Evaluación crítica de Algoritmos de segmentación del test Timed Up and Go (TUG) basados en IMU.
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
El Timed Up and Go (TUG) es una prueba funcional ampliamente utilizada para evaluar movilidad y riesgo de caídas, especialmente en adultos mayores y personas con patologías neurológicas. Sin embargo, su enfoque tradicional basado únicamente en el tiempo total presenta limitaciones al no considerar aspectos cualitativos del movimiento. La integración de sensores inerciales (IMU) y algoritmos ha permitido segmentar el TUG en subetapas, mejorando la detección de deficiencias motoras.
Este trabajo valida y desarrolla un algoritmo automático de segmentación del TUG, utilizando datos inerciales en personas de distintos rangos etarios. Se revisaron algoritmos existentes, se evaluó un modelo basado en reglas y se propuso una optimización mediante votación de jurados, comparando resultados con el sistema OptiTrack, estándar de referencia para este trabajo. El algoritmo se evaluó por precisión, sensibilidad, error promedio y correlación con segmentación manual. Se espera aportar herramientas más precisas para la evaluación clínica, facilitando la integración del TUG automatizado en entornos de salud.
En la validación con 75 participantes de diversos rangos etarios, el algoritmo alcanzó sensibilidad y precisión de hasta 0.97, con MAE menor a un segundo en la mayoría de las subetapas. Se observó alta correlación (r > 0.90) con la segmentación manual en la duración total y fases estables, confirmando su robustez. Además, el análisis por grupo etario mostró un aumento progresivo en los tiempos de ejecución del TUG, subrayando el potencial de la herramienta para identificar variaciones funcionales asociadas a la edad. Estos resultados demuestran que la segmentación automática con IMU complementa y potencia la evaluación clínica, proporcionando métricas objetivas para el monitoreo y prevención del riesgo de caídas.
The Timed Up and Go (TUG) test is widely used to assess mobility and fall risk, especially in older adults and individuals with neurological conditions. However, its traditional approach, based solely on total completion time, has significant limitations, as it fails to capture qualitative aspects of movement such as stability during turns and coordination when standing up. The integration of inertial measurement units (IMU) and algorithms has enabled the segmentation of the TUG into specific subtasks, improving the detection of motor deficiencies. This thesis aims to validate and develop an automatic segmentation algorithm for the TUG using IMU data from individuals of different age ranges. The methodology involved a critical review of existing algorithms, evaluation of a rule-based model, and the development of an optimized approach through jury-based voting techniques, with validation against the OptiTrack system, the gold standard for motion capture. The algorithm was evaluated in terms of accuracy, sensitivity, mean absolute error, and correlation with manual segmentation. The goal is to contribute more precise and accessible tools for clinical assessment, enhancing the objectivity of TUG analysis and facilitating its integration into healthcare settings. In the validation conducted with 75 participants across different age groups, the proposed algorithm achieved sensitivity and precision of up to 0.97, with a mean absolute error below one second in most subtasks. High correlation (r > 0.90) was observed with manual segmentation for total test duration and stable phases, confirming the method’s robustness. Age group analysis showed a progressive increase in TUG times, highlighting the tool’s potential to identify age-related functional variations. These results demonstrate that automatic IMU-based segmentation can complement and enhance traditional clinical assessment, providing objective and detailed metrics for the monitoring and prevention of fall risk.
The Timed Up and Go (TUG) test is widely used to assess mobility and fall risk, especially in older adults and individuals with neurological conditions. However, its traditional approach, based solely on total completion time, has significant limitations, as it fails to capture qualitative aspects of movement such as stability during turns and coordination when standing up. The integration of inertial measurement units (IMU) and algorithms has enabled the segmentation of the TUG into specific subtasks, improving the detection of motor deficiencies. This thesis aims to validate and develop an automatic segmentation algorithm for the TUG using IMU data from individuals of different age ranges. The methodology involved a critical review of existing algorithms, evaluation of a rule-based model, and the development of an optimized approach through jury-based voting techniques, with validation against the OptiTrack system, the gold standard for motion capture. The algorithm was evaluated in terms of accuracy, sensitivity, mean absolute error, and correlation with manual segmentation. The goal is to contribute more precise and accessible tools for clinical assessment, enhancing the objectivity of TUG analysis and facilitating its integration into healthcare settings. In the validation conducted with 75 participants across different age groups, the proposed algorithm achieved sensitivity and precision of up to 0.97, with a mean absolute error below one second in most subtasks. High correlation (r > 0.90) was observed with manual segmentation for total test duration and stable phases, confirming the method’s robustness. Age group analysis showed a progressive increase in TUG times, highlighting the tool’s potential to identify age-related functional variations. These results demonstrate that automatic IMU-based segmentation can complement and enhance traditional clinical assessment, providing objective and detailed metrics for the monitoring and prevention of fall risk.
Description
Tesis presentada para optar al título de Ingeniero/a Civil Biomédico.
Keywords
Algoritmos, Enfermedades neuromusculares, Accidentes por Caídas, Adultos mayores