0
Research Papers

Application of Digital Human Models to Physiotherapy Training OPEN ACCESS

[+] Author and Article Information
Takao Kakizaki

Department of Mechanical Engineering,
Nihon University,
Nakagawara 1, Tokusada, Tamura,
Koriyama 963-8642, Fukushima, Japan
e-mail: kakizaki.takao@nihon-u.ac.jp

Mai Endo

Department of Mechanical Engineering,
Nihon University,
Nakagawara 1, Tokusada, Tamura,
Koriyama 963-8642, Fukushima, Japan
e-mail: bluefascination@gmail.com

Jiro Urii

CAS Research,
44-4-105 Shimo,
Fussa 197-0023, Tokyo, Japan
e-mail: Jiro.URII@cas.fussa.tokyo.jp

Mitsuru Endo

Department of Mechanical Engineering,
Nihon University,
Nakagawara 1, Tokusada, Tamura,
Koriyama 963-8642, Fukushima, Japan
e-mail: m_endo@mech.ce.nihon-u.ac.jp

Contributed by the Design Engineering Division of ASME for publication in the JOURNAL OF COMPUTING AND INFORMATION SCIENCE IN ENGINEERING. Manuscript received December 23, 2016; final manuscript received April 27, 2017; published online July 18, 2017. Editor: Bahram Ravani.

J. Comput. Inf. Sci. Eng 17(3), 031014 (Jul 18, 2017) (16 pages) Paper No: JCISE-16-2157; doi: 10.1115/1.4036991 History: Received December 23, 2016; Revised April 27, 2017

The importance of physiotherapy is becoming more significant with the increasing number of countries with aging populations. Thus, the education of physiotherapists is a crucial concern in many countries. Information and communications technologies, such as motion capture systems, have been introduced to sophisticate the training methods used in physiotherapy. However, the methods employed in most training schools for physiotherapists and occupational therapists remain dependent on more conventional materials. These materials include conventional textbooks with samples of traditional gait motion photographs and video archives of patients' walking motion. Actual on-site clinical training is also utilized in current physiotherapy education programs. The present paper addresses an application of a previously developed digital human model called the kinematic digital human (KDH) to physiotherapy education with a focus on improving students' understanding of the gait motion of disabled patients. KDH models for use in physiotherapy were constructed based on Rancho Los Amigos National Rehabilitation Center (RLANRC) terminology, which is considered the preferred standard among clinicians. The developed KDH models were employed to allow the three-dimensional visualization of the gait motion of a hemiplegic patient.

Background.

According to a white paper on the aging society of Japan [1], the total population is approximately 127,080,000, and the number of people aged 65 yr or older is at a record high of 33 million. This paper also states that the elderly comprise approximately 26.0% of the total population of Japan (one in four people). Based on the current population trend, approximately 39.9% of the total population (two in five people) will be 65 or older by 2060 (Fig. 1). This indicates that the number of people with physical difficulties will increase in the future. Therefore, the importance of physiotherapy will gain significance as Japan's population ages.

The purpose of physiotherapy is to maintain and improve patients' ability to perform physical exercise by employing a variety of physical treatments using heat, electricity, water, and Xenon laser light [2]. Despite the increasing importance of physiotherapy, the number of physical therapists is much smaller than the number of elderly people requiring treatment. The number of licensed physical therapists who passed the national examination of Japan was 110,675 as of 2013. This means that there is one therapist for every 298 elderly people. Thus, physiotherapy training and education should be accelerated significantly to solve this critical problem [3].

Conventional instruction given in physiotherapy training courses typically uses the following tools and tactics. (1) Lectures are given on a variety of gait motions using figures and photographs from textbooks. (2) Students attempt to imitate the gait motions they learned from their textbooks to better understand them. (3) Students attempt to simulate gait motions corresponding to various disabilities using tools that partly restrict movement, such as the bending of the knee. In this way, they can experience the effects of some of the physical disabilities that are common among elderly people on gait motions. (4) If videos of disabled patients are available, students carefully analyze the recorded gait motions in slow motion. They then investigate the analyzed motion by creating their own sketches. Using these typical learning tools, even experienced teachers often have difficulties understanding the gait motions of disabled patients.

Hemiplegia and Duchenne muscular dystrophy are examples of disorders that induce typical gait dysfunctions. Hemiplegia is the paralysis of one side of the body. It is usually caused by a brain lesion, such as a tumor, or stroke syndrome. In contrast, the gait caused by Duchenne muscular dystrophy involves a trunk lean toward the affected stance limb. Moreover, the pelvis is either level or elevated on the swinging limb side during the single stance phase. Both gait patterns are accompanied by complex three-dimensional (3D) body motions. Hence, it is difficult to understand the details of the motion by using only photographic images or videos taken from one direction. Furthermore, quantitative instruction in rehabilitation using such tools is also difficult.

Therefore, leaders in physiotherapy hope to employ digital human models in physiotherapy education. Such models may include (1) simple motion modeling, (2) full 3D modeling with forward and inverse kinematics, and (3) visualizations of walking speed, steps, the gait motion rate, and the trajectory of the body, which are necessary for precise body motion analysis.

Among all physiotherapy patients, a large proportion are hemiplegic. Thus, it is considered that rehabilitation through activities of daily living (ADL) is important for such patients to function in their daily lives [47]. Hence, the introduction of an information and communications technology apparatus including a motion capture device to the field of physiotherapy education has been prioritized in recent years [8].

Related Studies.

The motion capture technology that is most widely used in industrial applications is based on optical methods. The motions of a human body are captured using markers attached to the main parts of the body so that a computer can estimate the motion in two-dimensional (2D) or 3D coordinate space.

