Low-Cost Gaze-Based Communication Interface for ALS Patients Using MediaPipe FaceMesh: Prototype Development and Functional Evaluation

Abstract

Background: Amyotrophic Lateral Sclerosis (ALS) progressively impairs motor function while often preserving ocular mobility, making gaze-based communication a viable alternative for interaction. However, commercial eye-tracking systems are frequently inaccessible due to their high cost.

Objective: To design, implement, and evaluate a low-cost gaze-based communication interface using open-source computer vision tools and standard consumer hardware.

Methods: A prototype was developed in Python using OpenCV and MediaPipe FaceMesh for real-time iris tracking. The system employed normalized iris-position ratios, an 8-frame smoothing filter, and a personalized 9-point calibration procedure. A communication interface with 12 predefined phrases distributed across three contextual pages was controlled through gaze direction and a dwell-time activation mechanism. Seven healthy adult participants were evaluated under controlled lighting and distance conditions.

Results: Directional accuracy reached 91.4% for left gaze, 88.6% for right gaze, and 85.7% for upward gaze, all exceeding the predefined acceptance threshold of 80%. Mean activation latency was 1,847 ± 143 ms, the false activation rate was 4.3%, and inter-user variability was 4.7%. Downward gaze achieved 71.4% accuracy and was identified as the principal limitation due to partial iris occlusion by the lower eyelid.

Conclusion: The results demonstrate the technical feasibility of a low-cost, open-source gaze-based communication system for augmentative and alternative communication applications. The prototype achieved reliable performance using only a standard webcam and consumer hardware, offering a potentially accessible solution for individuals with severe motor impairments in resource-limited settings. Future work includes clinical validation with ALS patients and improvements to downward gaze detection.