RESNA Annual Conference - 2019

EConHand: A Wearable BCI System for Stroke Rehabilitation

Yao Xu, Zhun Qin, Xinyu Sun, Mechanical Engineering, Shanghai Jiao Tong University

Jianghong Fu, Rehabilitation Medicine, Fudan University

1. Problem Statement/Research Question and Background

A stroke is a medical condition when inadequate blood flow to the brain results in cell death, which is usually caused by blood vessel burst or vessel blocked. The stroke may arouse the symptoms like loss of functions of movement, sensory, speaking, problem understanding, vision or even death.

With the social development and urbanization acceleration, an increasing number of residents are leading an unhealthy lifestyle, resulting in more and more cardiovascular disease occurrences. According to World Health Organization report in 2014, of the 17.5 million deaths due to cardiovascular disease in 2012, approximate 6.7 million were due to stroke [1]. In 2015 there were about 42.4 million people who experienced a stroke and were still alive [2]. However, 75 per cent of the survivors are suffering from the motor function loss in different levels.[1] The nursing and rehabilitation for these individuals has becoming a serious social issue.

The motor function of upper limbs plays an important role in our daily life and the deficiency of motor function caused by stroke will directly lead to the loss of activities of daily living (ADL). So the rehabilitation therapies have raised worldwide considerable concern for decades. Traditional methods require persons with stroke to passively receive treatment, such as mechanical traction, massage, electrical stimulation and so on, which often fail to achieve desired clinical outcomes. Since 21st century, active therapies, like mirror therapy and motor imagery, have appeared and offered more choices for stroke rehabilitation. In recent year, thanks to the development of neuroscience and computer science, the burgeoning brain-computer-interface (BCI) techniques have successfully combined the active therapies with robot techniques [3]. Many clinical and academic groups have proven the viability and usefulness of BCI- based active therapies in stroke rehabilitation [4]. But most of the active BCI systems are expensive, impractical and unavailable for common users, thus impeding the promotion of the emerging technique.

Our eConHand is an active wearable BCI system for hand function rehabilitation of persons with stroke. It consists of three devices and is a relatively cheap and available system which is suitable for clinical usages in hospitals.

2. Methods/Approach/Solutions Considered

The working principle of BCI rehabilitation system is based on neuroplasticity. When a person with stroke performs motor imagery (MI) or motor execution (ME), his/her motor cortex will be activated, which can be measured by electroenphalogram (EEG) devices. If we can precisely control the movement of the exoskeleton hand on users’ hand coupling in time domain to their own intentions, a better clinical rehabilitation outcome will be achieved according to the literatures. So there are four main components in our active rehabilitation system.

First, we needed to design a special EEG cap to collect the signal. Most of the commercial EEG cap are not only expensive but also hard to fit various sizes of heads. Meanwhile there are little research on the number of electrodes required and adjustment design for different heads. So we have conducted series experiments in the laboratory to answer these questions and come up with a new structure.

Second, we needed an exoskeleton hand to open and close the hand with disability. The basic design of the exoskeleton was a two-freedom mechanical structure which could realize the movements of grasp and release. Actually, it was not hard to design an exoskeleton hand, but how to design it ergonomically mattered. Considering the high muscular tension and possible morbidity in persons with stroke, we chose to combine the exoskeleton with a half glove in order to make it comfortable and easy to wear.

Third, we needed to design a control box to decode the EEG signal and control the movement of exoskeleton hand. Since the BCI required special bio-signal collection and data computation circuit, we chose the STM32 chip as the main processor and designed a single PCB board. For the decoding program, we transformed a machine learning method from MATLAB to C++ code. This method required calibration before decoding the data.

Forth, we needed to design a receiver for controlling the exoskeleton hand. This receiver also needed to offer the flashing lights to guide the movement of users. So, we design a PCB board with Bluetooth 4.0 and five LED lights on it.

When designing and testing the prototypes of the system, we found it tough to satisfy both the ergonomic design and functional requirements. The original design of the exoskeleton hand was not suitable in real situation as the center of gravity was located on the front part. Moreover, the overall accuracy for intention detecting were relatively low due to the simple classification algorithm. Then we have upgraded and re-designed the whole system in many aspects to fulfill the requirements.

3. Description of Final Approach and Design

1) EEG Cap Design Description

Figure 1: prototype of EEG cap and 3D model
The EEG cap is called eConMind as a product. It is specially designed to be a plastic light framework with 10 electrodes on it, which is obviously different from nowadays EEG cap. The structure is designed in SolidWorks and produced by 3-D printing. It can adjust the position of all electrodes slightly to fit different heads. These electrodes are placed around the motor cortex area regarding to the physiology and held by the framework. The GND is placed on the forehead and the reference is placed behind the right ear. We have left enough space to expose the upper surface of every electrode in order to inject the conductive glue. The metal parts of electrodes are made with AgCl and we design the trumpet-shaped supporting structures towards the skin so as to keep the glue from leaking and to reduce the contact pressure. The connecting lines of electrodes are hiding inside the structure and gather together at the backside of the cap, where we have left an area for connecting-peg and the control box. In the prototype, we have not hidden all of the lines because of the test and uncertainty. In the final design we are going to connect all of the lines to a standard port-pin peg and fix the peg on the cap. Also, the cap can be washed in the water for cleaning.

