← Back to All Projects
Research Prototype Clinical AI & Decision Support

CURES-AI-UDS

Explainable Urodynamic Study Classification & Specialist Report Drafting

Machine learning classifier categorizing lower urinary tract dysfunction with tree SHAP explainability and automated structured report drafting.

UrodynamicsSHAP ExplainabilityUrologyReport GenerationMachine Learning
Launch Live Application ↗
Actual Application Interface · High-Resolution Capture cures.studio
Interface of CURES-AI-UDS

The Clinical & Engineering Problem

Urodynamic study (UDS) interpretation requires manual pressure-flow curve analysis, where summary parameters often overlap with final diagnostic categories, causing variable clinical concordance.

The Technical Approach

Trains multi-label classifiers across 9 lower urinary tract diagnoses on 527 public studies, providing per-patient SHAP feature attributions and drafting a clinician-reviewable summary report.

Engineering Architecture

  • Nested 5x5 cross-validation evaluating one-vs-rest classifiers across 9 overlapping diagnostic classes
  • Label-proximity audit classifying whether inputs are direct definitional criteria or independent predictors
  • Tree SHAP explainability engine attributing predictions to storage vs voiding parameters
  • Specialist report generation interface requiring explicit clinician sign-off

Documented Evidence & Verification

527
Real Patient Studies
Curated urodynamic examinations from public clinical cohort (DOI: 10.17632/vr2brd5wt6.1)
+0.144
Macro-AP Gain
Improvement over core clinical baseline (95% CI +0.110 to +0.183)
9
Diagnoses Modeled
Concurrent detrusor, compliance, and bladder outlet classifications
Scientific Status & Regulatory Disclaimers

Research prototype; not an FDA-cleared diagnostic device. Must not be used for unassisted patient care decisions without specialist urologist review.

Open Working Workspace ↗ Inquire About This Project ← Back to All Projects