FHIR R4 SHAP XAI ● Live

AI agents collaborating through explainable biomarker analysis and FHIR-compatible healthcare workflows.

Speech Biomarker Screening

Adjust key vocal biomarkers below. The model analyses 753 speech features to detect Parkinsonian patterns.

Loading feature inputs…

Biomarker Insights

Key speech biomarkers show increased variability and irregular energy patterns in Parkinson's patients, indicating vocal instability and neuromotor impairment.

Top 5 Most Influential Biomarkers
  1. Loading...
Global Feature Importance (SHAP)
SHAP Feature Importance Chart
Parkinson's vs Healthy — Biomarker Comparison
Loading biomarker analysis...
ℹ️
Note: Wavelet decomposition levels correspond to different frequency bands. Features from Level 12 and Level 13 reflect behavior in different parts of the speech spectrum and are not directly comparable in magnitude.

Clinical Note: These opposing trends indicate a redistribution of spectral energy — from strong, stable peaks to irregular, noise-like fluctuations — a hallmark of Parkinsonian speech.

Values are standardized and derived from the training dataset · XGBoost + SHAP TreeExplainer · ROC AUC 0.946
Agent Workflow Pipeline
🎙️
Speech Biomarkers
753 vocal features
→
🔬
Screening Agent
XGBoost + SHAP
→
📊
Biomarker Analysis
Top-5 attribution
→
🩺
Clinical Summary
LLM narrative
→
📋
FHIR Report
R4 DiagnosticReport

Clinical AI Agent

One call. Three agents. A complete, interoperable clinical screening result.

Uses median biomarker values · Customise in the Screening tab first