​Djavad Mowafaghian
​​​​​​​ Research Center for Intelligent NeuroRehabilitation Technologies


No.11, Khark St., Enghelab St., Tehran, Iran

Past Project

One of the most common issues of cerebral palsy (CP) patients is movement dysfunction and foot deformity. The dynamic function of the foot can be assessed by Center of Pressure Progression (CoPP) and Ground Reaction Force (GRF), as has been investigated in normal subjects. So, the purpose of this research is using CP’s gait data to find new differences between any type of CP that were defined till now and classification type of CP’s walking issue and suggest new method to diagnose CP patients more effectively.​​​​​​​

Research Team: Ali Salehi
Supervisors: Hasan Zohoor, Farzam Farahmand, Mina Arab Baniasad

Classification Cerebral Palsy by the Differences in Patterns of Center of Pressure Progression and Ground Reaction Force

Now it is quite evident that during the walking, components of the ground reaction force and the route of the center of mass of the whole body are affected by the movement of the hands, trunk and walking speed, and by changing these values, the vertical oscillations of the center of mass are changed and energy consumption increases. Due to the interactions between the upper extremities and lower extremities in gait, various studies have been conducted, but they have many contradicted in biomechanical effects (kinetics, kinematics, and time spatial variables) of this issue. As upper limb oscillations effect on gait and lower extremities, it can be Indicates as the importance and emphasis on how and how to upgrade the upper limb in walking and gait rehabilitation. Considering the new findings regarding the importance of upper limb oscillations in improving dynamic stability during walking, the importance of knowing the biomechanical effect of upper limb movement on the lower limb and gait is very necessary for medicine, sports instructors, rehabilitation practitioners and mechanical engineers. However, the importance of this principle is not well documented and is usually considered upper limbs and trunk together as HAT in analyses, and most of the time the main focus is on the lower extremities. In the present study, it is hoped that a study of the effects of changes in the upper limb movement pattern on the behavioral biomechanical gait variables could be achieved by a comprehensive approach to the impact of these changes.​​​​​​​

Research Team: Razieh Yousefian Molla
Supervisors: Farzam Farahmand​​​​​​​


The Effect of Upper Extremity Movement Pattern changes on Spatio-Temporal parameters and 3D-Mechanical Muscle Power of Lower Limbs during Gait of Active Females

The purpose of this project is to investigate the effect of learning on kinematic and muscle synergies.

In this study we first put the subjects on a balance system which rotates randomly to various directions and ask them to control their balance while the system is working. Participants in the study were tested at 2 time points: (a) before any balance practice and (b) after some practice sessions. During the tests we collect EMG signals from different leg and trunk muscles. We also collect kinematic data with IRED markers to calculate the joint angles with Vicon Nexus system.

We hypothesized that training leading to skilled motor performance alters muscle and movement coordination during challenging as well as nominal everyday motor behaviors. We want to compare motor modules between professionally trained balance subjects and novices. We predicted that experts would use more motor modules than novices and also have modules composed of fewer coactive.

Research Team: Sina Esmaeili, Hojat Karami
Supervisors: Farzam Farahmand,  Mina Arab Baniasad

Studying the effect of motor learning on joint motion coordination and muscle synergies during balance control

In this study, the both sole foot geometry of 22 persons were evaluated by 3D scanning in weight bearing and none weight bearing. First, the subjects were asked to stand normally for data acquisition in the weight bearing mode. Next, they were asked to stand on one foot, while the other foot only touched the sole glass of scanner and didn’t exert force by that foot. Here, the abandoned foot that was in none weight bearing mode, was scanned. During the experiment, we tried to position the leg perpendicular to the scanner glass. Also it was essential that all points of sole touched the glass especially fingers.

To have 3D data acquisition repeatability of different subjects, we made an effort to have a same volume analysis. For achieving this purpose, we pasted a cotton by a double-sided adhesive on malleolus bones (the ankles) to have these parts more prominent. It is possible to eliminate added volume in CAD. After doing the experiments, some criteria based on the empty area and cavity between sole and ground were proposed.

Volume criteria is defined as the ratio of the cavity between the glass of scanner and foot sole to foot volume. In this method, 2 mallei are connected by a line. The midpoint of this line is assumed. A plane which is parallel to the sole plane, is crossed this point. All bulk above this plane is deleted and the remaining bulk is foot volume (fig 1A). The distance between this plane and sole plane is named foot height. 10% of foot height is considered as foot sole height. In the following, a plane that is located in the foot sole height from the sole is crossed; the above bulk is removed to achieve foot sole volume (fig 1B). To attaining the cavity volume, an insole is designed to fill it (fig 1C and 1D). Finally, cavity volume is divided to foot volume and dimensionless and comparable volume criteria for every persons is gained.

