Smartphones are often equipped with inertial sensors capable of measuring individuals' physical activities. Their role in monitoring the patients' physical activities in telemedicine, however, needs to be explored. The main objective of this study was to explore the correlation between a participant's daily step counts and the daily step counts reported by their smartphone. This prospective observational study was conducted on patients undergoing lower limb orthopedic surgery and a group of non-patients. The data collection period was from 2 weeks before until four weeks after the surgery for the patients and two weeks for the non-patients. The participants' daily steps were recorded by physical activity trackers employed 24/7, and an application recorded the number of daily steps registered by the participants' smartphones. We compared the cross-correlation between the daily steps time-series taken from the smartphones and physical activity trackers in different groups of participants. We also employed mixed modeling to estimate the total number of steps. Overall, 1067 days of data were collected from 21 patients (11 females) and 10 non-patients (6 females). The cross-correlation coefficient between the smartphone and physical activity tracker was 0.70 [0.53–0.83]. The correlation in the non-patients was slightly higher than in the patients (0.74 [0.60–0.90] and 0.69 [0.52–0.81], respectively). Considering the ubiquity, convenience, and practicality of smartphones, the high correlation between the smartphones and the total daily step time-series highlights the potential usefulness of smartphones in detecting the change in the step counts in remote monitoring of the patient's physical activity

Physiotherapy is a critical element in successful conservative management of low back pain (LBP). The aim of this study was to develop and evaluate a system with wearable inertial sensors to objectively detect sitting postures and performance of unsupervised exercises containing movement in multiple planes (flexion, extension, rotation). A set of 8 inertial sensors were placed on 19 healthy adult subjects. Data was acquired as they performed 7 McKenzie low-back exercises and 3 sitting posture positions. This data was used to train two models (Random Forest (RF) and XGBoost (XGB)) using engineered time series features. In addition, a convolutional neural network (CNN) was trained directly on the time series data. A feature importance analysis was performed to identify sensor locations and channels that contributed most to the models. Finally, a subset of sensor locations and channels was included in a hyperparameter grid search to identify the optimal sensor configuration and the best performing algorithm(s) for exercise classification. Models were evaluated using F1-score in a 10-fold cross validation approach. The optimal hardware configuration was identified as a 3-sensor setup using lower back, left thigh, and right ankle sensors with acceleration, gyroscope, and magnetometer channels. The XBG model achieved the highest exercise (F1=0.94±0.03) and posture (F1=0.90±0.11) classification scores. The CNN achieved similar results with the same sensor locations, using only the accelerometer and gyroscope channels for exercise classification (F1=0.94±0.02) and the accelerometer channel alone for posture classification (F1=0.91±0.03). This study demonstrates the potential of a 3-sensor lower body wearable solution (e.g. smart pants) that can identify proper sitting postures and exercises in multiple planes, suitable for low back pain. This technology has the potential to improve the effectiveness of LBP rehabilitation by facilitating quantitative feedback, early problem diagnosis, and possible remote monitoring

Wearable inertial sensors can detect abnormal gait associated with knee or hip osteoarthritis (OA). However, few studies have compared sensor-derived gait parameters between patients with hip and knee OA or evaluated the efficacy of sensors suitable for remote monitoring in distinguishing between the two. Hence, our study seeks to examine the differences in accelerations captured by low-frequency wearable sensors in patients with knee and hip OA and classify their gait patterns. We included patients with unilateral hip and knee OA. Gait analysis was conducted using an accelerometer ipsilateral with the affected joint on the lateral distal thighs. Statistical parametric mapping (SPM) was used to compare acceleration signals. The k-Nearest Neighbor (k-NN) algorithm was trained on 80% of the signals' Fourier coefficients and validated on the remaining 20% using 10-fold cross-validation to classify the gait patterns into hip and knee OA. We included 42 hip OA patients (19 females, age 70 [63–78], BMI of 28.3 [24.8–30.9]) and 59 knee OA patients (31 females, age 68 [62–74], BMI of 29.7 [26.3–32.6]). The SPM results indicated that one cluster (12–20%) along the vertical axis had accelerations exceeding the critical threshold of 2.956 (p=0.024). For the anteroposterior axis, three clusters were observed exceeding the threshold of 3.031 at 5–19% (p = 0.0001), 39–54% (p=0.00005), and 88–96% (p = 0.01). Regarding the mediolateral axis, four clusters were identified exceeding the threshold of 2.875 at 0–9% (p = 0.02), 14–20% (p=0.04), 28–68% (p < 0.00001), and 84–100% (p = 0.004). The k-NN model achieved an AUC of 0.79, an accuracy of 80%, and a precision of 85%. In conclusion, the Fourier coefficients of the signals recorded by wearable sensors can effectively discriminate the gait patterns of knee and hip OA. In addition, the most remarkable differences in the time domain were observed along the mediolateral axis

The tendency towards using inertial sensors for remote monitoring of the patients at home is increasing. One of the most important characteristics of the sensors is sampling rate. Higher sampling rate results in higher resolution of the sampled signal and lower amount of noise. However, higher sampling frequency comes with a cost. The main aim of our study was to determine the validity of measurements performed by low sampling frequency (12.5 Hz) accelerometers (SENS) in patients with knee osteoarthritis compared to standard sensor-based motion capture system (Xsens). We also determined the test-retest reliability of SENS accelerometers. Participants were patients with unilateral knee osteoarthritis. Gait analysis was performed simultaneously by using Xsens and SENS sensors during two repetitions of over-ground walking at a self-selected speed. Gait data from Xsens were used as an input for AnyBody musculoskeletal modeling software to measure the accelerations at the exact location of two defined virtual sensors in the model (VirtualSENS). After preprocessing, the signals from SENS and VirtualSENS were compared in different coordinate axes in time and frequency domains. ICC for SENS data from first and second trials were calculated to assess the repeatability of the measurements. We included 32 patients (18 females) with median age 70.1[48.1 – 85.4]. Mean height and weight of the patients were 173.2 ± 9.6 cm and 84.2 ± 14.7 kg respectively. The correlation between accelerations in time domain measured by SENS and VirtualSENS in different axes was r = 0.94 in y-axis (anteroposterior), r = 0.91 in x-axis (vertical), r = 0.83 in z-axis (mediolateral), and r = 0.89 for the magnitude vector. In frequency domain, the value and the power of fundamental frequencies (F. 0. ) of SENS and VirtualSENS signals demonstrated strong correlation (r = 0.98 and r = 0.99 respectively). The result of test-retest evaluation showed excellent repeatability for acceleration measurement by SENS sensors. ICC was between 0.89 to 0.94 for different coordinate axes. Low sampling frequency accelerometers can provide valid and reliable measurements especially for home monitoring of the patients, in which handling big data and sensors cost and battery lifetime are among important issues