Application of Machine Learning Algorithms in Urinary Tract Infections Diagnosis Based on Non-Microbiological...

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Clinical Microbiology

Background:

• Urinary tract infections (UTIs) are highly prevalent, particularly in women and hospitalized individuals.
• Current UTI diagnosis relies on clinical signs and urine cultures, which are time-consuming and resource-intensive.
• Optimizing UTI detection is crucial for timely patient management and resource allocation.

Purpose of the Study:

• To develop and evaluate artificial intelligence (machine learning) models for predicting positive urine cultures.
• To utilize non-microbiological laboratory parameters for expedited UTI diagnosis.
• To reduce unnecessary urine cultures and improve diagnostic efficiency.

Main Methods:

• Analysis of 4283 urine cultures from patients with suspected UTIs (2016-2020).
• Application of various machine learning algorithms, including Random Forest and Tree, to predict UTI and microorganism type.
• Identification of key predictive variables from non-microbiological tests.

Main Results:

• Random Forest achieved 82.2% accuracy and 87.1% AUC for predicting positive urine cultures.
• The Tree algorithm accurately predicted Gram-negative bacilli with 79.0% accuracy.
• Key predictors included urine dipstick (leukocytes, nitrites), white blood cell count, monocyte count, lymphocyte percentage, and creatinine levels.

Conclusions:

• AI algorithms integrated with non-microbiological parameters show significant promise for optimizing UTI diagnosis.
• Machine learning models can effectively predict UTIs, potentially reducing reliance on traditional urine cultures.
• Further clinical validation is necessary before widespread hospital practice integration.