Surface criteria is defined as the ratio of tangential surface to foot sole surface. In this case we first tangent a line to the medial part of the foot sole in a way that it begins from the hindfoot near heel and runs to the forefoot. The area formed between this line and the medial part of the foot, is considered as the tangential surface (fig 2). at the end, this area is divided to foot sole area and results another dimensionless number.

Results and discussions

 3D scanning results and surface and volume criteria are calculated and presented in table 1. the results are sorted in 3 columns; weight bearing, none weight bearing and the ratio of none weight bearing to weight bearing. All people with flatted feet, are highlighted in red. Normal people are highlighted in green.

According to results and doctor diagnosis after clinical assessment, the criteria for flatfoot diagnosis is defined below for surface analysis: those who has a bigger number than 2 in  column in surface analysis or has number less than 1 in WB column have flatted feet. The accuracy of the numbers is about 0.1%. it is essential to mention that 2 is not used absolutely; it means that from 1.9 to 2 subjects are placed in hesitate margin for more evaluations. This criterion can recognize those with acute situation. It is observed that there is no way to tangent a line to the medial part in weight bearing for this group (fig 3). it is obvious in the figure that there is an outbound (بیرون زدگی) in the hindfoot near heel which prevents the formation of tangent surface. After clinical assessment, it is noticed that these subjects have acute flatfoot. Their cure is surgery.

After clinical assessment, it is observed that there is no relation between flatfoot and volume criteria.

Research Team: Alimohammad Kooshesh
Supervisors: Farzam Farahmand

A new quantitative criterion for evaluating flatfoot by 3D scanning

Gait analysis requires the timing of a series of events, including Heel Contact, Toe Contact, Heel Off and Toe Off. These events are used to determine the sub-phases in gait cycles and help normalize gait data. These foot contact events are regularly recorded either using force plates or foot switches, or by manually analyzing kinematic data. These data collection methods may not be suitable for several reasons. Force plate size and position is not always suitable for subjects with short step lengths, and only one stride can be recorded using this method per trial. Also, Heel Off and Toe Contact is not detectable using force plates. Foot switches are commonly used to collect temporal parameters, but this data is not always available during gait analysis. Kinematic data collected by visual markers is the most common method used for gait event detection. But this analysis strongly differs between analysts and the repeatability for one analyst is also not high. This method also needs an experienced analyst in order to have acceptable results and is quite time consuming. Therefore, finding an autonomous gait event detection method that uses kinematic data can greatly increase the accuracy and reliability of gait analysis.

In our project, we design and implement a fuzzy inference system for the recognition of human gait phases in normal subjects. Two experienced personnel detected the events separating the main sub phases of the stance phase (Heel Contact, Toe Contact, Heel Off, Toe Off) and these detected frames were used as the gold standard in this project. We then examined the velocity, acceleration, and jerk of lower limb markers, calculated from previously recorded trajectories, in order to determine the appropriate rules for differentiating between gait phases. In order to find the best parameters to separate these events, we calculated the average and standard deviation of each parameter at the time of the detected event, and also 10% of a gait cycle before and after each event. Parameters with the lowest STD and greatest difference in between the event frame and the moments before and after each event were selected as suitable parameters and the fuzzy system was implemented on these events. We then tested the completed system using data measured from healthy subjects and results were compared with events found by human experts​​​​​​​.


Research Team: Yasamin Foroutani, Reihaneh Jahedan
Supervisors:  Farzam Farahmand, Mina Baniasad

Gait Phase Detection Using a Fuzzy Logic System


Diagnosis of Anterior Cruciate Ligament Injury Using Inertial Sensors​​​​​​​


Introduction: Rapid and low-cost diagnosis of anterior cruciate ligament (ACL) injury is important to prevent further complications. The objective of this study was to investigate the capability of inertial sensors in detecting the ACL rupture.

Method: A finite element model of the knee joint was developed using the CT and MRI data of a male subject, with the ligaments represented by tensile springs. The change in the linear acceleration of the tibia relative to femur, after simulating the ACL rupture, was predicted by the model in a number of different tests. The tests with substantial differences in the maximum acceleration peaks (MAP) were identified as those more likely to detect the ACL injury in practice. The identified tests were applied to the two knees of eight ACL-ruptured patients, four in conscious state and four under anesthesia. Each test was repeated three times and the tibiofemoral joint accelerations were recorded by the inertial sensors.