Motion capture without any markers is also possible, but its accuracy has certain limitations. For example, a new method of obtaining stroboscopic images from a video of a moving subject captured by camera has been presented [9]. Optical motion capture methods do not require significant body motion constraints in comparison with conventional mechanical methods [1013]. However, such methods require not only a time-consuming video setup but also a complicated camera calibration procedure, especially for precise 3D motion capture.

Recently, a number of studies have also reported motion capture applications using Kinect, a line of motion sensing input devices by Microsoft [14,15]. Kinect devices feature a red–green–blue camera, a depth sensor, and a multi-array microphone running proprietary software. However, it should be noted that all types of motion capture systems require a predefined precise digital human model.

Thus, a method of applying digital human models to physiotherapy training and education is addressed in this study. A simple marker-less motion capture system using only commercial devices can be combined with the precise digital human models.

Objective.

The objective of this paper is to present a method of applying precise digital human models to the field of physiotherapy to allow the 3D visualization and analysis of the gait motions of patients with conditions that affect their gait, such as hemiplegia or Duchenne muscular dystrophy. To achieve this objective, this study employed a digital human model previously developed by the authors [16].

By the interview of physiotherapy experts before starting the study, the following was clarified. At physical therapy schools, video lecture is often utilized, and guidance based on the therapist experience is carried out; however, the method is mainly based on language, and a more direct and quantitative method is required. By modeling the movement of the patient once using the proposed digital human model, it is possible to identify the site to be improved and share this information with the patient and therapist. With this model, the operation goal for the target site can be quantitatively set. For example, it can be concretely determined that the rotation of the ankle joint of the right foot is half of the target value. By quantifying the rehabilitation goal, it is possible to visually confirm the effect of rehabilitation for each site using a personalized motion model for each patient; this will give the patient a sense of accomplishment and increase their motivation to move on to the next rehabilitation stage.

The proposed method can be easily implemented in physiotherapy training facilities using a camcorder and a laptop personal computer (PC). The digital human model introduced here is a full 3D kinematic model with forward and inverse kinematics in Cartesian space. Thus, the gait patterns or postures of the digital human model can be intensively investigated from various angles in Cartesian space once the model is created from the patient's gait information. Furthermore, the angle and modeled speed of each joint can be inspected through joint space analysis. This is very effective because it enables the therapist to discuss the rehabilitation procedure with the patient. To validate the developed method, the digital human model was enhanced by considering conventional gait phases that are widely used in physiotherapy. It was demonstrated that even the insufficient 2D data of a disabled patient's gait motion can be effectively utilized as a 3D physiotherapy human model using the developed method.

In this paper, the digital human model is first explained in detail. Then, the model verification performed using a typical gait motion phase method employed in physiotherapy is described. Finally, examples of the application of the model to patient rehabilitation are presented.

This section describes the details of the kinematic digital human (KDH) model.

Description of Kinematic Digital Human Model.

Figure 2 shows the joint arrangement and the length of each part in the KDH model. In this model, each lower limb comprises a toe joint, an ankle joint, a knee joint, and a hip joint, which have one, two, two, and three degrees-of-freedom, respectively, whereas each upper limb comprises a wrist joint, an elbow joint, and a shoulder joint which have two, two, and three degrees-of-freedom, respectively. The trunk and head of the model comprise a trunk joint, a back/chest joint, a clavicle joint, an upper neck joint, and a lower neck joint which have three, three, one, three, and three degrees-of-freedom, respectively.

Figure 3 shows the arrangement of the joint coordinate frames of the head, trunk, left arm, and left leg of the KDH model; the coordinate frames of the right arm and leg are symmetric to those of the left arm and leg, respectively, about the vertical axis. All joint coordinate frames in the KDH model are right-handed.

In the leg part, for example, k1 expresses the first joint coordinate frame, and the KDH model contains eight coordinate frames (k1 − k8). The joint angle θ, which rotates about each joint axis, is used as a joint coordinate variable.

The origin of the global coordinate frame ΣHumanBase of the KDH model is located at the hip, and the origins of the local coordinate frames ΣHeadBase, ΣRightArmBase, and ΣRightLegBase are located at the neck, the right shoulder, and the right inner thigh, respectively. The origin of the local coordinate frame ΣRightHandPoint is located at the right hand, as in the case of an end effector of a robot. Similarly, the origin of the local coordinate frame ΣRightFootPoint is located at the right foot. The transformation from the ith joint to the jth joint is described by the homogeneous transformation matrix iTj (θj) ∈ R4 × 4. After this transformation, similar homogeneous transformations are performed from the trunk to the right shoulder, neck, and head; from the right shoulder to the right hand; and from the trunk to the right foot. Consequently, the KDH model is a precise kinematic model with 44 joint vectors (see Appendix A). Further details concerning the model are given in a study by Kakizaki et al. [1618].

The KDH model with full 3D kinematics can be used as a reference model during the 3D motion capture process. KDH models can also be applied to forward and inverse kinematics problems.

This section discusses the use of the Rancho Los Amigos National Rehabilitation Center (RLANRC) method, which is one of the gait motion classification methods used in the field of clinical kinematics. The modeling of the gait motion using the KDH model and the RLANRC method is then presented.

Rancho Los Amigos National Rehabilitation Center Method for Gait Analysis.

The RLANRC method is a gait motion taxonomy method developed by the Rancho Los Amigos National Rehabilitation Center. The RLANRC method has been widely used in the field of clinical kinematics for physiotherapy.

