<![CDATA[Introductions: Tuberculosis (TB) remains a major global health problem and has become one of the world’s leading infectious diseases, particularly affecting populations in low- and middle-income countries. Despite advancements in molecular testing, the accessibility, cost, and time requirements of conventional diagnostics limit early case detection. Exhaled breath analysis provides a promising non-invasive approach through the identification of volatile organic compounds (VOCs) produced during TB infection. This study aimed to develop and evaluate a portable diagnostic system that integrates VOC sensing, Artificial Intelligence (AI), and Internet of Things (IoT) technologies to enhance early TB screening in community and primary healthcare settings. Methods: A metal oxide semiconductor gas sensor array connected to an ESP32-S3 microcontroller was employed to capture VOC profiles from 33 participants (17 TB-confirmed patients and 16 healthy controls). The acquired data were preprocessed, reduced, and classified using Principal Component Analysis. Several machine learning algorithms, including Support Vector Machines (SVM), Random Forest, Gradient Boosting, and Artificial Neural Networks (ANN), were trained and validated to develop a TB recognition model. Results and Discussion: The ANN achieved the best performance, with an accuracy of 79%, sensitivity of 78%, specificity of 80%, and an AUC of 0.84. IoT integration enabled real-time data transfer and cloud-based visualization, demonstrating scalability and potential use in resource-limited settings. Conclusion: This portable AI-based breath analysis system offers a rapid, affordable, and non-invasive approach for early TB detection. With further validation, it might complement existing diagnostics and strengthen global TB elimination efforts.]]>
