Auto detection and segmentation of daily living activities during a Timed Up and Go task in people with Parkinson's disease using multiple inertial sensors.

摘要: Wearable sensors have the potential to provide clinicians with access motor performance of people movement disorder as they undergo intervention. However, sensor data often be manually classified and segmented before can processed into clinical metrics. This process time consuming. We recently proposed detection segmentation algorithms based on peak using Inertial Measurement Units (IMUs) automatically identify isolate common activities during daily living such standing up, walking, turning, sitting down. These were developed a homogenous population healthy older adults. The aim this study was investigate transferability these in Parkinson’s disease (PD). A modified Timed Up And Go task used since it is comprised activities, all performed continuous fashion. Twelve adults diagnosed early PD (Hoehn & Yahr ≤ 2) recruited for three trials 10 5-m TUG OFF state. They outfitted 17 IMUs covering each body segment. Raw from detrended, normalized filtered reveal kinematics peaks that corresponded different activities. Segmentation accomplished by identifying first minimum or maximum right left peaks. times compared results two examiners who visually Specificity sensitivity evaluate accuracy algorithms. Using same previous study, we able detect 97.6% 92.7% specificity (n = 432) population. modifications selection, 100% accuracy. Similarly, applying population, within ~500 ms visual segmentation. Re-optimizing filtering frequencies, reduce difference ~400 ms. demonstrates agility system accurately segment disorders.