Figure 4 shows the gait motion classification used in the RLANRC method [19]. According to this classification, the gait period is defined as the time from the initial grounding of the observation limb to the subsequent grounding of the same observation limb [19]. In general, other characteristic postures repeated during a gait can be considered as elements used to conceptually define the gait motion. However, the initial grounding is considered the most observable feature in the RLANRC method.

The phases of the gait period are classified into two major categories: Stance and Swing. In the RLANRC method, the Stance and Swing categories are divided into five and three phases, respectively. The Stance category includes five phases: (1) initial contact (IC), (2) loading response (LR), (3) midstance (MSt), (4) terminal stance (TSt), and (5) preswing (PSw). The Swing category includes three phases: (1) initial swing (ISw), (2) midswing (MSw), and (3) terminal swing (TSw).

Each phase has its own function. IC and LR represent the transfer of weight onto the observation limb. MSt and TSt correspond to single leg support. Finally, PSw, ISw, MSw, and TSw indicate the motion of the swinging leg as it achieves foot clearance and advances the trailing limb for the next gait period. The total gait period is distributed among the phases as follows, with the period progressing from 0% to 100% completion: IC 0%, LR 0–12%, MSt 12–31%, TSt 31–50%, PSw 50–62%, ISw 62–75%, MSw 75–87%, and TSw 87–100%.

It has been reported that the RLANRC method is applicable to both normal and pathological gait motion [15]. However, the exact limitations of the method have not yet been determined. In the following discussion, the eight phases explained previously are referred to as the eight representative phases of the RLANRC method.

Kinematic Digital Human Modeling of Normal Gait.

Figure 5 shows a schematic flow of the construction of a gait motion KDH model. The developed system is simple and inexpensive compared with conventional motion capture systems. In this modeling process, a video of the normal gait of a subject is taken prior to KDH modeling. The subject walks in a straight line for 4.0 m at normal gait speed. A video is taken by a camcorder at a position of 5.0 m from the path of the subject. Gait images are sampled from the captured video with sampling time of 100 ms. The images corresponding to RLANRC postures are then extracted onto a PC. Finally, KDH models are generated through joint parameter tuning.

Figure 6 shows a series of images of a subject with normal gait motion corresponding to the initial postures of the eight representative phases of the RLANRC method. This series of images was then transformed into a series of KDH models corresponding to the initial postures of the eight representative phases by utilizing KDH forward and inverse kinematics operations (Fig. 7). The gait motion KDH model is a full 3D kinematic model, meaning a movie showing the motion of the model can be produced by applying simple joint interpolation to transition between the phases represented by the eight KDH models. Hereafter, the term “gait motion KDH model” is used to refer to the series of eight KDH models corresponding to the initial postures of the eight representative phases of the RLANRC method and the transitional motion between these eight models.

The gait motion speed of the gait motion KDH model can be calibrated by setting its gait period to be equal to the measured gait period of the subject. In this experiment, the gait period of the subject was 1.1 s.

Figure 8 shows the gait motion KDH model-based coordinate frames used in this study. The global coordinate frame Σg is defined in Cartesian space. The coordinate frame ΣKDH, which determines the initial position and orientation of the KDH model, is defined such that its xy plane overlaps the xy plane of Σg. Furthermore, the moving coordinate frame ΣHB, whose origin is located at the midpoint between the hip joints, is defined as parallel to the Z-axis of ΣKDH. The KDH model is assumed to walk straight on the X-axis of ΣKDH. It should be noted that no yawing of the model about the Z-axis of ΣKDH is allowed in this case. In the initial conditions, all the joint angles of the gait motion KDH model are set to zero.

Gait Motion Parameters.

Table 1 shows the gait parameters used in the gait motion KDH model. H, F, and V are the step length, gait frequency, and gait motion speed of a normal gait, respectively. In general, H, F, and V can be measured from videos of normal gait motion. In most cases, the corresponding gait parameters of pathological patients may be given by these same three parameters. In such cases, the coefficients α and β can be determined from the gait motion observation.

However, the step length, gait frequency, and gait motion speed of a patient's gait are not available when only photographs can be obtained for KDH modeling. In such cases, appropriate values of α and β are assumed because the gait motion of a typical patient is very slow compared to normal gait motion. In the following, α = 0.42 and β = 0.73 are employed as the parameters of a pathological patient.

In this section, the gait motion KDH model of a hemiplegic patient is discussed. Some photographs from a textbook [20] were used to construct the hemiplegic gait motion KDH model. It also should be noted that each image provides limited 2D information with no explicit kinematic expression, although features of the disorder are clinically described.

Initial Kinematic Digital Human Models From Two-Dimensional Photographs.

Figure 9 shows gait motion photographs of a right hemiplegic patient [20]. The major gait motion characteristics of a right hemiplegic patient are as follows. (1) The forward step of the lower limb on the left (nonparalyzed) side is relatively limited in the extorsion direction. (2) The trunk, which is generally tilted forward, is tilted backward when the lower limb on the right (paralyzed) side is swung [20]. In the textbook considered here, only four photographs are used to explain the gait characteristics, as shown in Fig. 9. Similar examples can be found in other conventional clinical kinematics and physiotherapy textbooks. These photographs and the accompanying explanation may give students important knowledge on this pathological gait, but this knowledge may be insufficient for the students to quantitatively analyze the gait motion of a patient. Figure 9 shows the gait, proceeding from 1 to 4. However, the figure does not give certain information that is necessary to reproduce the gait motion, such as the step length and gait frequency. Therefore, KDH models of the patient were constructed using only this limited information.

