Project Topics
Get Complete Project Material File(s) Now! »
Human gait analysis and gait cycle phases description
This section presents an analysis of human gait and describes the gait cycle phases.
Human gait analysis
Gait analysis is the systematic study of bipedal locomotion in human. This analysis encompasses quanti cation as well as interpretation of measurable parameters charac-terizing the human bipedal locomotion from a given gait pattern [51].
Gait analysis is an open and competitive topic that is attracting increasing interest in various multidisciplinary elds. In sport, such an analysis is a very useful tool in coaching for improving the performances of athletes and preventing injuries. Gait analysis can also aid clinicians in gait recovery process monitoring in subjects following operations or in course of processes of rehabilitation. In the rehabilitation, events distinguishing gait phases may be used as functional electrical stimulation control variables. Moreover, gait analysis can be exploited in the design and control of wearable assistive devices such as exoskeletons, prostheses and orthoses, used for walking assistance for elderly and subjects with paretic lower limbs. Gait analysis can also be used in healthcare monitoring for the detection of abnormal gait and estimation of fall risk. An abnormal gait can be seen as a symptom indicating neurodegenerative diseases progression. For example, the presence of gait abnormalities in the elderly, is frequently an important predictor of the risk of dementia development. Finally, the generation of walking models for human gait imitation based humanoid robots, is another potential application eld of the human gait analysis [53].
The early work on gait analysis was conducted at the end of the 19th century and the rst applications in biomedical engineering arisen with the availability of video camera sys-tems [54{58]. Numerous gait laboratories have successfully developed and implemented a standard gait analysis method based on a force platform with the capability of mea-suring Ground Reaction Forces (GRFs) and on a multi-camera motion capture system [59, 60]. Yet, this standard method of gait analysis necessitates expensive equipment, specialized locomotion laboratories, and long-time of installation and post-processing. In addition, limitations have been observed in terms of the moving area and gait cycles for the observed subject. To overcome these limitations, other methods of gait analysis based on wearable sensors were proposed.
Human gait cycle description
Walking is a form of movement that represents the most important human physical activity because it is at the basis of several daily living activities. It represents a cyclical activity in which each stride (gait cycle) follows the other one continuously. The gait cycle describes how subjects walk; it represents a pattern of movement that is speci c to each individual [53]. A gait cycle starts when the toe of the Right Foot (RFo) and the heel of the Left Foot (LFo) touch the ground simultaneously and ends with the same con guration.
The gait cycle is mainly divided into two phases: the stance phase and the swing phase. The stance phase represents the period during which the foot is in contact with the ground while the swing phase corresponds to the period during which the foot is in the air for the advancement of the leg. The initial contact is the beginning of the stance phase and the swing phase starts when the foot is lifted from the ground (Fig. 2.4)[61].
Concerning the time percentage of each phase, the stance phase represents 60 % of the gait cycle time while the swing phase represents 40 % [62{64]. The exact duration of each phase changes according to the subject’s walking velocity [65, 66]. In a normal walking pace (80 m/min), the stance and the swing phases represent, respectively, 62 % and 38 % of the gait cycle time. Furthermore, it can be noted that the walking speed and the two phases durations have an inverse relationship ; swing and stance phases durations are shortened when the gait speed increases, and vice-versa.
The stance phase is generally subdivided into three intervals according to the sequence of contact between the two feet and the ground. The rst and third intervals imply a period of bilateral contact of the two feet with the ground (double stance) while the second interval corresponds to one contact with a single foot (single stance) (Fig. 2.4)[61]. In the following, a brief description of these intervals is presented.
The timing for the stance intervals is as follows: 10 % for each double stance sub-phase and 40 % for the single stance. Since the single stance of one foot and the swing phase of the second one occur at the same time, they have the same duration (see Fig. 2.4).
As presented above, the stance phase and the swing phase are the main phases of the gait cycle, but in practice, the cycle can be divided up to eight sub-phases (functional patterns): ve sub-phases in the stance phase and three sub-phases in the swing phase [61](see Fig. 2.4). The eight gait cycle sub-phases are de ned hereafter :
Loading Response (foot at): represents the initial double stance interval. It begins when the right foot touches the oor (heel strike) and continues until the left foot is lifted from the oor to initiate its swing phase. The sub-phase duration varies from 0 to 10 % of the gait cycle.
The number of sub-phases studied varies from one study to another. In [67, 68], the authors take into consideration the eight gait cycle sub-phases. In [69], Nordin and Frankel consider a seven sub-phases gait cycle: initial contact, mid-stance, terminal stance, pre-swing, initial swing, mid-swing and terminal swing. In some early literature ([70]), the initial contact is considered as a part of the loading response sub-phase. In [53], authors take into consideration six sub-phases, by eliminating the initial contact and the initial swing sub-phases. In [71], Williamson and Andrews consider four stance sub-phases (the loading response, the mid-stance, the terminal stance and the pre-swing) and the swing phase. In [72], the gait cycle is divided into stance, heel-o , swing and heel-strike sub-phases. Finally, several researchers only consider the stance and swing phases in their studies, as in [73, 74].
