Semi-supervised clustering of quaternion time series: Application to gait analysis in multiple sclerosis using motion sensor data.

Recent approaches in gait analysis involve the use of wearable motion sensors to extract spatio-temporal parameters that characterize multiple aspects of an individual's gait. In particular, the medical community could largely benefit from this type of devices as they could provide the clinicians with a valuable tool for assessing gait impairment. Motion sensor data are however complex and there is an urgent unmet need to develop sound statistical methods for analyzing such data and extracting clinically relevant information. In this article, we measure gait by following the hip rotation over time and the resulting statistical unit is a time series of unit quaternions. We explore the possibility to form groups of patients with similar walking impairment by taking into account their walking data and their global decease severity with semi-supervised clustering. We generalize a compromise-based method named hclustcompro to unit quaternion time series by combining it with the proper dissimilarity quaternion dynamic time warping. We apply this method on patients diagnosed with multiple sclerosis to form groups of patients with similar walking deficiencies while accounting for the clinical assessment of their overall disability. We also compare the compromise-based clustering approach with the method mergeTrees that falls into a sub-class of ensemble clustering named collaborative clustering. The results provide a first proof of both the interest of using wearable motion sensors for assessing gait impairment and the use of prior knowledge to guide the clustering process. It also demonstrates that compromise-based clustering is a more appropriate approach in this context.

ROBOGait: A Mobile Robotic Platform for Human Gait Analysis in Clinical Environments.

Mobile robotic platforms have made inroads in the rehabilitation area as gait assistance devices. They have rarely been used for human gait monitoring and analysis. The integration of mobile robots in this field offers the potential to develop multiple medical applications and achieve new discoveries. This study proposes the use of a mobile robotic platform based on depth cameras to perform the analysis of human gait in practical scenarios. The aim is to prove the validity of this robot and its applicability in clinical settings. The mechanical and software design of the system is presented, as well as the design of the controllers of the lane-keeping, person-following, and servoing systems. The accuracy of the system for the evaluation of joint kinematics and the main gait descriptors was validated by comparison with a Vicon-certified system. Some tests were performed in practical scenarios, where the effectiveness of the lane-keeping algorithm was evaluated. Clinical tests with patients with multiple sclerosis gave an initial impression of the applicability of the instrument in patients with abnormal walking patterns. The results demonstrate that the system can perform gait analysis with high accuracy. In the curved sections of the paths, the knee joint is affected by occlusion and the deviation of the person in the camera reference system. This issue was greatly improved by adjusting the servoing system and the following distance. The control strategy of this robot was specifically designed for the analysis of human gait from the frontal part of the participant, which allows one to capture the gait properly and represents one of the major contributions of this study in clinical practice.

Person Re-Identification Using Deep Modeling of Temporally Correlated Inertial Motion Patterns.

Person re-identification (re-ID) is among the essential components that play an integral role in constituting an automated surveillance environment. Majorly, the problem is tackled using data acquired from vision sensors using appearance-based features, which are strongly dependent on visual cues such as color, texture, etc., consequently limiting the precise re-identification of an individual. To overcome such strong dependence on visual features, many researchers have tackled the re-identification problem using human gait, which is believed to be unique and provide a distinctive biometric signature that is particularly suitable for re-ID in uncontrolled environments. However, image-based gait analysis often fails to extract quality measurements of an individual's motion patterns owing to problems related to variations in viewpoint, illumination (daylight), clothing, worn accessories, etc. To this end, in contrast to relying on image-based motion measurement, this paper demonstrates the potential to re-identify an individual using inertial measurements units (IMU) based on two common sensors, namely gyroscope and accelerometer. The experiment was carried out over data acquired using smartphones and wearable IMUs from a total of 86 randomly selected individuals including 49 males and 37 females between the ages of 17 and 72 years. The data signals were first segmented into single steps and strides, which were separately fed to train a sequential deep recurrent neural network to capture implicit arbitrary long-term temporal dependencies. The experimental setup was devised in a fashion to train the network on all the subjects using data related to half of the step and stride sequences only while the inference was performed on the remaining half for the purpose of re-identification. The obtained experimental results demonstrate the potential to reliably and accurately re-identify an individual based on one's inertial sensor data.

Validation of Thigh Angle Estimation Using Inertial Measurement Unit Data against Optical Motion Capture Systems.

Inertial measurement units are commonly used to estimate the orientation of sections of sections of human body in inertial navigation systems. Most of the algorithms used for orientation estimation are computationally expensive and it is difficult to implement them in real-time embedded systems with restricted capabilities. This paper discusses a computationally inexpensive orientation estimation algorithm (Gyro Integration-Based Orientation Filter-GIOF) that is used to estimate the forward and backward swing angle of the thigh (thigh angle) for a vision impaired navigation aid. The algorithm fuses the accelerometer and gyroscope readings to derive the single dimension orientation in such a way that the orientation is corrected using the accelerometer reading when it reads gravity only or otherwise integrate the gyro reading to estimate the orientation. This strategy was used to reduce the drift caused by the gyro integration. The thigh angle estimated by GIOF was compared against the Vicon Optical Motion Capture System and reported a mean correlation of 99.58% for 374 walking trials with a standard deviation of 0.34%. The Root Mean Square Error (RMSE) of the thigh angle estimated by GIOF compared with Vicon measurement was 1.8477°. The computation time on an 8-bit microcontroller running at 8 MHz for GIOF is about a half of that of Complementary Filter implementation. Although GIOF was only implemented and tested for estimating pitch of the IMU, it can be easily extended into 2D to estimate both pitch and roll.