Figure 10 shows the four initial KDH models of the patient. The models were constructed based on the photographs in Fig. 9. First, a preliminary KDH model was superimposed on each photograph, and forward and inverse kinematics operations were applied to the preliminary KDH model to adjust its posture to align with the posture shown in each photograph. The following assumptions were applied to the construction of the models. (1) The patient gait motion period is 1.5 s. (2) The patient gait motion progresses from 1 to 4 and repeats cyclically. (3) The link parameters of the KDH model of the hemiplegic patient are equal to those for a subject with a normal gait.

Sophisticated Model Based on Rancho Los Amigos National Rehabilitation Center Method.

A series of KDH models corresponding to the eight extracted images of normal gait motions were created by utilizing forward and inverse kinematics operations, as described in Sec. 3.2 (Fig. 7). Figure 11 shows the KDH models corresponding to the initial postures of the eight representative phases of the RLANRC method for a subject with a normal gait. Here, joint space linear interpolation (JSLI) was applied to these eight KDH models to generate transitional postures between each pair of consecutive RLANRC phases.

Then, as shown in Fig. 12, four gait postures corresponding to the initial KDH models of the right hemiplegic patient (Fig. 10) were selected from the interpolated postures.

These postures were selected as follows. The difference between the right and left hip joint angles was selected as a dominant metric to investigate the gait motion characteristics. For the ith representative posture (i = 1–4) of the hemiplegic patient, this difference (constant value) is given as Display Formula

(1)ΔθH,i=|θKR1,iθKL1,i|

where the subscripts KR1 and KL1 denote the right and left hip joints (see Appendix A), respectively, and the subscript i denotes the ith KDH model of the hemiplegic patient. Here, the normalized difference can be given as Display Formula

(2)ΔΘH,i=ΔθH,imax[ΔθH,i]

where i = 1–4.

In contrast, the difference between the right and left hip joint angles in the normal gait KDH model is interpolated at each sampling point s throughout all gait phases as Display Formula

(3)ΔθN(s)=|θKR1,N(s)θKL1,N(s)|
Display Formula
(4)ΔΘN(s)=ΔθN(s)max[ΔθN(s)]

where the subscript N denotes the normal gait. Equation (3) is considered to be a function of s. Then, the candidates of the appropriate posture around each RLANRC representative phase can be found considering the local minimum value of Φ(s) as Display Formula

(5)dΦ(s)ds=0
Display Formula
(6)Φ(s)=ΔΘN(s)ΔΘH,i

Here, it should be noted that the most appropriate posture is determined considering the hip joint motion among the hemiplegic and normal gait postures.

In this manner, the normal gait postures most closely resembling the four representative patient postures shown in Fig. 9 were determined. As shown in Fig. 13, the initial KDH patient gait motions could then be placed in appropriate positions within the RLANRC phases. In this case, two of the four postures corresponded to initial postures of RLANRC representative phases. Finally, by applying linear joint space interpolation to the four KDH models placed in appropriate locations in the gait period, the patient postures corresponding to the initial postures of the RLANRC representative phases were determined.

Figure 14 summarizes the modeling results. As explained before, the six missing links of the right hemiplegic gait motion KDH model were determined by joint space linear interpolation. Patient gait analysis in 3D space can be achieved using full kinematic KDH models. The procedure explained here is called RLANRC-based modeling in this paper.

The KDH model is a full 3D kinematics model. This type of model allows physiotherapists to investigate patient gait motion from various perspectives. In particular, gait motion analysis is possible in both Cartesian and joint space by applying forward and inverse kinematics to the KDH model. Neither still photographs nor video footage can directly provide any such rich information. Thus, quantitative analysis using a KDH model is effective for conveying gait motion knowledge in physiotherapy education.

Justification of Three-Dimensional Visualization of Kinematic Digital Human Model Gait Motions.

Figures 15(a) and 15(b) show the gait motion KDH models for subjects with normal and right hemiplegic gaits, respectively, determined in Sec. 4.2. All postures shown in this figure correspond to the initial postures of the RLANRC phases.

Figure 16 shows visualization examples of the gait motion KDH model in 3D space. In Fig. 16(b), it is apparent that the upper part of the body tends to tilt to the left when the right foot (on the paralyzed side) is swung. It also should be noted that the centerline of the body is slanted in this case, despite the assumption that the KDH model walks in a straight line. This motion is a feature of paralysis, as explained in previous studies [21,22]. Using the proposed model, students can intensively investigate the gait motions from various angles in Cartesian space as shown here.

Gait Analysis in KDH Joint Space.

In this section, the behavior of the gait motion KDH models developed in this study is compared to previously reported data related to RLANRC gait motion analysis.

Figure 17 shows the relationship between the RLANRC gait phase and the right hip joint angle θKR1 of the gait motion KDH models. Positive and negative joint angles correspond to joint flexion and extension, respectively. In the normal gait motion KDH model (Fig. 17(b)), the hip joint angle decreased during LR (phase 2), then reached the maximum extension value at the beginning of TSt (phase 4). This tendency agrees fairly well with results obtained in the previous study (Fig. 17(a)) [23]. Hence, the developed KDH model is effective for gait motion investigation in joint space kinematics related to patient motion. Based on these results, patient gait motion can be investigated in similar manner.

Figure 17(c) shows a hemiplegic gait motion KDH model developed in this study. Here, the hip joint angle remained positive, meaning the hip was constantly in a flexion state, although the angle sometimes decreased slightly. This means that the model analysis clearly indicates one of the important characteristics of hemiplegic gait.

