Augmentation of Infrared Microscopy of White Blood Cells and Medical Measures for Rapid and Accurate Diagnosis of Bacterial or Viral Infections in Febrile Pediatric Oncology Patients: An Expert System-Based Study

Infectious diseases, a major contributor to high mortality rates, often exhibit similar symptoms, despite variations in immune responses to bacterial or viral infections. Rapidly differentiating bacterial infections from viral infections in febrile pediatric oncology patients is critical to reduce unnecessary antibiotic use and improve patient outcomes. Current diagnostic procedures require 2-4 days, prompting physicians to rely on clinical measures like C-reactive protein (CRP), white blood cell (WBC) count, and absolute neutrophil count (ANC) despite their limited specificity, leading to unnecessary antibiotic treatment. This study aims to accelerate and enhance the infection etiology prediction of bacterial or viral infections. Thus, we first evaluated the maximum achievable diagnostic accuracy using CRP, WBC, and ANC and found a success rate of approximately 70%. Additionally, we explored the potential of infrared spectroscopy of isolated WBCs by applying machine learning algorithms, which yielded a 97% classification accuracy for bacterial vs viral infections. This involved implementing various analysis strategies and employing a decision system. Finally, augmenting the infrared spectra with CRP, WBC, and ANC data further boosted diagnostic accuracy to 98.6%. This study included 50 bacterial infections, 21 viral infections, and 39 control cases for medical measures. For infrared spectroscopy, data were collected from 59 bacterial infections, 29 viral infections, and 92 controls (pediatric oncology patients without fever). These findings underscore that augmenting infrared spectroscopic data with traditional clinical measures can shorten diagnosis time to roughly 1 h, improve infection etiology determination, and potentially curb the overuse of antibiotics in vulnerable pediatric oncology populations.