Selection of gait parameters for modified Gillette Gait Index using Hellwig Correlation Based Filter method, random forest method, and correlation methods

• Published in: Biocybernetics and biomedical engineering Vol. 40; no. 3; pp. 1267 - 1276
• Main Authors: Syczewska, Małgorzata, Kocel, Krzysztof, Święcicka, Anna, Graff, Krzysztof, Krawczyk, Maciej, Wąsiewicz, Piotr, Kalinowska, Małgorzata, Szczerbik, Ewa
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2020
• Subjects: Advanced Basic Science; Gait parameters; Gillette Gait Index; Hellwig method; Internal Medicine; Key words; Random forest; Key words; Random forest; Hellwig method; Gait parameters; Gillette Gait Index
• ISSN: 0208-5216
• DOI: 10.1016/j.bbe.2020.07.002

Objective gait analysis provide a large number of data, which are used for planning further treatment of the patient. Data from groups of patients are used for comparisons of different treatment methods, assessment of the severity of gait deviations, design of classification systems. The Gilette Gait Index (GGI) was designed to express the level of abnormality of the gait pattern of patients with cerebral palsy by one number: a measure of distance between the set of discrete gait parameters of a patient from a similar set of a healthy subject, based on 16 parameters. Gait pathology in other disorders is different, thus other variables may better describe their level of pathology. The aim was to see if modified GGI can be constructed based on other sets of gait variables. To decide which gait variables should be taken three different analytical methods were used: Hellwig Correlation Based Filter, random forest, and correlation methods. Gait data of 84 patients were retrospectively selected: 30 had spastic cerebral palsy, 24 scoliosis, 30 suffered the stroke. The results show, that in the final sets of the 16 parameters chosen by the analyses there are some parameters, which were not present in the original list of GGI. If the number of sixteen parameters is kept, there are more than one optimal set of parameters. In conclusion, the use of analytical methods enabled the choice of sets of 16 gait parameters which are specific for the medical problem, and the calculation of modified GGIs.