Risk Classification for Interstitial Cystitis/Bladder Pain Syndrome Using Machine Learning Based Predictions.

Area of Science:

• Urology
• Biomarkers
• Machine Learning

Background:

• Interstitial cystitis (IC)/bladder pain syndrome (BPS) diagnosis requires improvement.
• Current diagnostic methods lack sufficient accuracy.
• Novel approaches are needed to enhance IC/BPS classification.

Purpose of the Study:

• To develop an improved IC risk classification model using machine learning.
• To enhance the diagnostic accuracy of IC/BPS.
• To create a novel classification model integrating patient-reported outcomes (PROs) and urine biomarkers.

Main Methods:

• Developed a machine learning predictive classification model (IC-PIS Score).
• Utilized 1264 crowdsourced urine samples and PROs, plus 296 academic center samples.
• Measured urinary cytokine biomarker levels and compared models.

Main Results:

• The top model combined biomarker measurements and PROs, achieving an AUC of 0.87.
• This model outperformed PROs alone (AUC=0.83) and biomarkers alone (AUC=0.58).
• The combination of biomarkers and PROs yielded improved predictive effects for IC.

Conclusions:

• IC-PIS is a novel classification model enhancing IC/BPS diagnostic accuracy.
• The model integrates PROs and urine biomarkers.
• Large-scale crowdsourced data and ambient shipping methods support the findings' robustness and scalability.