Josef Christian1, Josef Kröll2, Gerda Strutzenberger3, Nathalie Alexander4, Michael Ofner5, Hermann Schwameder6. 1. Department of Sports Science and Kinesiology, University of Salzburg, Schlossallee 49, 5400 Hallein-Rif, Austria. Electronic address: josef.christian@sbg.ac.at. 2. Department of Sports Science and Kinesiology, University of Salzburg, Schlossallee 49, 5400 Hallein-Rif, Austria. Electronic address: josef.kroell@sbg.ac.at. 3. Department of Sports Science and Kinesiology, University of Salzburg, Schlossallee 49, 5400 Hallein-Rif, Austria. Electronic address: gerda.strutzenberger@sbg.ac.at. 4. Department of Sports Science and Kinesiology, University of Salzburg, Schlossallee 49, 5400 Hallein-Rif, Austria. Electronic address: nathalie.alexander@sbg.ac.at. 5. Institute of Pathophysiology and Immunology, Medical University of Graz, Heinrichstraße 31a, 8010 Graz, Austria. Electronic address: michael.ofner@medyco.net. 6. Department of Sports Science and Kinesiology, University of Salzburg, Schlossallee 49, 5400 Hallein-Rif, Austria. Electronic address: hermann.schwameder@sbg.ac.at.
Abstract
BACKGROUND: Gait analysis is a useful tool to evaluate the functional status of patients with anterior cruciate ligament injury. Pattern recognition methods can be used to automatically assess walking patterns and objectively support clinical decisions. This study aimed to test a pattern recognition system for analyzing kinematic gait patterns of recently anterior cruciate ligament injured patients and for evaluating the effects of a therapeutic treatment. METHODS: Gait kinematics of seven male patients with an acute unilateral anterior cruciate ligament rupture and seven healthy males were recorded. A support vector machine was trained to distinguish the groups. Principal component analysis and recursive feature elimination were used to extract features from 3D marker trajectories. A Classifier Oriented Gait Score was defined as a measure of gait quality. Visualizations were used to allow functional interpretations of characteristic group differences. The injured group was evaluated by the system after a therapeutic treatment. The results were compared against a clinical rating of the patients' gait. FINDINGS: Cross validation yielded 100% accuracy. After the treatment the score improved significantly (P<0.01) as well as the clinical rating (P<0.05). The visualizations revealed characteristic kinematic features, which differentiated between the groups. INTERPRETATION: The results show that gait alterations in the early phase after anterior cruciate ligament injury can be detected automatically. The results of the automatic analysis are comparable with the clinical rating and support the validity of the system. The visualizations allow interpretations on discriminatory features and can facilitate the integration of the results into the diagnostic process.
BACKGROUND: Gait analysis is a useful tool to evaluate the functional status of patients with anterior cruciate ligament injury. Pattern recognition methods can be used to automatically assess walking patterns and objectively support clinical decisions. This study aimed to test a pattern recognition system for analyzing kinematic gait patterns of recently anterior cruciate ligament injured patients and for evaluating the effects of a therapeutic treatment. METHODS: Gait kinematics of seven male patients with an acute unilateral anterior cruciate ligament rupture and seven healthy males were recorded. A support vector machine was trained to distinguish the groups. Principal component analysis and recursive feature elimination were used to extract features from 3D marker trajectories. A Classifier Oriented Gait Score was defined as a measure of gait quality. Visualizations were used to allow functional interpretations of characteristic group differences. The injured group was evaluated by the system after a therapeutic treatment. The results were compared against a clinical rating of the patients' gait. FINDINGS: Cross validation yielded 100% accuracy. After the treatment the score improved significantly (P<0.01) as well as the clinical rating (P<0.05). The visualizations revealed characteristic kinematic features, which differentiated between the groups. INTERPRETATION: The results show that gait alterations in the early phase after anterior cruciate ligament injury can be detected automatically. The results of the automatic analysis are comparable with the clinical rating and support the validity of the system. The visualizations allow interpretations on discriminatory features and can facilitate the integration of the results into the diagnostic process.
Authors: Fabian Horst; Sebastian Lapuschkin; Wojciech Samek; Klaus-Robert Müller; Wolfgang I Schöllhorn Journal: Sci Rep Date: 2019-02-20 Impact factor: 4.379