ScanWise A Deep Learning Approach for TB Detection in X-Rays With OpenCV and TensorFlow Integration

Abstract

Although there have been a great many advances in medical technology, tuberculosis continues to present itself as a substantial global health challenge, particularly in resource-poor areas. In this paper, we present ScanWise, a novel deep learning system based on the concept of integrating chest X-ray analysis with patient data for enhanced TB detection. The system comprises of convolutional neural networks (CNNs) that leverage state-of-the-art image processing and geographical information system integration to develop a holistic-diagnostic support tool for TB. Multiple experiments employed an equally sized dataset of chest X-ray along with medical records, achieving an overall accuracy of 85%, precision of 90%, recall of 81.8%, and an F-score of 85.7%. The results highlight the excellent prediction ability of the model, achieving an appropriate balance between sensitivity and specificity. Furthermore, the addition of location-based services for mapping a hospital adds to the healthcare accessibility in providing a novel recruitment method for TB screening. This paper describes the processes, challenges, and empirical findings in support of this methodology in automated medical diagnosis.

Introduction

Summary:

Tuberculosis (TB) remains a critical global health issue, especially in resource-limited regions, with millions of new cases and high mortality rates annually. Early detection is essential to controlling TB transmission, but conventional diagnostic methods (clinical exams, lab tests, imaging) are often slow, inaccurate, and resource-intensive. This motivates the development of automated, efficient, and accurate detection systems.

The study proposes a novel system leveraging deep learning with a custom Convolutional Neural Network (CNN) to analyze chest X-rays for TB detection. It integrates clinical and demographic data, provides an accessible user interface for healthcare providers, and incorporates location-based hospital mapping. The system uses advanced image preprocessing with OpenCV, a custom CNN architecture implemented in TensorFlow, and a Flask-based backend.

The CNN architecture processes 224x224 chest X-ray images through three feature extraction blocks with increasing filter sizes (32, 64, 128), batch normalization, ReLU activations, max pooling, and dropout to prevent overfitting. The final classification head uses global average pooling and a dense layer to output binary TB positive/negative predictions.

The dataset combines public chest X-ray collections from NLM, Belarus, NIAID, and RSNA, totaling thousands of TB and normal images along with clinical and demographic features.

Performance evaluation showed the model achieved 85% accuracy, 90% precision, 81.8% recall, and an F1-score of 85.7%, demonstrating strong predictive ability with some room for improvement to reduce overfitting and enhance generalization. Training loss steadily decreased while validation loss showed some fluctuations, suggesting potential overfitting that could be mitigated with techniques like increased dropout or early stopping.

Overall, this deep learning approach shows promise for rapid, reliable TB detection from chest X-rays, which can help improve early diagnosis and reduce disease transmission in resource-constrained settings.

Conclusion

Tuberculosis (TB) remains a major public health challenge on the global stage, especially in resource-limited areas. As per the World Health Organization data, TB ranks as one of the top 10 causes of death globally, with millions of new cases being diagnosed annually. Early diagnosis is the important key in controlling the transmission of TB, but presently used diagnostic methods do not grant rapid and adequate identification, which results in a delay in administering appropriate treatment and increases the incidence of transmission.

References

[1] AI-Assisted Tuberculosis Detection and Classification from Chest X-Rays Using a Deep Learning Normalization-Free Network Model Vasundhara Acharya 1, Gaurav Dhiman 2 3 4, Krishna Prakasha 5, Pranshu Bahadur 5, Ankit Choraria 5, Sushobhitha M 6, Sowjanya J 7, Srikanth Prabhu 1, Krishnaraj Chadaga 1, Wattana Viriyasitavat 8, Sandeep Kautish 9 Affiliations Expand PMID: 36225551 PMCID: PMC9550434 DOI: 10.1155/2022/2399428 [2] Machine and Deep Learning for Tuberculosis Detection on Chest X-Rays: Systematic Literature Review Authors of this article:Seng Hansun1, 2 Author Orcid Image ; Ahmadreza Argha3, 4, 5 Author Orcid Image ; Siaw-Teng Liaw6 Author Orcid Image ; Branko G Celler7 Author Orcid Image ; Guy B Marks1, 2 Author Orcid Image Comprehensive Approach to TB Diagnosis Powered by Machine Learning [3] Tawsifur Rahman, Amith Khandakar, Muhammad A. Kadir, Khandaker R. Islam, Khandaker F. Islam, Zaid B. Mahbub, Mohamed Arselene Ayari, Muhammad E. H. Chowdhury. (2020) \"Reliable Tuberculosis Detection using Chest X-ray with Deep Learning, Segmentation and Visualization\". IEEE Access, Vol. 8, pp 191586 - 191601. DOI. 10.1109/ACCESS.2020.3031384. [4] S. Jaeger, S. Candemir, S. Antani, Y.-X. J. Wáng, P.-X. Lu, and G. Thoma, \"Two public chest X-ray datasets for computer-aided screening of pulmonary diseases,\" Quantitative imaging in medicine and surgery, vol. 4 (6), p. 475(2014) [5] B. P. Health. (2020). BELARUS TUBERCULOSIS PORTAL [Online]. Available: http://tuberculosis.by/. [Accessed on 09-June-2020] [6] NIAID TB portal program dataset [Online]. Available: https://tbportals.niaid.nih.gov/download-data. [7] kaggle. RSNA Pneumonia Detection Challenge [Online]. Available: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data. [Accessed on 09-June-2020]

Copyright

Copyright © 2025 Tanmay Mahendra Dhanawde . This 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.