Results: The model predicted that the anterior drawer, Lachman, pivot shift, reverse pivot shift and lunge tests produce the highest MAP differences. The experimental results indicated an acceptable repeatability in the MAP data recorded for each knee in three repetitions of each test (coefficient of variation < 0.25). The mean of the MAP of the injured knee was higher than that of the healthy knee, for all individuals during the anterior drawer test and for subjects under anesthesia during the pivot shift test. Other tests and states did not provide consistent results.

Conclusion: Based on the early results, the inertial sensors are capable of detecting the ACL rupture during anterior drawer test in both conscious and anesthesia states, and during the pivot shift test under anesthesia.​​​​​​​

Research Team: Reza Ahmadi, Arash Sherafat Vaziri, Mohammad Naghi Tahmasebi
Supervisors: Farzam Farahmand

​​​​​​​Design and Development of a Mobility Recognition System in PD Patients for Tele-Rehabilitation

The purpose of the current work is to design an affordable and accurate device for mobility activity recognition in Parkinson’s patients. The type of sensors and the optimum number were found to minimize the cost while having fair accuracy. 34 activities were selected as the target: LSVT-BIG home training, fast and slow walking, going up and down the stairs, passing obstacles, sit and stand posture, standing up and sitting down, and turning. The collected data was separated within 2.5-second windows. Time and frequency domain as well as wavelet transform parameters were used for feature computation. Through a PCA algorithm, the best combination of features has been extracted. Among 4 classifiers trained for activity recognition, k-NN was the best with the accuracy of 99.7% and sensitivity of 94.1% using left thigh and shank, hip and left forearm sensors.​​​​​​​

Research Team: Amin Mohammadi-Nasrabadi,Amirhossein Tariverdi, Ahmadreza  Eslaminia
Supervisors: Saeed Behzadipour,Laila Alibiglou, Ghorban Taghizadeh​​​​​​​

Physical vs. VR Dual Task Exercises in Post-stroke Patients​​​​​​​

The purpose of this study was to investigation the effect of environmental exercises (physical and virtual reality environment) on the brain reorganization, molecular parameters and behavioral functions (cognitive and motor) in the chronic stroke patients.

In this study, participants participated randomly in 3 groups as follows.
Patients with stroke who do not receive any cognitive-motor exercise (control group)
Patients with stroke who take cognitive-motor physical exercises in the physical environment.
Stroke patients who receive cognitive-motor-physical exercises in the virtual reality environment.
The exercises consist of four categories of cognitive assignments (work memory, attention, response control, and executive function) as well as balance and movement acticities.

The following assessments are done on the subjects before and after receiving the treatments: postural control (force plate), upper limb movement and cognitive (functional tests), serum levels of brain-derived growth factor (immunohistochemical tests) and structural changes in various brain regions (laboratory imaging tests).​​​​​​​

Research Team: Soheila Fallah
Supervisors: Mohammad Taghi Joghataei, Ghorban Taghizadeh, Saeed Behzadipour

Development of integral indices based on a biomechanical model to evaluate upper limb motion quality of stroke patients​​​​​​​

Despite the importance of evaluation of upper-limb motion quality for diagnostic purposes and treatment process of post-stroke patients, the effectiveness of the present assessment methods remains controversial. The existing qualitative and quantitative motion quality assessment approaches, however, suffer from several shortcomings. Qualitative analyses, that are based on questionnaires and therapist’s observations, have subjective components. Furthermore, they fail to adequately evaluate short-term progresses in the patients’ physical performance as they are based on limited scoring scales. Quantitative approaches, that are based on the measurements of upper-limb kinematics are less subjective but their validity, reliability and responsiveness have been questioned. To resolve these shortcomings one needs to introduce a non-subjective and reliable motion quality index that can adequately describe patients’ performance during the treatment process. To this end, development of a new quantitative index has been set as the main objective of this study. The proposed index, is determined as an integral function of a biomechanical model signals (e.g. joints torque, muscles stress and joints compression forces) that have been barely used before. Aggregation of upper-limb motion information by integrating of the mentioned signals results in a more accurate motion quality evaluation. The proposed index will be evaluated based on experimental investigations.​​​​​​​
Research Team: Majid Abedi
Supervisors: Ghorban Taghizadeh, Saeed Behzadipour

Design of an Adaptive Virtual Reality Rehabilitation Exercise Based on Motor Fatigue Prediction for People with Stroke​​​​​​​

