Facial Droop Simulator

1. Diagnostic Overview

Facial paresis (muscle weakness) is a common clinical symptom that requires rapid medical evaluation. This simulator models unilateral facial droop, demonstrating the somatic motor signs that differentiate central cortical damage from peripheral cranial nerve injuries.

Understanding the biomechanics of asymmetric facial expressions is crucial for educators, students, and healthcare professionals. Visualizing the functional impacts of muscle sagging makes neurological anatomy accessible, helping users learn to spot warning signs.

Furthermore, this platform implements a voice sonification engine to model dysarthria (motor speech articulation deficits). Facial paralysis alters lips closure, causing breathy air leakage and muffled formant resonance, which is rendered audibly as the facial distortion rises.

2. Operating Instructions

1. Select Workspace Mode: Use the Virtual Mannequin for rapid anatomical study, or tap Live Webcam to project facial droop onto your own live video feed in real-time.
2. Configure Presets: Choose a neurological pathology (Stroke or Bell's Palsy) to auto-select clinical landmark clusters and dynamically update the diagnostics dashboard.
3. Interact Manually: In Custom Sandbox mode, directly tap individual key landmarks (green dots) on the mannequin or video feed to isolate custom muscle pairs.
4. Adjust Sagging: Drag the Droop Intensity slider to simulate localized tissue laxity.
5. Activate Sonification: Turn on the Voice Sound and move the slider to hear how facial droop distorts human vowel articulation (dysarthria simulation).

3. Technical Details

The simulator operates across two distinct graphics pipelines. The Live Webcam Mode leverages MediaPipe Face Mesh via TensorFlow.js to track facial landmarks. Texture mapping and sagging calculations are processed dynamically using affine transformation matrices inside cropped Delaunay triangulation meshes.

The Virtual Mannequin Sandbox maps anatomical coordinates to a vector rendering engine. When the slider is adjusted, coordinates undergo a customized anatomical gravity translation formula: $$\Delta y = \text{maxDroop} \times \text{factor} \times \text{intensity}$$ $$\Delta x = -\text{pullX} \times \text{intensity}$$ modeling muscle fiber sag and lateral inward pull.

The Vocal Sonification Engine implements dual peaking Web Audio BiquadFilter nodes to synthesize standard F1/F2 vocal tract formants. Parallel noise buffer sources simulate lip air escape, dynamically modulated by Q-factor scaling.

4. Future Directions

Ongoing iterations focus on incorporating volumetric 3D modeling pipelines (WebGL/Three.js) to map depth coordinates, mimicking realistic clinical lighting and facial crease changes.

Further developments will integrate machine learning speech classifiers that analyze vocal recordings to quantify speech degradation alongside visual asymmetries, working toward a unified educational scoring system.

Additionally, expanding the anatomical tracking to coordinate cranial nerves III, IV, and VI (extraocular muscle pathways) will enable the modeling of complex multi-neuropathy conditions like cavernous sinus syndromes.