The EEG cap can also be connected to some other common BCI experimental equipment with different adapters. For example, we can use our cap to connect to Neuroscan or JAGA devices to collecting brain signals as well. This feature extends the usage of the cap.

2) Exoskeleton Hand Design Description

Figure 2: prototype of hand and 3D model
The execution part of the exoskeleton hand is a two-freedom double rocker structure with one underactuated freedom. The index finger and thumb parts are driven by a stepping motor to realize the functions of grasp. The main holding parts are made with aluminum alloy and the shell are made with plastic by 3-D printing. The shell is designed to fit the human hands.


The receiver is installed on the back of the hand. It sends and receives the commands from the control box and control the movement of the hand. There is a switch and five lights separately placed on the surface. One light represents the battery state; One represents the Bluetooth connection state and the rest 3 lights are used to guide the actions of the practice. We can also change the practice modes by short pressing or holding the switch. In active mode, the receiver will get the commands from the control box on the cap by built-in Bluetooth 4.0 module when we conduct the active BCI therapies. In passive mode, the receiver will directly control the movement of the hand

according to the pre-set paradigms and act as a passive mechanical traction device. Also, we have a charging interface on the receiver. The receiver can be linked to or controlled by other devices like computers, which can be used in other conditions.

3) Control Box Design Description

Figure 3: prototype of control box PCB board and new version
The control box is used to gather and filter the signal, save the data and parameters and send out control signal to the receiver. There is a PCB board and a battery in it. We started by data collection by an amplifier chip, and then save the data. If we finish the data collection part, the STM32 processor will use the CSP and LDA methods to decode the movement hiding in the signal and send the commands to the receiver. The all procedures can be conducted on board with out the help of other additional facilities.


Also, the control box can collect signal from other commercial EEG caps with adapters and send the commands or decode results to the computers. The raw data will be kept in the SD card on the board for academic uses when doing the practice. We are going to put the PCB board and battery into a small box to match our caps, which is a challenge for us..

4) Usage of the System

When using the system, we need to do these steps. Step 1: put on the EEG cap and wear the exoskeleton hand. Step2: inject the conductive glue. Step 3: turn on the switch and wait for connection. Step4: follow the guidance of the flashing lights to take practice. There are two stages when we take practice. The first stage is used for calibration. People are expected to do actions according to lights. Synchronously the control box will collect the data but not send out the movement signal to the receiver. The exoskeleton hand will be kept still in this stage. After calibration, the control box will calculate the parameters used for decoding in a short time. For the second stage, people are still expected to do the same actions according to lights. Meanwhile the control box will save the data in memory card and decode the signal synchronously. If there is a movement intention detected, a command will be sent to the receiver to move the hand, thus building the coupling relationship between brain and hand. The second stage is the main practice procedure and it will take about 45 minutes. After practice, we simply need to disassemble the control box and wash the cap for cleaning up.

4. Outcome

The prototype is being tested under the supervision of the professional therapists. According to the test, the preparation time is 10 minutes for the person on the first try and 5 minutes for experienced users. The practice will last from 30 minutes to 1 hour

depending on the status of the user. When doing the motor execution, the users can get an accuracy of 80% on average in decoding the movement intention. We have conducted experiments on this practice paradigm with other commercial equipment and achieved significant clinical outcome in the past several years. In another aspect, the users consider the system to be more comfortable and easier to take practice comparing to other BCI methods they had experienced in hospital before and are willing to engage in the next rehabilitation stage.

5. Cost

This system is planned to be adapted in hospitals and it should be relatively cheap. The cost of each component is listed below:


6. Significance

EConHand is a brand new BCI system for hand rehabilitation. It helps improve the functional performance of persons with hand function loss in an active, effective way and is affordable, portable for most of the hospital, which is meaningful in rehabilitation practice. The system can also be promoted in rural areas due to its cheapness, thus helping more people return to normal life.

Including the hand function rehabilitation as a whole, each component of eConHand can work separately as an independent module for different purpose. These components benefit from universality and compatibility in cooperation with other devices. Not only the clinical purpose but also the academic data can be attained when we use the system.

7. References

[1] Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015,WTO

[2] Stroke,

[3] Bundy, D. T., Souders, L., Baranyai, K., Leonard, L., Schalk, G., Coker, R., ... & Leuthardt, E. C. (2017). Contralesional brain–computer interface control of a powered exoskeleton for motor recovery in chronic stroke survivors. Stroke, 48(7), 1908-1915.

[4] Pichiorri, F., Morone, G., Petti, M., Toppi, J., Pisotta, I., Molinari, M., ... & Mattia, D. (2015). Brain–computer interface boosts motor imagery practice during stroke recovery. Annals of neurology, 77(5), 851-865.