Stroke has been known as a major neural deficit leading to motor inabilities during the last decades. Benefitting from the concept of neuroplasticity, rehabilitation is the most efficient approach to compensate for such impairments. Recently, virtual reality (VR) solutions have been developed to facilitate the process. Nevertheless, one of the main complaints and sources of discouragement for the patients during exercises is motor fatigue. To this point, there is no adaptive VR system known which considers real-time fatigue index using an established measurement method or a fairly accurate model. Such a system will be the main concern of this thesis.
The proposed research is aimed to develop a fatigue-based adaptive VR rehabilitation exercise (upper limb reaching) to postpone fatigue occurrence and prolong the rehabilitation session. In order to predict the fatigue raised during the exercise, musculoskeletal modeling coupled with current dynamic muscle fatigue models is proposed. In fact, muscle forces required to produce the desired kinematics are computed using a patient-specific inverse dynamic model. Next, by feeding the muscle force history into a dynamic fatigue model, the fatigue index can be predicted. This index can be further used to tune the difficulty parameters of the exercise in real time using a control strategy.​​​​​​​

Research Team: Maliheh Fakhar 
Supervisors: Hassan Zohoor, Ghorban Taghizadeh, Saeed  Behzadipour

Quantitative Assessment of Parkinson Patient’s Upper Extremity Movements, Using Microsoft Kinect for Telerehabilitation​​​​​​​

Impairments caused by PD, in the event of progression, lead to a person’s dependency and inefficiency.

In this project, a VR assessment system is being developed to provide a better understanding of the upper body movement performance for the therapist. The system uses the Microsoft Kinect sensor as a motion capture device. Several exercise applications have been developed to test the motor performance of the patients. Various performance metrics that of the kinematic (kinematic biomechanical indicators) are calculated. A good performance metric needs to be valid, sensitive, and reliable which are analyzed in this research. Also, combinations of such metrics are under investigation to find the best correlation with the clinical assessment tools.

Research Team:
 Seyed Mostafa Alavian
Supervisors: Ghorban Taghizadeh, Saeed Behzadipour

Development of a tele-assessment system for balance in Parkinson's disease using a force plate​​​​​​​

The purpose of this research is to develop a tele-assessment system for balance using a force plate in patients with Parkinson’s disease. This system, as a subset of a tele-rehabilitation system, can reduce costs and create more comfort for patients and therapists. To assess the balance in this system, dynamic posturography is used, including Limits of Stability and Random Control tests. In these tests, the center of pressure signal is recorded using a force plate during balance tasks, and dynamic posturography measures are computed. In the current study, the reliability, validity, sensitivity and specificity of these measures will be evaluated.​​​​​​​

Research Team: Kosar Barati, Amir Zahedi, Naimeh Alizadeh
Supervisors: Saeed Behzadipour,Ghorban Taghizadeh​​​​​​​

The Development of a Postural Control Model for Parkinson’s Disease to Predict the Balance Rehabilitation Effects​​​​​​​

Parkinson disease (PD) patients seriously suffer from instability and impaired postural control. Rehabilitation helps patients to recover their ability through long-term practical sessions. For designing optimal tasks in each session, and in agreement with each patient’s state, it is essential to develop a computational model of such neuro-rehabilitation/neuroplasticity process. In this study, we aim for developing a postural control model of Parkinson’s disease (PD) with the capacity of implementing balance exercises and predict their effects on the improvements of the balance performance. In order to validate the predictions of the model about the motor learning dynamics of the rehabilitation process, a group of PD patients was given specific exercises with ‘Balance Robot’ (developed in Intelligent Therapeutic Instruments laboratory, Djavad Mowafaghian Research Center) for 18 sessions (2 months). During this new intervention, the balance performance of patients was evaluated with both laboratorial (in each session) and clinical assessments (before, mid, and after the intervention). This first-ever data of rehabilitative dynamics in PD patients, particularly in view of a computational model, will bring us a fresh insight into the factors contributing to the postural instability in PD as well as the factors improving via rehabilitation programs.​​​​​​​

Research Team: Zahra Rahmati
supervisors: Saeed Behzadipour , Ghorban Taghizadeh

Stroke neuromusculoskeletal modeling in order to understand rehabilitative interventions​​​​​​​

The goal of this project is to develop a computational model that can be used to simulate the neuromuscular system of the upper extremity, in order to analyze the movement execution, and to plan rehabilitation programs. This model is a subject-specific computational neurorehabilitation model of arm movements in stroke patients. The integration of an artificial neural network and a musculoskeletal model helps to simulate various intervention programs and their effects on the movement qualification.

