A new fractal-based kinetic index to characterize gait deficits with application in stroke survivor functional mobility assessment
Introduction
Surface electromyography (sEMG) signal records the electrical activities of skeletal muscles. It infers muscle function that generates body movement [1]. sEMG has been widely used by researchers and clinicians to perform gait analysis [[2], [3], [4]]. It could be characterized by techniques involving time and frequency domain analysis. For example, root mean square value (RMS), zero-crossing (ZC) rate, median frequency (MDF), mean frequency (MNF), determine muscle fatigue and muscle energy expenditure. These methods reveal specific properties in the linear system context. However, sEMG signal is nonlinear in nature [5]. Rodrick and Karwowski observed positive Lyapunov exponents existed in sEMG of the biceps muscle in some work postures. These suggested chaotic-liked behaviors [6]. Ouyang et al. [7] revealed the characteristics of sEMG during different hand movements using recurrence plots. Besides that, fractal analysis is another common approach to identify nonlinear characteristics of sEMG signals.
Fractal dimension (FD) measures self-affine and dominant complexity of a signal [8]. Detrended fluctuation analysis (DFA) [9], correlation dimension, Katz method [10], box counting method, Higuchi fractal dimension (HFD) [11] and bi-phase power spectrum [12] are common methods to estimate FD of a time series. These techniques have been widely applied to correlate the sEMG FD and its interference patterns [12,13]. Besides that, fractal analysis is also commonly applied in sEMG signal classification [[14], [15], [16]]. In a recent study, FD of rectus femoris muscle sEMG was strongly correlated to the height of vertical jump [17]. Besides that, FD was used to estimate the contraction force from different muscles [18]. In gait analysis, Beretta-Piccoli et al. [19] extracted FD from the quadriceps femoris muscle sEMG to scrutinize fatigue. Boccia et al. [20] correlated the rate of change of FD from vastus lateralis and medialis muscles sEMG to fatigue contraction.
While there are extensive reports on sEMG fractal analysis, its applications in patients with neurological disorders are relatively limited. This could be attributed to the requirement of analyzing long time-series [21,22]. Nevertheless, this technique shows great potential as a quantitative gait assessment tool for neurological pathologies [22]. FD of a time series could serve as a normalized indicator since its value varies between 1 (line) and 2 (plane). Comparatively, raw sEMG signal would have large variation in amplitude across the stroke survivors. Therefore, it requires normalization with maximum voluntary contraction (MVC) which cannot be reliably established in individuals with neurological disorders.
Timed Up and Go test (TUG) is a simple test to evaluate the basic mobility, balance and locomotor skills of elderly and patients with neurological disorders such as stroke. The procedure of TUG test can be studied in [23]. TUG test is often used to monitor the recovery progress of disabled patients after certain intervention [[24], [25], [26]]. Besides that, TUG test can also be applied to classify stroke from healthy [27], type of amputation of lower limbs [28], type of walking aid [29] etc. This forms the motivation of current research work. We aim to investigate the feasibility of applying fractal analysis on sEMG signals from stroke survivors to characterize their gait deficits and to classify the gait deficits based on their TUG score.
Gait classification among stroke is to identify homogeneous subgroups of stroke survivors, which could enable physiotherapist to deliver treatment that is more effective during rehabilitation. This is particularly important to those researchers who do not have full access to collect necessary gait data; such a method would also facilitate communication between clinicians [30]. Besides that, proper classification can help to organize and manage large amounts of complex gait data. These gait data were generated by instrumented gait analysis such as kinematic and EMG data [31]. Many authors have attempted to identify homogeneous subgroup of gait pattern among stroke survivors using methods such as cluster analysis [32,33] and artificial neural network [30,34]. However, most of these methods require multiple inputs, which are very subjective and generally based on observation by visual inspection from researchers or clinicians [33]. Meanwhile, single category of parameter often yielded functionally heterogeneous results [33]. Therefore, it is worth applying classification methods to classify the fractal features mentioned earlier and compared it to conventional classification results.
The objectives of this study are: (i) to formulate a new kinetic index using the fractal features from sEMG of different muscles and to correlate to TUG score; (ii) to classify the gait pattern of stroke subjects into homogenous subgroup using different approaches; and (iii) to compare the classification results. To achieve these objectives, 30 strokes survivors with different walking functionalities were recruited in this experimental study. The subject’s sEMG from different muscles were analyzed.
Section snippets
Higuchi fractal dimension (HFD)
In this study, Higuchi algorithm [11] is used and it is briefly described as follows:
Consider a sEMG time series , where N is the total number of samples in the time series. A total of k new time series , are constructed and defined as Eq (1):where m and k are integer numbers which represent the initial time and the interval time respectively, indicates the integer part of the expression. The length of the curve is computed as Eq
Experiment procedure
30 stroke survivors with different levels of gait impairments (i.e. different TUG scores) were recruited for this experiment. This experiment was conducted at the local medical center. Ethics approval was obtained at this medical center. Stroke survivors were asked to perform the TUG test and allowed to use the assistive devices they commonly used. All participants were then instructed to wear a gait sensor system. The system included two force sensitive resistors (FSRs) (SEN-09375) on toe and
Correlation between K.I. and TUG score
Fig. 6 shows the correlation between K.I. and TUG scores. The correlation coefficient r was 0.9222. The result suggested that K.I. was strongly correlated to TUG scores. Table 1 shows the means, standard deviations (SD) and 95% confidence interval of the K.I. for the 3 different stroke groups. In particular, stroke survivors with TUG scores ranged from 10 to 19 s had the lowest K.I. value ( = 33.1, SD = 2.45). Subjects with TUG score of 20–29 s had increased K.I. value ( = 45.7,
Discussion on methods
To test the quality of sEMG signal, the SNR was determined using Eq (10) for stroke survivors from all three categories. The SNR of raw sEMG from the first group (TUG 10–19 s) ranges between 24–30 dB, the second group (20–29 s) ranges between 22–28 dB and the third group (30 s and above) ranges between 21–26 dB. SNR results showed that the third group has lower quality compared to two other categories due to lower level of muscle contraction. However, the difference is not significant.
