With this your competent? doctor and therapists will have an objective damage diagnosis of your gait problems. Leading them to create EXACT REHAB RECOVERY PROTOCOLS! At least they would if they were competent at all! Are they?
Enhanced gait tracking measures for individuals with stroke using leg-worn inertial sensors
Journal of NeuroEngineering and Rehabilitation volume 21, Article number: 219 (2024)
Abstract
Background
Clinical gait analysis plays a pivotal role in diagnosing and treating walking impairments. Inertial measurement units (IMUs) offer a low-cost, portable, and practical alternative to traditional gait analysis equipment, making these techniques more accessible beyond specialized clinics. Previous work and algorithms developed for specific clinical populations, like in individuals with Parkinson’s disease, often do not translate effectively to other groups, such as stroke survivors, who exhibit significant variability in their gait patterns. The Salarian gait segmentation algorithm (SGSA) has demonstrated the potential to detect gait events and subsequently estimate clinical measures of gait speed, stride time, and other temporal parameters using two leg-worn IMUs in individuals with Parkinson’s disease. However, the distinct gait impairments in stroke survivors, including hemiparesis, spasticity, and muscle weakness, can interfere with SGSA performance. Thus, the objective of this study was to develop and test an enhanced gait segmentation algorithm (EGSA) to capture temporal gait parameters in individuals with stroke.
Methods
Forty-one individuals with stroke were recruited from two acute rehabilitation settings and completed brief walking bouts with two leg-worn IMUs. We compared foot-off (FO), foot contact (FC), and temporal gait parameters computed from the SGSA and EGSA against ground truth measurements from an instrumented mat.
Results
The EGSA demonstrated greater accuracy than the SGSA when detecting gait events within one second, for both FO (96% vs. 90%) and FC (94% vs. 91%). The EGSA also demonstrated lower error than the SGSA when detecting paretic FC, and FO events in slow, asymmetrical, and non-paretic footfalls. Temporal gait parameters from the EGSA had high reliability (ICC > 0.90) for stride time, step time, stance time, and double support time across gait speeds and levels of asymmetry.
Conclusion
This approach has the potential to enhance the accuracy and validity of IMU-based gait analysis in individuals with stroke, thereby enhancing clinicians’ ability to monitor and intervene for gait impairments in a rehabilitation setting and beyond.
Introduction
Inertial measurement units (IMUs) can obtain high-resolution gait parameters in diverse environments with minimal obtrusiveness. With judicious signal processing algorithms, these wearable sensors offer precision comparable to that of cumbersome and expensive motion analysis equipment [1, 2]. These sensors can be leveraged in clinical gait analysis and rehabilitation medicine to detect abnormal gait patterns and biomarkers of impairment [3, 4]. Integrating these devices into therapy planning would enable clinicians to monitor changes in gait impairment, supplementing traditional functional outcomes of gait speed, endurance, and balance [5,6,7].
Effective, reliable algorithms are needed to extract meaningful gait measures from raw IMU signals. Although numerous studies have demonstrated the validity and reliability of IMU-based gait parameters within various populations and clinical settings (i.e., by detecting and segmenting key parts of the gait cycle) [8,9,10], these measures have remained largely untested across individuals with different gait impairments. Examining an algorithm’s validity in different subgroups of patients is critical to define and understand its potential use cases. If an algorithm performs insufficiently, it may not be practical for clinical use. Further development of the population-specific or even person-specific algorithms may be warranted [11]. This approach can maximize the validity of gait parameters for effective implementation, and it can guide clinicians and researchers to interpret measures generated by different algorithms [12, 13].
Stroke is one such condition that may challenge the performance and usability of IMU systems for clinical gait analysis. Distinct motor impairments after stroke (including hemiparesis, spasticity, muscle weakness, limited balance and coordination) result in asymmetric, slow, and irregular gait patterns, as well as limited knee flexion and ankle dorsiflexion [14, 15]. We previously examined IMU-derived gait parameters from a commercial IMU system in a chronic stroke population, finding greater error during slower walking and more asymmetric gait [16]. Such errors would limit the usability of an IMU system in Inpatient Rehabilitation Facilities (IRFs), where patients experience more extensive and fluctuating gait impairments compared to chronic populations [17]. Notably, there have been a limited number of algorithms developed or validated specifically in stroke populations [4]. In one example, Yang et al. (2013) applied IMUs on the lower legs in 13 individuals with post-stroke hemiparetic gait [18]. They utilized acceleration and angular velocity signals to compute walking speed and to segment different parts of the gait cycle, respectively. While their approach generally performed well when estimating walking speed, there was no comparative ground truth measurement to validate the gait segmentation performance, and the authors acknowledged that their algorithm failed for subjects with abnormal shank angular velocities. Thus, more robust methods are needed to address the range of gait patterns seen in broader stroke populations. Specialized gait segmentation algorithms may be necessary to adapt to and accurately interpret these varied stroke-related gait abnormalities [19].
One such gait segmentation algorithm, presented by Salarian and colleagues, was designed to detect mid-swing, foot-off (FO), and foot contact (FC) gait events based on characteristic angular velocity patterns during walking, as measured by two IMUs on the lower legs [20]. Originally designed using IMU data from people with Parkinson’s disease, which typically manifests in gait as a global impairment, marked by bradykinesia and shuffling [21], this algorithm has been applied to various healthy and clinical populations for gait monitoring [22], deriving spatial gait parameters [23], and computing features for predictive models [24]. However, the sole reliance of Salarian’s gait segmentation algorithm (SGSA) on angular velocity could lead to misdetections or failures when applied to populations with different gait abnormalities, such as stroke survivors.
Unlike Parkinson’s disease, stroke-related gait impairments often include hemiparesis, spasticity, and muscle weakness, leading to asymmetric, slow, and irregular gait patterns [25]. Specifically, hemiparesis leads to uneven weight distribution and altered swing phases between the paretic and non-paretic legs [26, 27], causing irregular patterns that gyroscopes may not consistently detect [28]. Spasticity and muscle weakness contribute to unpredictable and hesitant movements, resulting in irregular angular velocity signals that can obscure the precise identification of gait events [29]. Additionally, slower walking speeds amplify the variability in limb movements, making it more challenging for gyroscope-based algorithms to maintain accuracy in event detection. These complexities can hinder an algorithm’s ability to reliably segment the gait cycle by making it difficult to differentiate between genuine gait events and other motions introduced by post-stroke impairments. To address these challenges, we developed and tested an enhanced gait segmentation algorithm (EGSA), an adaptation of the SGSA specifically for individuals with subacute stroke.
The objectives of this study were twofold: (i) to benchmark the gait events and temporal gait parameters derived from the SGSA and the EGSA against a commonly used reference system, the GAITRite instrumented mat, and (ii) to compare the performance of the EGSA against the SGSA across different levels of speed and symmetry in subacute stroke. We hypothesized that the EGSA would be more sensitive than the SGSA when identifying gait events for individuals with stroke exhibiting slow and asymmetric gait.
No comments:
Post a Comment