Applying AI‐Driven Predictive Models for Diagnosis, Treatment Modality, and Duration of Cutaneous Leishmaniasis

Abstract

Background Cutaneous leishmaniasis (CL) is an infectious disease that affects thousands of individuals annually in tropical and subtropical regions. Early diagnosis, accurate prognosis, and proper disease classification are critical for selecting an appropriate treatment modality. Objective This study aims to apply machine learning (ML) and deep learning (DL) techniques to determine the appropriate treatment modality and duration for CL patients in Semnan Province, Iran, with the ultimate goal of providing a faster, more effective therapeutic approach. Materials and Methods We collected the data used in this research from the Ministry of Health and Medical Education of Iran. Seventeen unique and independent patient features were selected to predict treatment modality and duration. We then applied several AI‐derived predictive models, including multilayer perceptron (MLP), k‐nearest neighbors, random forests, support vector machines, and decision trees, to classify the collected data and compared their results using classification evaluation metrics. Results The MLP model demonstrated superior performance in diagnosing treatment modality, achieving an accuracy of 80%, sensitivity of 63%, and an area under the curve (AUC) of 81%. For predicting treatment duration in the systemic antimony group, the MLP model achieved 86% accuracy, 80% sensitivity, and an AUC of 94%. For predicting treatment duration in the local antimony group, the MLP model achieved 78% accuracy, 60% sensitivity, and an AUC of 89%. Among the features analyzed, lesion size had the greatest impact on the classification model, while lesion history had the least. Conclusion The findings emphasized the effectiveness of the ML models in facilitating the rapid diagnosis and management of CL. This approach significantly improved diagnostic accuracy and therapeutic decision‐making for the disease. Furthermore, the present study established a foundation for extending the use of ML algorithms to the analysis and management of other diseases.

Ähnliche Arbeiten