Assessing inertial measurement unit locations for freezing of gait detection and patient preference

Abstract

Background Freezing of gait, a common symptom of Parkinson’s disease, presents as sporadic episodes in which an individual’s feet suddenly feel stuck to the ground. Inertial measurement units (IMUs) promise to enable at-home monitoring and personalization of therapy, but there is a lack of consensus on the number and location of IMUs for detecting freezing of gait. The purpose of this study was to assess IMU sets in the context of both freezing of gait detection performance and patient preference. Methods Sixteen people with Parkinson’s disease were surveyed about sensor preferences. Raw IMU data from seven people with Parkinson’s disease, wearing up to eleven sensors, were used to train convolutional neural networks to detect freezing of gait. Models trained with data from different sensor sets were assessed for technical performance; a best technical set and minimal IMU set were identified. Clinical utility was assessed by comparing model- and human-rater-determined percent time freezing and number of freezing events. Results The best technical set consisted of three IMUs (lumbar and both ankles, AUROC = 0.83), all of which were rated highly wearable. The minimal IMU set consisted of a single ankle IMU (AUROC = 0.80). Correlations between these models and human raters were good to excellent for percent time freezing (ICC = 0.93, 0.89) and number of freezing events (ICC = 0.95, 0.86) for the best technical set and minimal IMU set, respectively. Conclusions Several IMU sets consisting of three IMUs or fewer were highly rated for both technical performance and wearability, and more IMUs did not necessarily perform better in FOG detection. We openly share our data and software to further the development and adoption of a general, open-source model that uses raw signals and a standard sensor set for at-home monitoring of freezing of gait.

Walking Speed Classification from Marker-Free Video Images in Two-Dimension Using Optimum Data and a Deep Learning Method

“…Second, the performance (in terms of the mean ± SD classification accuracy and mean ± SD training time) of a biLSTM deep learning-based walking speed classification model was evaluated using walking speed patterns created using all possible combination of one, two, three and four out of five ratio-based body measurements. The combination with the fewest ratio-based body measurements (i.e., less than five ratio-based body measurements) for the establishment of walk patterns was deemed optimal if it yielded a mean ± SD classification accuracy higher than or within 2% less [ 23 , 24 ] of the mean ± SD classification accuracy obtained in our previous study [ 13 ], and the ratio-based body measurements used for defining the walk pattern exhibited low correlations among them.…”

“…This study also evaluated the performance (in terms of the mean ± SD classification accuracy and mean ± SD training time) of a biLSTM deep learning-based walking speed classification model using the walking speed patterns created by all possible combinations of one, two, three and four ratio-based body measurements among five ratio-based body measurements (HW1, HW2, HW3, A1, and A2). The walk pattern created by the combination of fewest ratio-based body measurements (i.e., less than five ratio-based body measurements) was defined as optimal in the study if it was able to classify the walking speed with a mean ± SD classification accuracy higher than or within 2% less [ 23 , 24 ] of that obtained in our previous study [ 13 ], and the ratio-based body measurements in the walk pattern showed low correlations among them. This study hypothesized that walking speed patterns identified from few ratio-based body measurements can be used to classify walking speed using deep learning-based methods with high accuracy if the correlations among the body measurements are low.…”

Smartwatch-based prediction of single-stride and stride-to-stride gait outcomes using regression-based machine learning

“…IMUs can also be instrumented on multiple lower limb segments and combined with biomechanical constraints to model foot trajectories and extract spatiotemporal parameters 2,41 . A major challenge with each of these IMU approaches is that the sensors are not positioned on locations preferred by most individuals, as exemplified in a recent investigation of persons with Parkinson's disease 32 , questioning the feasibility of these IMU positions for long-term continuous monitoring.…”

“…Smartwatches are widely available and continue to grow in popularity as a smart wearable device due, in part, to perceived usefulness, enjoyment, and ease of use 31 . In conjunction with machine learning techniques, wrist-based IMUs have been used for gait recognition 21 , freezing of gait detection 28,32 , fall detection 43 , and spatiotemporal feature estimation 12 . For example, Erdem and colleagues 12 used regression-based machine learning on linear acceleration and angular velocity features of smartwatches worn on both wrists to predict step length, swing time, and stance time to within 5.3 cm, 0.05 s, and 0.09 s, respectively.…”

Machine Learning Approach to Support the Detection of Parkinson’s Disease in IMU-Based Gait Analysis

“…When assessing the gait of pwPD using IMUs, reducing the number of wearable devices to a single lumbar-mounted sensor allows for sufficient identification of gait abnormalities without significant information loss, reducing the wearability burden in nonlaboratory conditions [ 5 , 37 , 38 ]. Although machine learning algorithms are increasingly being used for gait analysis [ 39 ], few studies have applied ML-based classification to gait data derived from a single lumbar-mounted IMU to assess gait abnormalities of pwPD [ 10 , 36 , 39 , 40 , 41 , 42 , 43 ], and few studies have compared the classification performances of ML algorithms using lower trunk acceleration data in pwPD [ 36 , 44 ]. Comparing the performance of the most commonly used ML algorithms on trunk acceleration-derived gait data could provide useful information on which to use in clinical settings and which to implement into gait detection software.…”