Figure 18 shows the relationship between the RLANRC gait phases and the right knee joint angle θKR4. The knee joint angle in the normal gait motion KDH model (Fig. 18(b)) was positive regardless of the RLANRC phase, and it reached maximum flexion during PSw (phase 5). This tendency coincides with the data obtained in the previous study (Fig. 18(a)) and reflects the actual behavior of a normal knee joint. Conversely, the knee joint angle in the hemiplegic gait motion KDH model (Fig. 18(c)) remained at a constant 20 deg throughout the gait period. This implies that this patient's hemiplegia resulted in the immobility of the knee joint. The result demonstrates the ability of the proposed model to identify another important characteristic of hemiplegic gait.

Figure 19 shows the relationship between the RLANRC gait phases and the right ankle joint angle θKR7. Positive and negative ankle joint angles correspond to dorsiflexion and plantarflexion, respectively. In the normal gait data (Fig. 19(a)) [20], maximum dorsiflexion occurred during MSt and TSt (phases 3 and 4), and maximum plantarflexion occurred during PSw (phase 5). Dorsiflexion and plantarflexion were found to occur in an alternating pattern. Conversely, no plantarflexion occurred in the normal gait KDH model (Fig. 19(b)). This may be because the ankle joint of the examinee was relatively stiff. This is quite natural because there exist certain individual differences among people with normal gait motion. In the hemiplegic gait KHD model (Fig. 19(c)), the ankle joint angle remained nearly constant at 10 deg throughout the gait period. This is similar to the behavior of the knee joint, as shown in Fig. 18(c). This is yet another important characteristic of hemiplegic gait that can be identified using the proposed model.

Consequently, gait motion KDH models developed on the basis of the RLANRC method are effective for gait motion visualization and analysis in clinical kinematics and physiotherapy. As stated previously, only basic devices, including a commercial camcorder and a laptop PC, are required to utilize this modeling method in physiotherapy training.

In this section, an example of the clinical application of KDH to the treatment of a hemiplegic patient is presented. Prior to the treatment, the patient gait motion was captured using a camcorder mounted on a tripod. No extra apparatus was necessary for this process. Next, the patient KDH model was created from the captured video images. The KDH model can present a variety of patient gait motions in 3D space. Using the video and the KDH model, both the patient and physical therapist can share information, such as the target sites for treatment, their joint displacements, and other gait motion parameters. The patient profile is given in Table 2. Figures 20(a) and 20(b) show the side and front views, respectively, of continuous images of the gait motion represented by the patient KDH model.

Table 3 shows the major findings of the patient gait analysis performed by the physical therapist before treatment. These findings can be presented to patients to help them understand the target sites for treatment and necessary action corresponding to the specified joints in the KDH model, such as KR or KN (see Appendices A and B). The most notable finding is that the swing of the upper and lower limbs is insufficient. Therefore, treatment with electromyogram-induced electrical stimulation to the right upper and lower limbs was performed. Specifically, the following four sets of automatic assistance exercises were performed for 5 min each while alternately stimulating the antagonistic muscles of each joint of the right upper and lower limbs with a surface electrode: (i) flexion and extension of the right elbow joint, (ii) flexion and dorsiflexion of the right hand palm joint, (iii) flexion and extension of the right knee joint, and (iv) dorsiflexion and base flexion of the right ankle joint.

After treatment, the patient gait motion was again recorded and the findings by the physical therapist were shared with the patient. Based on the gait motion, the patient KDH model after treatment was created as before the treatment. Figures 21(a) and 21(b) show the side and front views, respectively, of continuous images of the patient KDH gait motion after the treatment.

Table 4 shows the major findings of the patient gait analysis performed by the physical therapist after treatment. The patient and physical therapist determined the effects by comparing the gait before and after treatment while confirming the each joint motion of the KDH model. Table 5 also shows the patient's walking motion data before and after treatment. After treatment, the walking speed and stride increased by 16% and approximately 30%, respectively.

Based on the KDH data before and after treatment, the physical therapist performed gait analysis. As a result, the following changes to the patient's gait were confirmed. (i) The front bending of the trunk and head during walking was improved, (ii) posterior pelvic inclination during walking was improved, (iii) the inclination of the head and trunk to the right was improved when standing on the right foot, and (iv) the stride and walking speed were improved. The patient reported that his incentive for rehabilitation was greatly improved by the KDH-based physical therapy and its ability to provide a visual and quantitative understanding of the treatment results.

This paper addressed the application of precise digital human models to the field of physiotherapy to allow the 3D visualization and analysis of the gait motion of hemiplegic patients. The proposed method can be easily utilized in physiotherapy training facilities using only a camcorder and a laptop PC. In this paper, a digital human model with full 3D kinematic features was considered based on the RLANRC gait classification scheme, which is widely used in physiotherapy. Gait motion analysis was conducted for normal and hemiplegic gait using the developed 3D physiotherapy human models. The effectiveness of the presented method was demonstrated by the gait motion visualization and analysis performed in both Cartesian space and joint space. The presented method was also applied to the rehabilitation of a patient in a clinical setting to confirm the effectiveness of the method. Based on certain physical assumptions, the developed KDH model can be sufficiently utilized in physiotherapy education and rehabilitation even when limited gait information is available for a given patient. The results obtained here indicate that the developed model can be introduced as an effective tool for physiotherapy education programs.

The authors would like to thank Mr. Yasuhiro Kudo and Ms. Emi Yokoyama, Special Needs Education School for the Visually Impaired, University of Tsukuba, for their advice on physiotherapy, including clinical kinematics education programs and the application example for the rehabilitation of a hemiplegic patient. The authors would also like to thank Mr. Hiromasa Fukushi and Daichi Yamauchi for their significant efforts in KDH modeling and verification. This work was supported by JSPS KAKENHI Grant No. 24510234.

