ATBdiscrimination: An in Silico Tool for Identification of Active Tuberculosis Disease Based on Routine Blood Test and T-SPOT.TB Detection Results

Abstract

Tuberculosis remains one of the deadliest infectious diseases worldwide. Only 5–15% of people infected with Mycobacterium tuberculosis develop active TB disease (ATB), while others remain latently infected (LTBI) during their lifetime, which has a completely different clinical treatment schedule. However, most current clinical diagnostic methods are based on the immune response of M. tuberculosis infections and cannot distinguish ATB from LTBIs. Thus, the rapid diagnosis of active or latent tuberculosis infections remains a serious challenge for clinicians. In this work, based on the test data of a total of 478 patients, 36 blood biochemical data were specially included with T-SPOT.TB detection results which are all from routine clinical practice as commercially available. Then a discrimination method to detect ATB infections was successfully developed based on these data by the random forest algorithm. This method presents a robust classification performance with AUC as 0.9256 and 0.8731 for the cross-validation set and the external validation set, respectively. This work suggests an innovative strategy for identification of ATB disease from a single drop of blood with advantages of being timely, efficient, and economical. It also provides valuable information for the comprehensive understanding of TB with deep associations between TB infection and routine blood test data. The web server of this identification method, called ATBdiscrimination, is now available online at .

Severity detection for the coronavirus disease 2019 (COVID-19) patients using a machine learning model based on the blood and urine tests

“…Its model complexity renders itself computation-intensive and RF runs slower than many prediction algorithms. RF is another popular algorithm for building the prediction models using the clinical data [27,28].…”

Differential Diagnosis of Latent Tuberculosis Infection and Active Tuberculosis: A Key to a Successful Tuberculosis Control Strategy

“…As mentioned above, the immunological methods that are frequently used, such as skin tests and IGRAs, are intrinsically unable to discriminate LTBI from aTB. Recently, with the development of artificial intelligence (AI) and bioinformatics, new strategies have been introduced to improve diagnostic performance in distinguishing LTBI from aTB, such as the ImmunoScore (IS) model, Cox proportional hazards model, expectation maximization algorithm, silico mapping algorithm, and random forest algorithm (Villate et al, 2006;Zhou et al, 2017;Ndzi et al, 2019;Wu et al, 2019;Rambaran et al, 2021). IS is a novel and promising prognostic tool that is widely used in tumors.…”

Artificial Intelligence Predicts Severity of COVID-19 Based on Correlation of Exaggerated Monocyte Activation, Excessive Organ Damage and Hyperinflammatory Syndrome: A Prospective Clinical Study

“…In this study we investigated the possibility of predicting the severity of COVID-19 by using cytokines/blood test data. To solve the binary classification problem, we built a logistic regression model, named Support Vector Machine and Random Forest, which is widely used to construct clinical prediction models (36)(37)(38).…”