Gait cycle of Parkinsonian subjects
Parkinson’s disease (PD) subjects usually su er from gait alterations that increase with progression of the disease.
Freezing of gait: Freezing of gait is de ned as a brief, episodic absence or marked reduction of forward progression of the feet despite the intention to walk. This phenomenon is, in general, temporary, and the gait may be at a normal pace after few steps. This gait alteration can be triggered by contextual (eg. emotional, cognitive) and environmental factors, such as: walking through a doorway or a narrow passageway, changing directions, approaching one’s destination (such as a chair or couch), crossing a street, and simply when a patient feels like he/she is being rushed [15]. For PD patients, gait freezing induces increased risk and frequency of falls [2].
Shu ing gait: Subjects with PD may walk slowly with their chest bent forward, with short fast shu ing steps. This results in reduced stride length and walking speed. Subjects show also less arm and body movement which gives them a sti appearance [75].
Festinating gait: characterized by a short and fast stride. Short stride and fast step lead to a quite ine cient gait, which makes the walking person tired and frustrated [76].
Hypokinesia: characterized by a small amplitude movements [77]. This alter-ation is associated with di culties throughout the walking process, from prepar-ing to initiation and nally performing walking. Execution of simultaneous and sequential of the walking movements is hampered [15].
These gait alterations increase the risk and the rate of fall. Falls may lead to severe injuries and fractures. The fear of falling is another consequence which results in a restriction of daily activities that in turn lead to a loss of independence and increased mortality. The fear of falling has psychological consequences, such as isolation with less social interactions and depression risk [77]. The e ects of PD on the evolution of the stride-to-stride variability during a gait cycle have been extensively studied in the liter-ature [78]. Yogev et al. [79] studied the cognitive function and the e ects of di erent types of dual tasks on the gait of subjects with PD. The outcomes of this study show that the executive function [80{82] is deteriorated in the subjects with PD. In [37], the authors discuss gait asymmetry (GA) in subjects with PD. The outcomes of this study show that when gait becomes impaired and less automatic, Gait Asymmetry apparently relies on cognitive input and attention. In the same context, Hausdor et al. [38] focused on the gait dynamics to evaluate the e ect of Rhythmic Auditory Stimulation (RAS), which consists of using musical stimuli to enhance the gait performance of neurologi-cal conditions subjects (e.g., subjects with PD). It has been shown that RAS promotes more automatic movement and reduces stride-to-stride variability in subjects with PD. The study conducted in [64], showed that the ability to maintain a steady gait with low stride-to-stride variability decreases in subjects with PD. In [39], the authors showed that swing time variability is independent of gait speed in subjects with PD; therefore, this can be used as a marker of rhythmicity and gait steadiness. The obtained results show also an increase in the variability of stride time and swing time at comfortable walking speeds for the subjects with PD compared to control subjects.
Gait assessment techniques
This section presents the semi-subjective and objective techniques used for gait anal-ysis, as well as the di erent types of sensors used for measuring and estimating gait parameters.
Semi-subjective techniques
The tests and measurements traditionally used for analyzing gait parameters in clinical conditions are carried out by therapists by observing and evaluating the patient’s gait-related parameters while he/she walks along a pre-determined circuit. Several semi-subjective techniques are traditionally used in clinics [83]:
Timed 25-Foot Walk (T25-FW): known as the 25 foot walk test. During this test, therapist measures the time taken by the patient to walk a distance of 7.5 m in a straight line [84].
Multiple Sclerosis Walking Scale (MSWS-12): is a self-assessment scale which measures the impact of multiple sclerosis on walking. It consists of 12 questions concerning the walking di culties due to multiple sclerosis during the past 2 weeks [85].
Tinetti Performance-Oriented Mobility Assessment (POMA): the protocol used in this assessment is as follows, (1) the patient is seated in a chair without armrests;
(2) the patient is asked to stand up, if possible without leaning on the armrests, a balance test in the standing position is then performed; (3) the patient must turn 360 ; (4) the patient must walk at least 3 meters forward, turn around and return back quickly to the chair. He must use his usual technical assistance (cane or walker); (5) the patient must sit on the chair [86]. This test allows to precisely evaluate gait disorders and elderly subjects’ balance in daily life situations [83].
Timed Get up and Go (TUG): in this test, the clinician measures the time taken by the subject to get up from a sitting position, walk a short distance, turn 180 , then return back to the chair and sit down again [87].
