the isolation ward of a hospital in Arak, Iran, with suspected TB or pneumonia. Experimental results show that the proposed method outperforms the compared machine learning methods for early differential diagnosis of pulmonary tuberculosis from pneumonia with AUC of $90.26\pm 2.30$ and accuracy of $91.37\pm 2.08$ with $95\%$ CI and final decision making with AUC of $92.81\pm 2.72$ and accuracy of $93.89\pm 2.81$ with $95\%$ CI.