Fractal
Conclusions
In this study, a new Kinetic Index K.I. is proposed to characterize stroke survivor's gait functionality. 30 stroke survivors with different gait functionalities were recruited. Their sEMG from Tibialis Anterior and Gastrocnemius Lateral muscles were acquired in a 5-meter walk experiment. Results showed that K.I. has strong correlation to the TUG scores (r = 0. 9222, p < 0.05). The proposed method allows survivors gait deficits to be examined at neuromuscular level.
Hierarchical Cluster Analysis
Acknowledgement
The authors wish to acknowledge the support and help from physiotherapists from University Malaya Medical Center. The approval number of medical ethics committee is 201411-835.
References (44)
- et al.
Abnormal EMG muscle activity during gait in patients without neurological disorders
Gait Posture
(2008) - et al.
Detection of the onset of gait initiation using kinematic sensors and EMG in transfemoral amputees
Gait Posture
(2014) Fractals and the analysis of waveforms
Comput. Biol. Med.
(1988)Approach to an irregular time series on the basis of the fractal theory
Physica D
(1988)- et al.
Fractal analysis of surface electromyography signals: a novel power spectrum-based method
J. Electromyogr. Kinesiol.
(2009) Fractal analysis of surface EMG signals from the biceps
Int. J. Med. Inform.
(1997)- et al.
Fractal analysis features for weak and single-channel upper-limb EMG signals
Expert Syst. Appl.
(2012) - et al.
Assessing fractal dimension methods as feature extractors for EMG signal classification
Eng. Appl. Artif. Intel.
(2014) - et al.
Linear correlation between fractal dimension of surface EMG signal from Rectus Femoris and height of vertical jump
Chaos Soliton Fract.
(2014) - et al.
Allometric control, inverse power laws and human gait
Chaos Soliton Fract.
(1999)
Effects of twice-weekly intense aerobic exercise in early subacute stroke: a randomized controlled trial
Arch. Phys. Med. Rehabil.
The timed up & go test: its reliability and association with lower-limb impairments and locomotor capacities in people with chronic stroke
Arch. Phys. Med. Rehabil.
The timed “up and go” test: reliability and validity in persons with unilateral lower limb amputation
Arch. Phys. Med. Rehabil.
Timed up & go test score in patients with hip fracture is related to the type of walking aid
Arch. Phys. Med. Rehabil.
Intelligent data analysis of instrumented gait data in stroke petients-a systematic review
Comput. Biol. Med.
Gait Pattern categorization of stroke participants with equinus deformity of the foot
Gait Posture
Use of cluster analysis for gait pattern classification of patients in the early and late recovery phases following stroke
Gait Posture
Use of the Higuchi’s fractal dimension for the analysis of MEG recordings from Alzheimer’s disease patients
Med. Eng. Phys.
Fractal dimension analysis for crack identification in beam structures
Mech. Syst. Sig. Process.
Hemiparetic gait following stroke. Part I: characteristics
Gait Posture
Bioelectrical Signal Processing in Cardiac and Neurological Applications
Qualitative evaluations of gait rehabilitation via EMG muscle activation pattern: repetition, symmetry, and smoothness
IEEE Int. Conf. ROBIO
Cited by (5)
Systematic review of automatic post-stroke gait classification systems
2024, Gait and PostureReconstruction of incomplete surface electromyography based on an adversarial autoencoder network
2023, Biomedical Signal Processing and ControlHiguchi fractal dimension: An efficient approach to detection of brain entrainment to theta binaural beats
2021, Biomedical Signal Processing and ControlCitation Excerpt :Now that the concept of complexity and HFD have been discussed, taking a glance at a regular signal (e.g., a sinusoidal signal) and a very complex and irregular signal (z variable in Lorenz system [69]) would provide further insights into these concepts (Fig. 1). HFD has been successfully utilized for various purposes: characterization of gait deficits [70], assessment of depth of anesthesia [71], estimation of the bispectral index score [72], discriminating between depressed patients and normal subjects [73], depression detection [74], discriminating between schizophrenic patients and controls [75,76], measuring the complexity of lung sounds [77], investigation of heart rate variability (HRV) [58,78], classification of fetal heart rate [79], magnetoencephalogram (MEG) complexity measurement in Alzheimer’s disease patients [80], assessment of EEG complexity changes caused by Hypoglycemia in Type 1 diabetes [81] (for more details on the applications of HFD, see [82]). EEG complexity could be under the influence of task and recording conditions.
Fractal analysis tools for early assessment of liver inflammation induced by chronic consumption of linseed, palm and sunflower oils
2020, Biomedical Signal Processing and ControlCitation Excerpt :Recently, different optimization techniques and machine learning strategies have been combined towards the automated lesion segmentation and classification [8,25,26], which could further be integrated with the compatible decision support systems (DSS). Amongst the proposed methods, fractal and multifractal analyses have been very successful in biomedical signal classification [27–33], as well as in tissue classification and analysis [25,34–42], which can be attributed to the fractal nature of tissues and organs [43–45] and the underlying multiplicative processes governing their growth [46,47]. In the study of biomedical signals, fractal features are well established for the assessment of heart rate variability (HRV) [27,29], including the fetal heart rate (FHR) [7], gait rhythm [26,33], muscle contraction variations [28], as well as electroencephalogram (EEG) signals [31,32].