Gait Abnormality Rating Scale (GARS): is a video-based analysis used to rate the subjects’ gait according to 16 variables using a 4-point scale (0 = normal, 1 = mildly impaired, 2 = moderately impaired, 3 = severely impaired). These variables can be classi ed into: general categories (5 variables), lower extremity categories (4 variables), and trunk, head, and upper extremity categories (7 variables). The GARS is obtained by summing each of the individual variables. More impaired gait is characterized by a higher score [88].
Extra-Laboratory Gait Assessment Method (ELGAM): is a method to quantify gait in the community or home. In this method several parameters are studied such as initial starting style of walking, walking speed, step length, static balance and ability to turn head during walking [89].
Tests and measurements used in semi-subjective techniques are usually followed by a self-assessment step in which the patient is asked to give a subjective evaluation of his/her gait quality. The aformentioned methods have the disadvantage of relying on subjective measurements that may have an incidence on the quality of diagnosis and treatment [83].
Objective techniques of gait analysis
With advances of sensor technologies, new techniques have been introduced as an alter-native to semi-subjective techniques. These techniques have the advantage to provide a more reliable information related to gait parameters, and therefore, a more objective evaluation of gait quality as well as a more reliable diagnosis. The objective techniques of gait analysis di er from the semi-subjective ones by the use of di erent sensors for measuring and estimating the gait parameters. Sensors used for gait analysis can be classi ed into two main categories: non-wearable or wearable (Fig. 2.5).
Figure 2.5: Sensors used for gait analysis: Wearable and Non-Wearable sensors.
Non-wearable sensors can be classi ed into two sub-categories. The rst sub-category in-cludes oor sensors such as force platforms and pressure-measurement systems; pressure sensors and Ground Reaction Force (GRF) sensors are used to extract gait information by measuring the force exerted by the subjects feet on the oor during walking. The sec-ond sub-category includes mainly vision-based systems such as optical motion capture (OMC) systems (Fig. 2.6). An OMC system consists of multiple optoelectronic cameras and allows an objective and accurate measurement of gait parameters. As alternative non-wearable sensors for gait analysis, infrared sensors, Time-of-Flight (ToF) cameras, laser range nders (LRF) placed at prede ned positions [90] or mounted on robotic rollators are also used [91]. Non-wearable sensors are generally expensive and present several constraints that limit their use to instrumented and indoor environments.
Table of contents :
1 Introduction
2 Diagnosis of Parkinson’s Disease
2.1 Introduction
2.2 Parkinson’s Disease
2.3 Diagnosis of Parkinson’s Disease
2.4 Human gait analysis and gait cycle phases description
2.4.1 Human gait analysis
2.4.2 Human gait cycle description
2.5 Gait cycle of Parkinsonian subjects
2.6 Gait assessment techniques
2.6.1 Semi-subjective techniques
2.6.2 Objective techniques of gait analysis
2.6.2.1 Non-Wearable sensors
2.6.2.2 Wearable sensors
2.7 Positioning of the thesis
3 Data-driven approach to aid Parkinson’s disease diagnosis
3.1 Introduction
3.2 General background
3.2.1 Pre-processing
3.2.1.1 Features computation
3.2.1.2 Features Selection
3.2.1.3 Features Extraction
3.2.2 Classication Techniques
3.2.3 Performance evaluation
3.2.3.1 Generalization performance
3.2.3.2 Classier performance evaluation
3.3 Related works
3.4 Parkinson’s disease classication
3.4.1 Dataset Description
3.4.2 Data preprocessing
3.4.3 Results of feature extraction process
3.5 Results and Discussion
3.5.1 Parameters settings
3.5.1.1 Supervised methods
3.5.1.2 Unsupervised methods
3.5.2 Parkinson’s disease classication results
3.5.2.1 Results of feature selection process
3.5.2.2 Classication results
3.6 Conclusion
4 CDTW-based classication for Parkinson’s Disease diagnosis
4.1 Introduction
4.2 Time series similarity measures
4.3 Dynamic Time Warping (DTW)
4.3.1 Dynamic Time Warping (DTW) formulation
4.3.2 Continuous Dynamic Time Warping (CDTW) formulation
4.4 Gait cycle similarity evaluation using Dynamic Time Warping (DTW)
4.5 Data pre-processing for PD classication
4.5.1 Features extraction
4.5.2 Features selection
4.6 Results and discussion
4.6.1 PD classication using CDTW-based features
4.6.2 PD classication based on feature selection
4.7 Conclusion
5 Multidimensional CDTW-based classication for Parkinson’s Disease diagnosis
5.1 Introduction
5.2 Multidimensional CDTW formulation
5.3 PD subjects classication using multidimensional CDTW-based features
5.3.1 Parameter settings
5.3.2 Results and discussion
5.4 Conclusion
6 Conclusion and Perspectives
6.1 Conclusion
6.2 Perspectives