Appendix A: Kinematic Digital Human (KDH) Model and Its Joint Arrangement

Figure 22 shows the detail of the joint coordinate frame arrangement in the KDH model. The subscripts KN, KB, KRA, and KRL denote neck part, trunk part, right arm part, and right leg part, respectively. The left arm and leg are mirror-symmetrically arranged to the right side coordinate frame. Thus, the notation for the left-hand parts of the KDH model can be expressed by replacing subscript R with subscript L as shown in the figure.

Appendix B: Mapping of Terminology Between Physiotherapy and KDH

In the physical therapy field, a total of 68 specifically defined exercise terminologies are used in whole body such as “abduction” and “extension” (19). Therefore, extracting the motion of the right half of the body including cervical spines can express the physical therapy state value as a numerical vector pR68. On the other hand, right half degree-of-freedom of KDH is expressed as joint displacement vector kR25. Hence, a mapping of patient motion is defined between them and the mapping is expressed as follows using the transformation matrix pTkR25×68Display Formula

(B1)p=pTkk

Here, the matrix is a diagonal and its elements are arbitrary constants including 0. Although the terminology does not strictly indicate numerical values, it can be quantified by KDH model if treatment part is determined.

Cabinet Office of Japanese Government, 2015, “ Ageing Society White Paper,” Cabinet Office of Japanese Government, Tokyo, Japan, accessed Jan. 6, 2016, http://www8.cao.go.jp/kourei/whitepaper/w-2015/gaiyou/pdf/1s1s.pdf (in Japanese)
Japanese Physical Therapy Association, 2015, “ Membership Statistics,” Japanese Physical Therapy Association, Tokyo, Japan, accessed Dec. 2, 2015, http://www.japanpt.or.jp/aboutpt/physicaltherapy/ (in Japanese)
Saito, H. , 2013, “ The History of the Physiotherapy in Japan, Vol. 2,” Bulletin of Institute of Tokyo Medical Care University, University of Tokyo Health Science, Tokyo, Japan (in Japanese).
Machiel, F. , and Reinkensmeyer, D. , 2008, “ Rehabilitation and Health Care Robotics,” Springer Handbook of Robotics, Springer, Berlin, pp. 1223–1251.
Broadbent, E. R. , Stafford, R. , and MacDonald, B. , 2009, “ Acceptance of Healthcare Robots for the Older Population: Review and Future Directions,” Int. J. Soc. Rob., 1(4), pp. 319–330. [CrossRef]
Smarr, C.-A. , Fausset, C. B. , and Rogers, W. A. , 2011, “ Understanding the Potential for Robot Assistance for Older Adults in the Home Environment,” Human Factors and Aging Laboratory Technical Reports, Georgia Institute of Technology, Atlanta, GA, pp. 6–24.
Isezaki, M. , Takano, K. , and MochidukiI, Y. , 1999, “ Relationship Between Quality of Life (QOL, Life Satisfaction) and Activities of Daily Living (ADL) in Elderly People With Diseases,” Vol. 16, Bulletin of Yamanashi Medical University, Yamanashi Medical University, Kofu, Japan, pp. 71–75 (in Japanese).
Hirose, N. , Nakayama, A. , and Maruyama, H. , 2014, “ Educational Effect of Physical Therapy With the Use of Information and Communication Technology—New Attempt to the Fixing of the Ability of Problem Solving Leading to Clinical Reasoning,” Jpn. J. Sports Rehabil., 3, pp. 37–42 (in Japanese).
Shimada, S. , Suzuki, A. , Yonemura, S. , and Kojima, A. , 2012, “ A Presentation Method of Image Information for Sports Motion Analysis by Video Viewing,” J. Inst. Image Electron. Eng. Jpn., 41(1), pp. 65–72.
Ijaz, A. , Gibbs, L. , Abboud, R. , Wang, W. , Ming, D. , and Wan, B. , 2012, “ Analysis of Knee Joint Kinematics During Walking in Patients With Cerebral Palsy Through Human Motion Capture and Gait Model-Based Measurement,” IEEE International Conference on Virtual Environments Human–Computer Interfaces and Measurement Systems (VECIMS), Tianjin, China, July 2–4.
Tao, W. , Liu, T. , Zheng, R. , and Feng, H. , 2012, “ Gait Analysis Using Wearable Sensors,” Sensors, 12(2), pp. 2255–2283. [CrossRef] [PubMed]
Lee, L. , and Grimson, W. , 2002, “ Gait Analysis for Recognition and Classification,” Fifth IEEE International Conference on Automatic Face and Gesture Recognition (AFGR), Washington, DC, May 20–21.
Hong, J. , Kang, J. , and Price, M. , 2012, “ Gait Analysis and Identification,” 18th International Conference on Automation and Computing (ICAC), Loughborough, UK, Sept. 7–8. http://ieeexplore.ieee.org/document/6330517/
Gabel, M. , Gilad-Bachrach, R. , Renshaw, E. , and Schuster, A. , 2012, “ Full Body Gait Analysis With Kinect,” 34th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), San Diego, CA, Aug. 28–Sept. 1, pp. 1964–1967.
Sun, B. , Liu, X. , Wu, X. , and Wang, H. , 2014, “ Human Gait Modeling and Gait Analysis Based on Kinect,” 2014 IEEE International Conference on Robotics and Automation (ICRA), Hong Kong, China, May 31–June 7, pp. 3173–3178.
Kakizaki, T. , Urii, J. , and Endo, T. , 2012, “ A Three-Dimensional Evacuation Simulation Using Digital Human Models With Precise Kinematic Joints,” ASME J. Comput. Inf. Sci. Eng., 12(3), p. 031001. [CrossRef]
Kakizaki, T. , Urii, J. , and Endo, T. , 2014, “ Post-Tsunami Evacuation Simulation Using 3D Kinematic Digital Human Models and Experimental Verification,” ASME J. Comput. Inf. Sci. Eng., 14(2), p. 021010. [CrossRef]
Kakizaki, T. , Urii, J. , and Endo, T. , 2015, “ Experimental Study of an Airplane Accident Evacuation/Rescue Simulation Using Three-Dimensional Kinematic Digital Human Models,” ASME J. Comput. Inf. Sci. Eng., 15(3), p. 031006. [CrossRef]
GoÌ^tz-Neumann, K. , Tsukishiro, K. , Yamamoto, S. , and Ehara, Y. , 2005, The Gait Analysis by Observation, Igaku Shoin, Tokyo, Japan (in Japanese).
Saito, H. , 2011, Clinical Kinematics, 3rd ed., Ishiyaku Publishers, Tokyo, Japan, p. 486 (in Japanese).
Nara, I. , 2001, Standard Physiotherapy Studies—Specialized Field (Clinical Movement Analysis), Igaku-Shoin, Tokyo, Japan, p. 128.
Saito, E. , 2015, “ Walk Analysis and Movement Analysis,” Fujita Health University Rehabilitation Section, Tokyo, Japan, p. 86 (in Japanese).
Allen, R. , Guthrie, M. , and Buford, J. , 2008, “ Gait I: Overview, Overall Measures, and Phases of Gait,” Course Note of Kinesiology Course Material, Washington State University, Pullman, WA, pp. 11–28.
Copyright © 2017 by ASME
Topics: Kinematics
View article in PDF format.

