An enhancing diagnostic pulmonary diseases diagnostic method for differentiating talaromycosis from tuberculosis.

Talaromycosis (TSM) affects immunocompromised individuals, particularly those with human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS), causing varied pulmonary abnormalities on chest computed tomography (CT). These features overlap with pulmonary tuberculosis, making accurate differentiation essential for appropriate treatment. This study utilized real patient data from the First Affiliated Hospital of Wenzhou Medical University. A machine learning model, termed bIPCACO-FKNN, was developed, integrating an ant colony optimization (ACO) algorithm with a fuzzy k-nearest neighbors (FKNNs) classifier. This model introduces an incremental proportional-integral-derivative control strategy to enhance the search efficiency of ACO. Comparative analysis with several algorithms in the CEC 2017 benchmark functions confirms the superior performance of the IPCACO. Applying the bIPCACO-FKNN model for the prediction of pulmonary TSM achieved a prediction accuracy of 98.196% and a specificity of 99.500%, thus demonstrating its significant efficacy in accurately distinguishing between pulmonary TSM and tuberculosis. This provides an efficient and reliable machine learning tool for the differentiation between pulmonary TSM and tuberculosis.