Artificial Intelligence in Differential Diagnostics of Meningitis: A Nationwide Study

Area of Science:

• Medical Informatics
• Infectious Diseases
• Clinical Diagnostics

Background:

• Accurate differential diagnosis between bacterial and viral meningitis is critical for appropriate patient management.
• Current diagnostic methods can be time-consuming or lack specificity, necessitating improved approaches.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in differentiating bacterial from viral meningitis.
• To identify key predictors for accurate meningitis diagnosis across different age groups.

Main Methods:

• Applied multiple logistic regression (MLR), random forest (RF), and naïve-Bayes (NB) algorithms to patient data.
• Utilized cerebrospinal fluid (CSF) and blood markers including neutrophils, lymphocytes, C-reactive protein (CRP), albumin, glucose, and neutrophil-to-lymphocyte ratio (NLR) as predictors.
• Analyzed data from two age groups (0-14 and >14 years) using both conventional and molecular diagnostic methods.

Main Results:

• Machine learning models achieved high prediction accuracy, exceeding 95% for viral and 78% for bacterial meningitis.
• MLR and RF algorithms demonstrated optimal performance, particularly when utilizing CSF neutrophils, CSF lymphocytes, NLR, albumin, glucose, gender, and CRP.
• The neutrophil-to-lymphocyte ratio (NLR) was reconfirmed as a highly accurate diagnostic marker.

Conclusions:

• Machine learning, particularly MLR and RF, offers a powerful tool for the differential diagnosis of meningitis.
• Specific clinical and laboratory markers, notably NLR, are highly effective in distinguishing between bacterial and viral meningitis.
• These findings support the integration of ML-based approaches into clinical practice for improved meningitis diagnosis.