Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Cost-Effective Eye-Tracker for Early Detection of Mild Cognitive Impairment

Danilo Greco, Francesco Masulli, Stefano Rovetta, Alberto Cabri, Davide Daffonchio

TL;DR

This work addresses the need for accessible early detection of Mild Cognitive Impairment by leveraging Visual Paired Comparison tests with a low-cost, webcam-based eye-tracker. It introduces a two-subsystem architecture (Measurement and Test Management) that uses Viola–Jones detection and a dense neural network for gaze calibration, along with optional heart-rate variability monitoring via photoplethysmography from a forehead ROI. Quantitative calibration results show an average gaze localization error of about $3.00\%$ and perfect left/right discrimination in testing, while the HRV module provides a noninvasive stress indicator during testing. The proposed approach emphasizes practicality and scalability, aiming for broad diffusion in clinics and low-resource settings, contingent on future clinical validation.

Abstract

This paper presents a low-cost eye-tracker aimed at carrying out tests based on a Visual Paired Comparison protocol for the early detection of Mild Cognitive Impairment. The proposed eye-tracking system is based on machine learning algorithms, a standard webcam, and two personal computers that constitute, respectively, the "Measurement Sub-System" performing the test on the patients and the "Test Management Sub-System" used by medical staff for configuring the test protocol, recording the patient data, monitoring the test and storing the test results. The system also integrates an stress estimator based on the measurement of heart rate variability obtained with photoplethysmography.

A Cost-Effective Eye-Tracker for Early Detection of Mild Cognitive Impairment

TL;DR

This work addresses the need for accessible early detection of Mild Cognitive Impairment by leveraging Visual Paired Comparison tests with a low-cost, webcam-based eye-tracker. It introduces a two-subsystem architecture (Measurement and Test Management) that uses Viola–Jones detection and a dense neural network for gaze calibration, along with optional heart-rate variability monitoring via photoplethysmography from a forehead ROI. Quantitative calibration results show an average gaze localization error of about and perfect left/right discrimination in testing, while the HRV module provides a noninvasive stress indicator during testing. The proposed approach emphasizes practicality and scalability, aiming for broad diffusion in clinics and low-resource settings, contingent on future clinical validation.

Abstract

This paper presents a low-cost eye-tracker aimed at carrying out tests based on a Visual Paired Comparison protocol for the early detection of Mild Cognitive Impairment. The proposed eye-tracking system is based on machine learning algorithms, a standard webcam, and two personal computers that constitute, respectively, the "Measurement Sub-System" performing the test on the patients and the "Test Management Sub-System" used by medical staff for configuring the test protocol, recording the patient data, monitoring the test and storing the test results. The system also integrates an stress estimator based on the measurement of heart rate variability obtained with photoplethysmography.
Paper Structure (9 sections, 1 equation, 8 figures)

This paper contains 9 sections, 1 equation, 8 figures.

Table of Contents

  1. Introduction
  2. System Implementation
  3. System Design
  4. Calibration procedure and assesment
  5. Heart Rate Variability Monitoring
  6. Single-Board Computer Version
  7. Sub-systems synchrony
  8. Example of Test Protocol
  9. Discussion and Conclusions

Figures (8)

  • Figure 1: Raspberry Pi 4 and Raspberry Pi Camera Module 2.
  • Figure 2: Circle following a predefined pattern, used for calibration.
  • Figure 3: Click-points pattern used to measure the system's accuracy.
  • Figure 4: Accuracy evaluation: comparison of the horizontal position of the click-point vs the neural network estimate.
  • Figure 5: Calibration phase: the ellipse turns green if there is a face inside and the eyes are detected. BPM is also estimated. This figure compares also the optical heart rate monitor of our system with a commercial pulse oximeter device.
  • ...and 3 more figures