Development and clinical validation of a stroke-specific Gait Deviation Index
Stroke affects over 101 million individuals worldwide, leaving many to cope with long-term consequences. These often include physical and cognitive impairments such as muscle weakness, loss of coordination, and challenges in performing daily activities. Several quantitative indices have been develop...
| Autores: | , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2026 |
| País: | España |
| Institución: | Consejo Superior de Investigaciones Científicas (CSIC) |
| Repositorio: | DIGITAL.CSIC. Repositorio Institucional del CSIC |
| OAI Identifier: | oai:dnet:digitalcsic_::a3c98ae12244326f09fe0522ab83b1c2 |
| Acceso en línea: | http://hdl.handle.net/10261/429835 |
| Access Level: | acceso abierto |
| Palabra clave: | Gait Deviation Index Stroke Gait Kinematics Regression |
| Sumario: | Stroke affects over 101 million individuals worldwide, leaving many to cope with long-term consequences. These often include physical and cognitive impairments such as muscle weakness, loss of coordination, and challenges in performing daily activities. Several quantitative indices have been developed to objectively quantify gait deviations with a focus on ensuring ease of use and accessibility within clinical settings. Among those, the most cited one is the Gait Deviation Index (GDI), originally developed with data from a population of children with cerebral palsy. This characteristic probably reduces the ability of the GDI to interpret stroke-specific deviations. To overcome this limitation, one possibility is to derive a GDI for each specific population, as demonstrated for spinal cord injury. In this study, we developed a stroke-specific GDI (STR-GDI) considering a database of 22 healthy controls and 69 post-stroke subjects. The STR-GDI with a 17-feature basis provided a more reliable reconstruction of non-native data compared to the 15-feature original GDI basis. Both GDI and STR-GDI showed excellent reliability across limbs and datasets (ICC = 0.99), although STR-GDI showed higher ICCs in the test set for the non-paretic limbs (0.95 compared to 0.91 for the GDI). Regression models between GDI and STR-GDI yielded R-2 values of 0.76 and 0.81, respectively for the paretic and non-paretic limbs. The linear regressions between indexes and clinical and gait variables were less consistent but still meaningful, with R-2 values lower than 0.23. These results may indicate that these metrics offer complementary information, and their combined use could provide a more comprehensive understanding of post-stroke gait impairments. STR-GDI showed slightly higher correlations than GDI with the lower-limb Fugl-Meyer Assessment, the percentage of stance phase, step length, and cadence. In conclusion, in this study we successfully adapted the original GDI methodology to create a STR-GDI to capture gait deviations specific to post-stroke individuals. This study makes a meaningful methodological contribution by developing and validating a complete automatic pipeline that could be adapted to generate other condition-specific GDIs for other neurological or musculoskeletal disorders, provided that datasets are available. Future work will expand the STR-GDI using a larger and more diverse stroke population, assess its inter-session reliability, and develop a compatible version for inertial measurement units and markerless motion capture systems, facilitating broader clinical and real-world applications. |
|---|