References

Cabinet Office of Japanese Government, 2015, “ Ageing Society White Paper,” Cabinet Office of Japanese Government, Tokyo, Japan, accessed Jan. 6, 2016, http://www8.cao.go.jp/kourei/whitepaper/w-2015/gaiyou/pdf/1s1s.pdf (in Japanese)
Japanese Physical Therapy Association, 2015, “ Membership Statistics,” Japanese Physical Therapy Association, Tokyo, Japan, accessed Dec. 2, 2015, http://www.japanpt.or.jp/aboutpt/physicaltherapy/ (in Japanese)
Saito, H. , 2013, “ The History of the Physiotherapy in Japan, Vol. 2,” Bulletin of Institute of Tokyo Medical Care University, University of Tokyo Health Science, Tokyo, Japan (in Japanese).
Machiel, F. , and Reinkensmeyer, D. , 2008, “ Rehabilitation and Health Care Robotics,” Springer Handbook of Robotics, Springer, Berlin, pp. 1223–1251.
Broadbent, E. R. , Stafford, R. , and MacDonald, B. , 2009, “ Acceptance of Healthcare Robots for the Older Population: Review and Future Directions,” Int. J. Soc. Rob., 1(4), pp. 319–330. [CrossRef]
Smarr, C.-A. , Fausset, C. B. , and Rogers, W. A. , 2011, “ Understanding the Potential for Robot Assistance for Older Adults in the Home Environment,” Human Factors and Aging Laboratory Technical Reports, Georgia Institute of Technology, Atlanta, GA, pp. 6–24.
Isezaki, M. , Takano, K. , and MochidukiI, Y. , 1999, “ Relationship Between Quality of Life (QOL, Life Satisfaction) and Activities of Daily Living (ADL) in Elderly People With Diseases,” Vol. 16, Bulletin of Yamanashi Medical University, Yamanashi Medical University, Kofu, Japan, pp. 71–75 (in Japanese).
Hirose, N. , Nakayama, A. , and Maruyama, H. , 2014, “ Educational Effect of Physical Therapy With the Use of Information and Communication Technology—New Attempt to the Fixing of the Ability of Problem Solving Leading to Clinical Reasoning,” Jpn. J. Sports Rehabil., 3, pp. 37–42 (in Japanese).
Shimada, S. , Suzuki, A. , Yonemura, S. , and Kojima, A. , 2012, “ A Presentation Method of Image Information for Sports Motion Analysis by Video Viewing,” J. Inst. Image Electron. Eng. Jpn., 41(1), pp. 65–72.
Ijaz, A. , Gibbs, L. , Abboud, R. , Wang, W. , Ming, D. , and Wan, B. , 2012, “ Analysis of Knee Joint Kinematics During Walking in Patients With Cerebral Palsy Through Human Motion Capture and Gait Model-Based Measurement,” IEEE International Conference on Virtual Environments Human–Computer Interfaces and Measurement Systems (VECIMS), Tianjin, China, July 2–4.
Tao, W. , Liu, T. , Zheng, R. , and Feng, H. , 2012, “ Gait Analysis Using Wearable Sensors,” Sensors, 12(2), pp. 2255–2283. [CrossRef] [PubMed]
Lee, L. , and Grimson, W. , 2002, “ Gait Analysis for Recognition and Classification,” Fifth IEEE International Conference on Automatic Face and Gesture Recognition (AFGR), Washington, DC, May 20–21.
Hong, J. , Kang, J. , and Price, M. , 2012, “ Gait Analysis and Identification,” 18th International Conference on Automation and Computing (ICAC), Loughborough, UK, Sept. 7–8. http://ieeexplore.ieee.org/document/6330517/
Gabel, M. , Gilad-Bachrach, R. , Renshaw, E. , and Schuster, A. , 2012, “ Full Body Gait Analysis With Kinect,” 34th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), San Diego, CA, Aug. 28–Sept. 1, pp. 1964–1967.
Sun, B. , Liu, X. , Wu, X. , and Wang, H. , 2014, “ Human Gait Modeling and Gait Analysis Based on Kinect,” 2014 IEEE International Conference on Robotics and Automation (ICRA), Hong Kong, China, May 31–June 7, pp. 3173–3178.
Kakizaki, T. , Urii, J. , and Endo, T. , 2012, “ A Three-Dimensional Evacuation Simulation Using Digital Human Models With Precise Kinematic Joints,” ASME J. Comput. Inf. Sci. Eng., 12(3), p. 031001. [CrossRef]
Kakizaki, T. , Urii, J. , and Endo, T. , 2014, “ Post-Tsunami Evacuation Simulation Using 3D Kinematic Digital Human Models and Experimental Verification,” ASME J. Comput. Inf. Sci. Eng., 14(2), p. 021010. [CrossRef]
Kakizaki, T. , Urii, J. , and Endo, T. , 2015, “ Experimental Study of an Airplane Accident Evacuation/Rescue Simulation Using Three-Dimensional Kinematic Digital Human Models,” ASME J. Comput. Inf. Sci. Eng., 15(3), p. 031006. [CrossRef]
GoÌ^tz-Neumann, K. , Tsukishiro, K. , Yamamoto, S. , and Ehara, Y. , 2005, The Gait Analysis by Observation, Igaku Shoin, Tokyo, Japan (in Japanese).
Saito, H. , 2011, Clinical Kinematics, 3rd ed., Ishiyaku Publishers, Tokyo, Japan, p. 486 (in Japanese).
Nara, I. , 2001, Standard Physiotherapy Studies—Specialized Field (Clinical Movement Analysis), Igaku-Shoin, Tokyo, Japan, p. 128.
Saito, E. , 2015, “ Walk Analysis and Movement Analysis,” Fujita Health University Rehabilitation Section, Tokyo, Japan, p. 86 (in Japanese).
Allen, R. , Guthrie, M. , and Buford, J. , 2008, “ Gait I: Overview, Overall Measures, and Phases of Gait,” Course Note of Kinesiology Course Material, Washington State University, Pullman, WA, pp. 11–28.

