GAITPRO: Intelligent gait analysis (GAIT) for predictive clinical prognosis in anomaly detection and early warning of age-related diseases

GAITPRO addressed a particularly relevant challenge in Asturias, the oldest autonomous community in Spain, where demographic change has a direct impact on the onset of age-related chronic diseases. Human gait is a biometric parameter whose deviations can be associated with conditions such as Parkinson's disease and other neurodegenerative diseases, so its continuous analysis makes it possible to detect these ailments earlier.

The objectives of the project were:

• Implement an ambulatory gait assessment system that can be used in the home environment, as opposed to laboratory systems, which are more accurate but far more costly and restrictive.
• Design artificial intelligence algorithms for gait assessment, conceived to be computed at the edge.
• Facilitate early detection so that prevention and early care protocols can be implemented, reducing the burden of chronic conditions on the health system.

The system was built around three pillars integrated through Web of Things (WoT) standards under the edge computing paradigm:

• Measurement wearable: a proprietary device with a nine-degrees-of-freedom inertial sensor (accelerometer, gyroscope, and magnetometer), microcontroller, battery, and local storage, housed in a 3D-printed case and integrated into an adjustable belt that places the sensor in the lower back, close to the body's center of gravity. The device offered 19 hours of battery life in continuous recording.
• Gait modeling: signal cleaning algorithms and automatic identification of walking segments, with classification of the four phases of the gait cycle and supervised and unsupervised machine learning models.
• Visualization platform: a multi-device web application with differentiated profiles for clinical staff, support staff, and patients, complemented by a desktop application for configuring the device and downloading the recordings.

CTIC developed GAITPRO in clinical collaboration with the Geriatrics Clinical Management Area of the Hospital Monte Naranco in Oviedo, which provided healthcare support to the project and obtained the favorable authorization of the Research Ethics Committee of the Principality of Asturias. The study collected data from 54 people aged between 23 and 92. The supervised models achieved an accuracy rate of 99.75% in estimating a person's age from their gait pattern, and 97.3% when the personal variables of height and weight were excluded. The architecture was designed with an adaptive approach that allows new modeling algorithms to be incorporated, so the system can be extended to the early detection of other ailments.