Algorithm for Turning Detection and Analysis Validated under Home-Like Conditions in Patients with Parkinson’s Disease and Older Adults using a 6 Degree-of-Freedom Inertial Measurement Unit at the Lower Back

Abstract

IntroductionAging and age-associated disorders such as Parkinson’s disease (PD) are often associated with turning difficulties, which can lead to falls and fractures. Valid assessment of turning and turning deficits specifically in non-standardized environments may foster specific treatment and prevention of consequences.MethodsRelative orientation, obtained from 3D-accelerometer and 3D-gyroscope data of a sensor worn at the lower back, was used to develop an algorithm for turning detection and qualitative analysis in PD patients and controls in non-standardized environments. The algorithm was validated with a total of 2,304 turns ≥90° extracted from an independent dataset of 20 PD patients during medication ON- and OFF-conditions and 13 older adults. Video observation by two independent clinical observers served as gold standard.ResultsIn PD patients under medication OFF, the algorithm detected turns with a sensitivity of 0.92, a specificity of 0.89, and an accuracy of 0.92. During medication ON, values were 0.92, 0.78, and 0.83. In older adults, the algorithm reached validation values of 0.94, 0.89, and 0.92. Turning magnitude (difference, 0.06°; SEM, 0.14°) and duration (difference, 0.004 s; SEM, 0.005 s) yielded high correlation values with gold standard. Overall accuracy for direction of turning was 0.995. Intra class correlation of the clinical observers was 0.92.ConclusionThis wearable sensor- and relative orientation-based algorithm yields very high agreement with clinical observation for the detection and evaluation of ≥90° turns under non-standardized conditions in PD patients and older adults. It can be suggested for the assessment of turning in daily life.

Motor, cognitive and mobility deficits in 1000 geriatric patients: protocol of a quantitative observational study before and after routine clinical geriatric treatment – the ComOn-study

“…We will also use modern technology for the assessment of movement deficits, including but not limited to gait, balance, transfers, sleep and mobility. We will only apply algorithms that are validated for these populations for the extraction and appraisal of movement episodes and mobility patterns (e.g., [123][124][125][126]). It is expected that yet unknown symptoms will be detected that are not visible with the usual "clinical eye" [30].…”

Application of the ActiGraph GT9X IMU for the assessment of turning during walking and running

“…However, turn count error was lowest for the 0.25 Hz filter and it was the only gyroscope filter frequency that did not have significant main effects or interactions for detecting the number of turns. Previous studies have achieved turn detection sensitivities and specificities greater than 76% using a low-pass filter (18) or device orientation estimates via Kalman filtering (55). El-Gohary et al (18) low-pass filtered IMU data from the lower-back at 1.5 Hz to detect turns of 45°, 90°, 135°, and 180° during walking.…”

“…Cross-validation against 3Dmotion analysis resulted in a sensitivity and specificity of 90% and 76%, respectively. Pham et al (55) processed raw IMU data from the lower-back with a Kalman filter to detect turns completed during ADL. Cross-validation against direct observation resulted in a sensitivity and specificity of 94% and 89%, respectively.…”

“…Turn degree was estimated within 2.2° of the measured turn degree when using the gyroscope. Pham et al (55) estimated turn degree within 0.06° using an IMU placed on the lower back, though they used a Kalman filter to estimate position. By using a Kalman filter, they utilize all components of the IMU to estimate position.…”

“…An IMU is often used in navigation systems (63), though has received growing interest for the assessment of human movement such as gait analysis (65), postural orientation (45), and fall risk prevention (24) . The gyroscope and magnetometer components have also been used to detect turns and quantify turning characteristics (53,55,68), though there are limited studies that examine their application for estimating EE. walking, jogging, biking, and stair ascent and descent.…”