Human Pose Estimation for Clinical Analysis of Gait Pathologies.

Gait analysis serves as a critical diagnostic tool for identifying neurologic and musculoskeletal damage. Traditional manual analysis of motion data, however, is labor-intensive and heavily reliant on the expertise and judgment of the therapist. This study introduces a binary classification method for the quantitative assessment of gait impairments, specifically focusing on Duchenne muscular dystrophy (DMD), a prevalent and fatal neuromuscular genetic disorder. The research compares spatiotemporal and sagittal kinematic gait features derived from 2D and 3D human pose estimation trajectories against concurrently recorded 3D motion capture (MoCap) data from healthy children. The proposed model leverages a novel benchmark dataset, collected from YouTube and publicly available datasets of their typically developed peers, to extract time-distance variables (e.g. speed, step length, stride time, and cadence) and sagittal joint angles of the lower extremity (e.g. hip, knee, and knee flexion angles). Machine learning and deep learning techniques are employed to discern patterns that can identify children exhibiting DMD gait disturbances. While the current model is capable of distinguishing between healthy subjects and those with DMD, it does not specifically differentiate between DMD patients and patients with other gait impairments. Experimental results validate the efficacy of our cost-effective method, which relies on recorded RGB video, in detecting gait abnormalities, achieving a prediction accuracy of 96.2% for Support Vector Machine (SVM) and 97% for the deep network.

Petri net transition times as training features for multiclass models to support the detection of neurodegenerative diseases.

This paper proposes the transition times of Petri net models of human gait as training features for multiclass random forests (RFs) and classification trees (CTs). These models are designed to support screening for neurodegenerative diseases. The proposed Petri net describes gait in terms of nine cyclic phases and the timing of the nine events that mark the transition between phases. Since the transition times between strides vary, each is represented as a random variable characterized by its mean and standard deviation. These transition times are calculated using the PhysioNet database of vertical ground reaction forces (VGRFs) generated by feet-ground contact. This database comprises the VGRFs of four groups: amyotrophic lateral sclerosis, the control group, Huntington's disease, and Parkinson disease. The RF produced an overall classification accuracy of 91%, and the specificities and sensitivities for each class were between 80% and 100%. However, despite this high performance, the RF-generated models demonstrated lack of interpretability prompted the training of a CT using identical features. The obtained tree comprised only four features and required a maximum of three comparisons. However, this simplification dramatically reduced the overall accuracy from 90.6% to 62.3%. The proposed set features were compared with those included in PhysioNet database of VGRFs. In terms of both the RF and CT, more accurate models were established using our features than those of the PhysioNet.

A comprehensive survey on gait analysis: History, parameters, approaches, pose estimation, and future work.

Human gait is a periodic motion of body segments-the analysis of motion and related studies is termed gait analysis. Gait Analysis has gained much popularity because of its applications in clinical diagnosis, rehabilitation methods, gait biometrics, robotics, sports, and biomechanics. Traditionally, subjective assessment of the gait was conducted by health experts; however, with the advancement in technology, gait analysis can now be performed objectively and empirically for better and more reliable assessment. State-of-the-art semi-subjective and objective techniques for gait analysis have limitations that can be mitigated using advanced machine learning-based approaches. This paper aims to provide a narrative and a comprehensive analysis of cutting-edge gait analysis techniques and insight into clinical gait analysis. The literature of the previous surveys during the last decade is discussed. This paper presents an elaborated schema, including gait analysis history, parameters, machine learning approaches for marker-based and marker-less analysis, applications, and performance measures. This paper also explores the pose estimation techniques for clinical gait analysis that open future research directions in this area.

A Survey of Human Gait-Based Artificial Intelligence Applications.

We performed an electronic database search of published works from 2012 to mid-2021 that focus on human gait studies and apply machine learning techniques. We identified six key applications of machine learning using gait data: 1) Gait analysis where analyzing techniques and certain biomechanical analysis factors are improved by utilizing artificial intelligence algorithms, 2) Health and Wellness, with applications in gait monitoring for abnormal gait detection, recognition of human activities, fall detection and sports performance, 3) Human Pose Tracking using one-person or multi-person tracking and localization systems such as OpenPose, Simultaneous Localization and Mapping (SLAM), etc., 4) Gait-based biometrics with applications in person identification, authentication, and re-identification as well as gender and age recognition 5) "Smart gait" applications ranging from smart socks, shoes, and other wearables to smart homes and smart retail stores that incorporate continuous monitoring and control systems and 6) Animation that reconstructs human motion utilizing gait data, simulation and machine learning techniques. Our goal is to provide a single broad-based survey of the applications of machine learning technology in gait analysis and identify future areas of potential study and growth. We discuss the machine learning techniques that have been used with a focus on the tasks they perform, the problems they attempt to solve, and the trade-offs they navigate.