Figures

Grahic Jump Location
Fig. 1

Projected growth of elderly population in Japan

Grahic Jump Location
Fig. 2

KDH model and its joint arrangements

Grahic Jump Location
Fig. 3

Joint-centered coordinate frames in KDH model

Grahic Jump Location
Fig. 4

Gait motion classification by RLANRC method

Grahic Jump Location
Fig. 5

Schematic of gait motion KDH models construction

Grahic Jump Location
Fig. 6

Series of normal gait motion images corresponding to initial postures of eight representative phases of RLANRC method

Grahic Jump Location
Fig. 7

Gait motion KDH models for normal gait motion

Grahic Jump Location
Fig. 8

Gait motion KDH model-based coordinate frames

Grahic Jump Location
Fig. 9

Gait motion photographs of right hemiplegic patient [20]

Grahic Jump Location
Fig. 10

Initial KDH models of right hemiplegic patient

Grahic Jump Location
Fig. 11

KDH models corresponding to initial postures of eight representative phases of RLANRC method for subject with normal gait (similar to Fig. 7)

Grahic Jump Location
Fig. 12

Four normal gait postures corresponding to photographs of right hemiplegic patient shown in Fig. 9

Grahic Jump Location
Fig. 13

Appropriate RLANRC positions of initial KDH models of hemiplegic patient shown in Fig. 10

Grahic Jump Location
Fig. 14

RLANRC KDH models of hemiplegic patient gait

Grahic Jump Location
Fig. 15

Gait motion KDH models obtained by RLANRC-based modeling

Grahic Jump Location
Fig. 16

Visualizations of gait motion KDH model in 3D space: (a) side view, (b) front view, (c) top view, and (d) orthographic view

Grahic Jump Location
Fig. 17

Relationship between RLANRC gait phase and right hip joint angle: (a) normal gait data [23], (b) normal gait motion KDH model, and (c) hemiplegic gait motion KDH model

Grahic Jump Location
Fig. 18

Relationship between RLANRC gait phase and right knee joint angle: (a) normal gait data [23], (b) normal gait motion KDH model, and (c) hemiplegic gait motion KDH model

Grahic Jump Location
Fig. 19

Relationship between RLANRC gait phase and right ankle joint angle: (a) normal gait data [23], (b) normal gait motion KDH model, and (c) hemiplegic gait motion KDH model

Grahic Jump Location
Fig. 20

Continuous images of gait motion represented by patient KDH model before treatment: (a) side view and (b) front view

Grahic Jump Location
Fig. 21

Continuous images of gait motion represented by patient KDH model after treatment: (a) side view and (b) front view

Grahic Jump Location
Fig. 22

Detail of the joint coordinate frame arrangement in the KDH model

Tables

Table Grahic Jump Location
Table 1 Gait parameters used in the KDH model
Table Grahic Jump Location
Table 2 Patient profile data
Table Grahic Jump Location
Table 3 Findings of gait analysis (before treatment)
Table Grahic Jump Location
Table 4 Findings of gait analysis (after treatment)
Table Grahic Jump Location
Table 5 Walking motion data of patient

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In