A new device for the measurement of upper-limb kinematics has been designed and constructed to collect data needed for the modeling. The device measures motion parameters of the planar movement of the hand in an augmented reality environment that patient should hit and track targets.

50 stroke patients participate in this study who track during their training program. This project will provide pre-clinical insights into neuromuscular and brain functions to allow us to reach a deeper understanding of neuroplasticity and propose the optimum rehabilitation approach.​​​​​​​

Research Team: Majid Hajihosseinali
Supervisors: Ghorban Taghizadeh,  Saeed Behzadipour

Quantification of the muscle fatigue in functional rehab exercises​​​​​​​

Neuromuscular fatigue has an essential impact on the efficiency of clinical rehabilitation in post stroke patients but received little attention in the literature. The aim of this research is to devise an easy-to-use method for detection and quantification of fatigue level in stroke patients while performing a functional upper extremities rehabilitation exercises. This may be used by the therapist or a VR rehab system to optimally adjust the type and intensity of the exercise for best performance.​​​​​​​

Research Team: Alireza Zare 
Supervisors: Ghorban Taghizadeh, Saeed Behzadipour

Robotic Rehabilitation​​​​​​​

Lower extremity exoskeletons are among the most sought after technologies to replace wheelchair in patients with gait disabilities.

The objective of this research is to develop a methodology for gait design in these robots using arm crutch. These exoskeletons are usually under-actuated providing active torque only to hip flexion/extension and knee flexion/extension. Lacking ankle torque complicates the gait design. For this purpose, a biomechanical model of the body linked to the mechanical model of the exoskeleton is developed to determine the feasibility of a given gait pattern in terms of the stability and traction subject to the limits on the admissible joint torques and forces in the body and the robot. The next step is to use this model to find the optimum gait pattern for a given exoskeleton and move on to optimum design of the mechanics and drivetrain.​​​​​​​

Research Team: Reza Norouzzadeh
Supervisors: Saeed Behzadipour​​​​​​​

The effects of the connection stiffness of robotic exoskeletons on the gait quality and comfort​​​​​​​

Increase in the interaction forces/torques at the exoskeleton-human connection due to kinematic mismatches may consequently result in the user discomfort and/or lower performance of the exoskeleton. The stiffness of the exoskeleton-human connection elements plays a key role in this issue. The purpose of this study is to assess the effects of the exoskeleton-human connection stiffness on the user comfort and limb tracking error during normal gait. A biomechanical model of the leg was built and connected to a model of an exoskeleton by elastic connections whose stiffness tensors were identified experimentally for one study participant. The effects of the connection stiffness on the gait performance was investigated using two indices: discomfort index, DI, and tracking error (i.e., the difference between the human joint angle and that of the exoskeleton), TEI.  DI was calculated based on the mechanical energy stored in the elastic connections elements. Although increase of the stiffness magnitude in every direction results in DI growth, the torsional stiffness of the shank connection was found to be the most sensitive for DI. TEI, on the other hand, showed both increasing and decreasing trends with the stiffness increase in different directions. It was found that there are optimal values for some connection stiffness elements, especially shank connection, possibly due to intricacy of the knee joint, that would improve both DI and TEI. For instance, decrease of the torsional and mediolateral shank connection stiffness improved DI and TEI by an average of about 50%.

Stiffness measurement of the thigh connection: (a) The coordinate systems (CS) of the F/T sensor and anatomical CS (b) The loading handle and chair CS (c) The experimental setup arranged for the thigh connection

Research Team: Morteza Shafiei
Supervisors: Saeed Behzadipour

Telerehabilitation for postural instability in patients with Parkinson's disease​​​​​​​

Telerehabilitation for postural instability in patients with Parkinson’s disease.
The first purpose of this study is to investigate the effect of a package of exercise in telerehabilitation form, on the brain reorganization, molecular parameters and behavioral functions in patients with Parkinson’s disease.

The following assessments are done on the subjects before and after receiving the treatment:

behavioral assessment: postural control (force plate, EMG, motion analysis), cognition and mood

Molecular assessment: serum levels of brain-derived growth factor

Brain reorganization assessment: structural changes in various brain regions (laboratory imaging tests)

The second purpose of this research is to study Parkinson’s patients’ postural control in multitasking condition by considering clinical phenotype of Parkinson’s disease.​​​​​​​

Research Team: Naeemeh Haji alizadeh, Kosar Brati, Amir Zahedi
Supervisors: Mohammad Taghi Joghataei, Ghorban Taghizadeh, Saeed Behzadipour​​​​​​​