Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{205399,
        author = {Suresh S and Dhanalakshmi S},
        title = {Advancements in Machine Learning for Tuberculosis: A Systematic Analysis of Clinical Symptom Severity Datasets and Multimodal Fusion},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {6284-6299},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=205399},
        abstract = {In this systematic analysis, the findings concerning the development and utilization of datasets with symptom severity measures in clinical TB cases within the machine learning (ML) pipeline have been outlined. First, the aims encompass assessing the diversity and source of the datasets employed, benchmarking some ML approaches by using their symptom severity, as well as researching multimodal fusion strategies. Moreover, estimating labelling strategies and, ultimately, analysing the interpretability strategies adopted by the selected studies. In total, over 43 investigations conducted in numerous areas around the globe have been scrutinized, incorporating diverse modalities into the examination including clinical, acoustic, radiological, and genomic information. As per the results, fusing cough sound, clinical, and imaging data leads to improved diagnostic performance and more accurate predictions when employing advanced multimodal algorithms that produce AUCs above 0.9. Nevertheless, highly annotated datasets have significantly advanced ML algorithms' performance; however, patient adherence to the annotation guidelines and lack of consistency between each other is still a major challenge in the low resource settings. Techniques such as SHAP and feature importance assessment help build clinical trust but are, however, not extensively used across the literature. The information accumulated demonstrates that amalgamated longitudinal symptom intensity datasets possess significant promise for diagnosis and therapeutic oversight of tuberculosis utilizing machine learning. According to the review, data must be standardized, interpretable models must be developed to ensure scalable and clinically meaningful solutions in different healthcare settings.},
        keywords = {Tuberculosis; Machine Learning; Clinical Symptom Severity; Multimodal Data Fusion; Cough Sound Analysis; Explainable Artificial Intelligence (XAI).},
        month = {June},
        }