Human-Data Interaction in Predicting Clinical Outcomes in Tuberculosis: A Web Prototype with Streamlit for Explainability and Visualization

• Ronilson W. S. Pereira Rio de Janeiro State University http://orcid.org/0000-0002-3851-1338
• Igor Falcão Federal University of Pará
• Saul Carneiro Federal University of Pará
• Marcos Seruffo Federal University of Pará
• Karla Figueiredo Rio de Janeiro State University

DOI: https://doi.org/10.5753/ihc_estendido.2025.13301

Abstract

Introduction: Machine learning models; such as Random Forest; have proven effective in predicting clinical outcomes; although the interpretability of these models still poses a challenge. Objective: This paper presents an interactive web prototype; developed with Streamlit; aimed at explaining and visualizing clinical predictions in patients with tuberculosis. Methodology or Steps: The solution uses the Random Forest algorithm to generate predictions and incorporates interpretability techniques; such as SHAP; allowing interaction with the data and providing visual explanations about the contribution of variables to the results. Expected Results: The system is expected to promote an accessible and understandable interface; bringing artificial intelligence closer to clinical practice through human-data interaction.

Keywords: Machine learning, Random Forest, Interpretability, SHAP, Tuberculosis

References

Hartshorn, S. (2016). Machine learning with random forests and decision trees: A visual guide for beginners. Kindle edition.

Kanesamoorthy, K. e Dissanayake, M. B. (2021). Prediction of treatment failure of tuberculosis using support vector machine with genetic algorithm. The International Journal of Mycobacteriology, 10(3):279–284.

Ma, F.-q., He, C., Yang, H.-r., Hu, Z.-w., Mao, H.-r., Fan, C.-y., Qi, Y., Zhang, J.-x., e Xu, B. (2023). Interpretable machine-learning model for predicting the convalescent covid19 patients with pulmonary diffusing capacity impairment. BMC Medical Informatics and Decision Making, 23(1):169.

Orjuela-Cañón, A. D., Jutinico, A. L., Awad, C., Vergara, E., e Palencia, A. (2022). Machine learning in the loop for tuberculosis diagnosis support. Frontiers in Public Health, 10:876949.

Paixão, G. M. d. M., Santos, B. C., Araujo, R. M. d., Ribeiro, M. H., Moraes, J. L. d., e Ribeiro, A. L. (2022). Machine learning na medicina: revisão e aplicabilidade. Arquivos Brasileiros de Cardiologia, 118(1):95–102.

Tendedez, H., Ferrario, M.-A., McNaney, R., e Gradinar, A. (2022). Exploring humandata interaction in clinical decision-making using scenarios: co-design study. JMIR human factors, 9(2):e32456.

World Health Organization (2024). Global Tuberculosis Report. Number September.

Yasin, P., Yimit, Y., Cai, X., Aimaiti, A., Sheng, W., Mamat, M., e Nijiati, M. (2024). Machine learning-enabled prediction of prolonged length of stay in hospital after surgery for tuberculosis spondylitis patients with unbalanced data: a novel approach using explainable artificial intelligence (xai). European Journal of Medical Research, 29(1):383.