ENHANCED DIAGNOSTIC STRATEGIES FOR SEVERE PNEUMONIA IN ATHLETES: INTEGRATING CHEST CT AND PLATELET PARAMETERS WITH DEEP LEARNING ALGORITHMS

Abstract

Objective: This study evaluates the effectiveness of using deep learning algorithms to integrate chest CT and platelet parameters for diagnosing severe pneumonia in athletes. Methods: We conducted a retrospective analysis of 65 athletes admitted with severe pneumonia between April 2015 and February 2021. These were divided based on outcomes into survival and non-survival groups, comprising 37 and 28 athletes respectively. Data analyzed included demographic characteristics, pre-admission antibiotic usage, severity scores such as the Acute Physiology and Chronic Health Evaluation II (APACHE II), and mechanical ventilation usage. Platelet count (PLT), mean platelet volume (MPV), and platelet distribution width (PDW) were measured at admission and discharge, evaluating their correlation with patient outcomes. Results: The study found significant differences in mechanical ventilation use and APACHE II scores between the survival and non-survival groups. Pathogenic cultures revealed common bacteria such as Acinetobacter baumannii and Staphylococcus aureus. Notably, changes in platelet indices, particularly the rates of change in PLT, MPV, and PDW, significantly correlated with patient outcomes. These indices provided predictive insights into the prognosis, with combined models of platelet rate changes enhancing predictive accuracy significantly. Conclusion: Integrating platelet parameters with deep learning-analyzed chest CT scans offers a promising approach to enhancing diagnostic accuracy for severe pneumonia in athletes. This method not only aids in early diagnosis but also in predicting patient outcomes, thus optimizing treatment strategies and potentially reducing the rate of rehospitalization.