Methods for real-world gait analysis based on wearable sensors: mobility assessment on healthy and pathological subjects

The deterioration of walking ability has a significant impact on quality of life, as low gait speed is associated with mortality, dementia, cognitive decline, and fall risk. This holds true not only for individuals affected by motor impairment-related diseases but also for healthy elderly individuals as they age. Conversely, maintaining good mobility levels ensures individual independence and overall well-being in daily life. Therefore, assessing gait in real-life scenarios, considering internal and external factors, becomes essential in understanding and preventing this decline. Traditional laboratory gait analysis alone is insufficient as it primarily evaluates motor capacity; hence, it should be complemented with performance measures. Wearable sensors, particularly inertial measurement units (IMUs), represent an ideal solution for this transition. IMUs can extract both temporal and spatial gait variables while minimizing costs and inconvenience. However, their validation still primarily occurs in laboratory settings, necessitating validation under conditions resembling real-world use. Currently, a technically valid wearable solution providing accurate gait parameters for real-world conditions is lacking. Accurate gait assessment requires correctly identifying gait events to segment walking sequences into gait cycles and subphases (stance and swing). Existing IMU-based algorithms mostly rely on indirect methods and cannot serve as accurate references. On the other hand, technologies providing direct estimates are often time-consuming (e.g., body cameras), expensive (e.g., pressure mapping insoles), or possess low spatial resolution (e.g., foot switches). To address this, a new algorithm based on low-cost pressure insoles for accurate gait event detection was developed and validated. The algorithm utilized data from pressure insoles to detect gait events using a cluster-based approach, considering the activation or deactivation of at least three sensors in the same foot region as initial (IC) or final contact (FC). Testing on nine healthy participants against force platforms yielded low average root mean square errors for both gait events (approximately 20 ms for IC and 10 ms for FC) and temporal parameters (20 ms for stance duration and <10 ms for step duration). This suggests that this solution can serve as an accurate wearable reference, although further validation is needed, especially in non-rectilinear walking. While pressure insoles provide temporal gait characterization, spatial variables are also necessary for a comprehensive description. Therefore, a new algorithmic pipeline using a multi-sensor wearable system (INDIP) was developed and validated across various cohorts, including patients with different diseases. The system, comprising pressure insoles, IMUs on feet and lower back, and distance sensors, demonstrated excellent performance in structured tests and real-world activity simulations. While further improvements in sensor fusion and stride selection are possible, the INDIP system is a feasible reference for real-world gait analysis. Moreover, the same multi-sensor system can function as a mobile gait laboratory for clinical and rehabilitation applications, such as assessing the effectiveness of medical treatments in real-world conditions. Finally, the thesis demonstrates the application of the INDIP system in quantifying gait changes in Multiple Sclerosis patients before and after physical rehabilitation. The study suggests that the system can provide relevant gait measures for clinical use, although further research with larger participant groups and control groups is needed for conclusive results. In summary, this thesis presents solutions for validating and assessing gait in real-world conditions. The first two contributions enable the estimation of reference temporal and spatial gait parameters using affordable wearable solutions, while the third explores the system's use in clinical applications.

The deterioration of walking ability has a significant impact on quality of life. Low gait speed is linked to mortality, dementia, cognitive decline, and fall risk. This applies to those with motor impairment-related diseases and healthy elderly individuals. Maintaining mobility ensures independence and well-being. Assessing gait in real-life scenarios, considering internal and external factors, is crucial in preventing decline. Traditional lab gait analysis is insufficient, focusing on motor capacity. Wearable sensors, like inertial measurement units (IMUs), offer a solution. IMUs provide temporal and spatial gait data, minimizing costs and inconvenience. However, validation mostly occurs in labs, lacking real-world accuracy. A new algorithm using pressure insoles was developed to detect gait events. It showed promise with low errors in healthy participants. Nevertheless, further validation is needed, especially in non-rectilinear walking scenarios. Spatial variables require a multi-sensor wearable system (INDIP). It performed exceptionally well in structured tests and real-world simulations, serving as a feasible reference for real-world gait analysis. Additionally, the same multi-sensor system can function as a mobile gait laboratory for clinical and rehabilitation applications. The INDIP system was used to quantify gait changes in Multiple Sclerosis patients before and after rehabilitation, showing potential clinical utility. This application demonstrates the system's versatility and relevance in clinical settings. In summary, this thesis offers innovative solutions for real-world gait assessment using wearable technology, addressing both temporal and spatial parameters. These advancements have the potential to significantly impact the fields of healthcare, rehabilitation, and gerontology, enhancing the overall quality of life for individuals across diverse populations