Automatic gait events detection with inertial measurement units: healthy subjects and moderate to severe impaired patients

 With this your competent? doctor could objectively determine the problems in your gait. Thus pointing DIRECTLY  to those EXACT stroke protocols that fix those problems.  At least this would be possible if there was any competency anywhere in the stroke medical world.

Automatic gait events detection with inertial measurement units: healthy subjects and moderate to severe impaired patients

• Cyril Voisard,
• Nicolas de l’Escalopier,
• Damien Ricard &
• Laurent Oudre 

Journal of NeuroEngineering and Rehabilitation volume 21, Article number: 104 (2024) Cite this article

• 30 Accesses

• Metrics details

Abstract

Background

Recently, the use of inertial measurement units (IMUs) in quantitative gait analysis has been widely developed in clinical practice. Numerous methods have been developed for the automatic detection of gait events (GEs). While many of them have achieved high levels of efficiency in healthy subjects, detecting GEs in highly degraded gait from moderate to severely impaired patients remains a challenge. In this paper, we aim to present a method for improving GE detection from IMU recordings in such cases.

Methods

We recorded 10-meter gait IMU signals from 13 healthy subjects, 29 patients with multiple sclerosis, and 21 patients with post-stroke equino varus foot. An instrumented mat was used as the gold standard. Our method detects GEs from filtered acceleration free from gravity and gyration signals. Firstly, we use autocorrelation and pattern detection techniques to identify a reference stride pattern. Next, we apply multiparametric Dynamic Time Warping to annotate this pattern from a model stride, in order to detect all GEs in the signal.

Results

We analyzed 16,819 GEs recorded from healthy subjects and achieved an F1-score of 100%, with a median absolute error of 8 ms (IQR [3–13] ms). In multiple sclerosis and equino varus foot cohorts, we analyzed 6067 and 8951 GEs, respectively, with F1-scores of 99.4% and 96.3%, and median absolute errors of 18 ms (IQR [8–39] ms) and 26 ms (IQR [12–50] ms).

Conclusions

Our results are consistent with the state of the art for healthy subjects and demonstrate a good accuracy in GEs detection for pathological patients. Therefore, our proposed method provides an efficient way to detect GEs from IMU signals, even in degraded gaits. However, it should be evaluated in each cohort before being used to ensure its reliability.

Background

The study of gait in medicine is an essential tool for evaluating the health and progression of patients with various diseases [1]. In this context, quantitative gait analysis can be used to finely evaluate the patient’s functional abilities, track the progression of their disease, measure the effectiveness of treatments, and develop personalized rehabilitation plans [2, 3].

In recent years, inertial measurement units (IMUs) have become widely developed in gait analysis due to their compact size, low cost, and ease of integration [4, 5]. They allow for objective and quantitative gait analysis, easy to use in healthy subjects, athletes [6] and patients, for example with neurological or orthopedic diseases [7]. They allow therefore the measurement of indicators on the gait semiology of patients, such as speed, stride length, double-support time, and balance [8]. Some of these features depend on the segmentation of strides and steps. Based on the historical description of gait in healthy subjects, 4 GE occur in a stride, in this order: Heel-Off (HO), Toe-Off (TO), Heel-Strike (HS), and Foot-Flat (FF) [9, 10]. Two main phases are described: the Stance Phase (StP) when the foot is on the ground between HS and TO, and the Swing Phase (SwP) when the foot is in the air between TO and HS [11, 12].

To accurately identify GEs from IMU signals, many techniques have therefore been developed. Currently, automatic detection of GEs on IMU recordings in healthy subjects has achieved a high degree of accuracy and continues to improve, with many efficient algorithmic techniques detecting GEs with a median absolute error of less than a tenth of a second [13,14,15]. However, the results on pathological subjects are often less precise, especially when gait is severely degraded. Due to the complexity of the sensory and motor commands that control gait, patients with advanced neurological pathologies can have particularly unstructured gaits [16,17,18]. For example, Ji et al. showed a 4 times lower accuracy in detecting the end of the step in hemiplegic subjects compared to healthy subjects [19]. Moreover, most of the algorithms tested on impaired patients have only been tested in a few specific diseases [20,21,22,23]. Therefore, one of the current challenges is to improve GE detection in pathological gaits.

A recent literature review has referenced the most commonly used mathematical principles in the exercise of IMU-based gait analysis [24]. This study summarizes research practices regarding IMU positioning, algorithmic methods, and algorithm validation processes. Over the past few years, the 3 most commonly found types of methods have been the Hidden Markov Model (HMM) [25], the Wavelet Transform (WT) [9], inspired by the study of ECGs, and rule-based detections (RBD) utilizing various mathematical tools [13, 26]. The study recommends the use of RBD with IMUs placed on the ankle or foot, validated using pressure sensors as ground truth. Recently, Deep Learning algorithms have also proven successful in detecting GEs [23].

Among the RBD algorithms, template-based methods have shown promising results in segmenting GEs in healthy subjects [10, 27, 28]. This technique relies on the creation of a reference dictionary of steps, which is used for segmentation of the signal by pattern recognition and extraction [29]. The use of a unique step dictionary allows for GE segmentation in healthy subjects [10, 29]. However, in neurological diseases, this method requires the addition of a solution for imprinting the patient’s step, such as an instrumented mat [28], making the deployment in routine clinical practice impossible. Another area for improvement concerns the use of different pattern extraction methods, such as Dynamic Time Warping (DTW) [10, 30]. DTW is a technique that promises precise detection of signal variations and similarity, and has already been used in gait analysis [27, 31].

The objective of this study is to develop a reliable algorithm for detecting degraded gaits using only IMUs, making it suitable for routine clinical use. To achieve this, we propose a new method for detecting Gait Events (GEs) in degraded gaits using IMUs. This method combines mathematical tools such as autocorrelation [32], the matrix profile algorithm [33] and the multiparametric Dynamic Time Warping (mDTW) algorithm [10].

This new automatic GE detection algorithm is based on the raw acceleration and gyration data collected by the IMUs and requires neither manual annotation nor external tools. To evaluate its efficiency, we tested it in comparison with an instrumented mat, considered the gold standard in gait segmentation [34], as well as with template-based state-of-the-art techniques [10, 29].