Millimeter wave radar data of people walking.

This dataset contains complex signals coming from a mmWave FMCW radar system. Signals were acquired during a measurement campaign taken indoor and aimed to assess people's different ways of walking. Measurement setup and devices are described. The dataset consists of the acquisitions of six different types of activities, performed by 29 subjects who repeat each activity several times. Therefore, the dataset contains multiple different experiments for each activity, for a total of 231 acquisitions. The subjects walk without any constraint or do not follow any pattern, thus making this dataset useful not only for human gait recognition but also to evaluate the performance of different radar data processing algorithms.

Ensembling rules in automatic analysis of pressure on plantar surface in children with pes planovalgus.

PURPOSE: This paper presents a method of ensembling rules obtained through induction of several basic types of decision trees. MATERIAL AND METHODS: The proposed method uses rules generated by means of well-known decision trees: CART, CHAID, exhaustive CHAID and C4.5. The method was tested on data describing pressure distribution under foot during gait in children with pes planovalgus (PV) and typical foot. Children with pes planovalgus underwent surgical intervention and were re-examined. Overall, 316 gait cycles have been used in analysis. RESULTS: The obtained results consist of a set of rules for all considered cases and show that the proposed method may be a useful tool of gait analysis on the basis of parameters that have a physical interpretation. CONCLUSIONS: The presented method for mining rules useful in this respect may be successfully used by persons with a typically medical knowledge and could improve the understanding of the human gait phenomenon. There is obviously no reason why this method could not be used in the case of other data as well.

Functional, impulse-based quantification of plantar pressure patterns in typical adult gait.

BACKGROUND: Dynamic pedobarography is used to measure the change in plantar pressure distribution during gait. Clinical methods of pedobarographic analysis lack, however, a standardized, functional segmentation or require costly motion capture technology and expertise. Furthermore, while commonly used pedobarographic measures are mostly based on peak pressures, progressive foot deformities also depend on the duration the pressure is applied, which can be quantified via impulse measures. RESEARCH QUESTION: Our objectives were to: (1) develop a standardized method for functionally segmenting pedobarographic data during gait without the need for motion capture; (2) compute pedobarographic measures that are based on each segment's vertical impulse; and (3) obtain a normative set of such pedobarographic measures for non-disabled gait. METHODS: Pedobarographic data was collected during gait from sixty adults with normal feet. Using the maximum pressure map for each trial, an expert and novice rater independently identified the hallux, heel, medial forefoot, and lateral forefoot and computed nine normalized vertical impulse measures. RESULTS: From the computed impulse measures, the Heel-to-Forefoot Balance was 33.3 ± 5.5%, the Medial-Lateral Forefoot Balance (with hallux) 59.2 ± 8.0%, the Medial-Lateral Forefoot Balance (without hallux) 53.5 ± 7.7%, and the Hallux-to-Medial Forefoot Balance 21.0 ± 8.9% (mean ± standard deviation). The intra- and inter-rater reliability ranged between 0.93 and 1.00 and between 0.89 and 0.99, respectively (ICC(2,1)). SIGNIFICANCE: We developed a simple, stand-alone method for pedobarographic segmentation that is mechanistically linked to relevant anatomical regions of the foot. The normative impulse measures exhibited excellent reliability. This normative dataset is currently used in the clinical assessment of different foot deformities and gait impairments, and in the evaluation of treatment outcomes.

Three-dimensional kinematic analysis of upper and lower limb motion during gait of post-stroke patients

The aim of this study was to analyze the alterations of arm and leg movements of patients during stroke gait. Joint angles of upper and lower limbs and spatiotemporal variables were evaluated in two groups: hemiparetic group (HG, 14 hemiparetic men, 53 ± 10 years) and control group (CG, 7 able-bodied men, 50 ± 4 years). The statistical analysis was based on the following comparisons (P ≤ 0.05): 1) right versus left sides of CG; 2) affected (AF) versus unaffected (UF) sides of HG; 3) CG versus both the affected and unaffected sides of HG, and 4) an intracycle comparison of the kinematic continuous angular variables between HG and CG. This study showed that the affected upper limb motion in stroke gait was characterized by a decreased range of motion of the glenohumeral (HG: 6.3 ± 4.5, CG: 20.1 ± 8.2) and elbow joints (AF: 8.4 ± 4.4, UF: 15.6 ± 7.6) on the sagittal plane and elbow joint flexion throughout the cycle (AF: 68.2 ± 0.4, CG: 46.8 ± 2.7). The glenohumeral joint presented a higher abduction angle (AF: 14.2 ± 1.6, CG: 11.5 ± 4.0) and a lower external rotation throughout the cycle (AF: 4.6 ± 1.2, CG: 22.0 ± 3.0). The lower limbs showed typical alterations of the stroke gait patterns. Thus, the changes in upper and lower limb motion of stroke gait were identified. The description of upper limb motion in stroke gait is new and complements gait analysis.