IpsiHand: Direct Recoupling of Intention and Movement (Washington University in St. Louis)

Tile image shows complete prototype system laid out including 1)Emotiv EPOC EEG recording headset, 2)laptop, and 3)modified hand orthotic for grasp control

Sam Fok, Raphael Schwartz, Mark Wronkiewicz, Charles Holmes, Jessica Zhang, Nathan Brodell, Thane Somers (Washington University in St. Louis)


Stroke and traumatic brain injury (TBI) cause long-term, unilateral loss of motor control due to brain damage on the opposing (contralateral) side of the body. Conventional neurological therapies have been found ineffective in rehabilitating upper-limb function after stroke. Brain computer interfaces (BCIs), devices that tap directly into brain signals, show promise in providing rehabilitation but remain in research. Also, BCIs cannot work if the target signals have been eliminated due to injury. Therefore we present a novel BCI, the IpsiHand, which combines advances in neurophysiology, electronics, and rehabilitation. Recent studies show that during hand movement, the cortical hemisphere on the same (ipsilateral) side of the body as the hand also activates. IpsiHand uses electroencephalography (EEG) to record these signals and control a powered hand orthosis.  The undamaged hemisphere can then control both hands, and through neural plasticity IpsiHand will strengthen ipsilateral neural pathways to enhance ipsilateral motor control.


Stroke and TBI combined are the leading cause of disability in the US, with around a million cases annually; this poses a significant cost to the economy and decreases quality of life for affected individuals. Half report trouble with hand movement, and conventional physical therapy produces little significant improvement after 3 months post injury [1]. Lasting disabilities result in a typical lifetime cost between $100k and $2M per patient, including inpatient care, rehabilitation, and follow-up [2] [3][4]. The most effective therapies have patients actively controlling their limb, which is not an option in cases of severe paralysis. While BCIs promise new hope for treatment, they remain in the research stage.  In addition, conventional BCIs cannot be applied to cases of brain injury since the classical motor signals in cortex contralateral to the target limb needed would be gone with the injury.

We address the problem of applying BCI technology to rehabilitation following stroke and TBI.

We developed a device, IpsiHand, for rehabilitation that synthesizes recent developments in neurophysiology, electronics, and physical therapy into a BCI hand orthosis.  A recent study found signals associated with hand movements in cortex ipsilateral to the hand. These signals were present in cortex anterior to ipsilateral primary motor cortex and at frequencies below 40Hz [5], which are accessible via EEG. To acquire these intent-to-move signals we used a non-invasive, low cost EEG consumer headset to record from cortex, controlling an orthosis which opens and closes a patient’s hand. Tactile and proprioceptive feedback provided from this device will facilitate neural plasticity, strengthening existing and developing new neural pathways ipsilateral to the affected hand that will ultimately restore motor control. Allowing the patient to regain hand control with their thoughts alone should also provide tremendous encouragement in the rehabilitation process.

Our objective is to directly recouple the intent-to-move a hand with hand motion in order to improve outcomes of recovery, reduce the lifetime cost of brain injury, and improve quality of life for those affected by stroke or TBI.


IpsiHand integrates low-cost, commercial EEG acquisition hardware, signal processing software, and a modified orthosis to control hand grasping in real time (Figure 1).

Block diagram of data flow through IpsiHand system. Brainwave EEG data is acquired and digitized by an Emotiv EPOC headset. Digitized data is sent to a laptop for signal processing. BCI2000 and LabView software conduct spatial filtering, signal feature selection, and control signal generation from EEG data. Control signal transmitted via USB to mechanical actuator mounted on adjustable, prefabricated hand orthosis.

Figure 1- Block diagram of data flow through IpsiHand system. Brainwave EEG data is acquired and digitized by an Emotiv EPOC headset. Digitized data is sent to a laptop for signal processing. BCI2000 and LabView software conduct spatial filtering, signal feature selection, and control signal generation from EEG data. Control signal transmitted via USB to mechanical actuator mounted on an adjustable, prefabricated hand orthosis.

  1. Signal Acquisition: An Emotiv EPOCTM EEG headset records EEG signals from the scalp with 14 channels digitized at 128Hz, and transmits wirelessly to a computer.
  2. 2. Signal Processing and Control: EEG is the least invasive recording technique and practical for immediate application in the clinic, but EEG signals tend to be smaller and more spatially diffuse compared invasive recording techniques. Therefore, we must maximize the signal-to-noise ratio. We use a large Laplacian reference [6], which filters noise from wide areas of the scalp, detects signals specific to a particular brain areas, and makes IpsiHand resilient to electrode placement variations. On a computer, BCI2000, a software dedicated to BCI applications [7], extracts a control signal from the EEG data with spatial filtering, frequency analysis and feature selection, and statistical normalization. For optimal control, the user trains the algorithm by alternating between periods of attempted hand movement and periods of rest. After several trials, IpsiHand locates specific areas and brainwave frequencies that consistently changed in power between hand movement and rest conditions. The resulting control signal is sent to LabView software for conversion into actuator commands.
  3. Mechanical Actuation: A Becker Oregon TalonTM prefabricated orthosis (Figure 2) is fitted with a Firgell L16 linear actuator. Driven by signals from LabView, it flexes and extends the patient’s fingers for grasping. The orthosis was chosen for adjustable sizing, which is key in a clinical setting with a variety of patients. To mechanically prevent hyperextension or hyperflexion of the hand, The range of actuator motion is mapped only onto the natural range of finger joint rotation.
Procedure for modifying a Becker Oregon Talon prefabricated wrist driven hand orthosis (Model TAL100):  Wrist-hand linkage is removed, and replaced with Firgelli L16 linear actuator mounted on a modified plate linking the hand and forearm sections of the orthosis. This allows for powered control of hand grasp.

Figure 2 - Procedure for modifying a Becker Oregon Talon prefabricated wrist driven hand orthosis (Model TAL100): Wrist-hand linkage is removed, and replaced with Firgelli L16 linear actuator mounted on a modified plate linking the hand and forearm sections of the orthosis. This allows for powered control of hand grasp.

Our intended treatment plan is designed to integrate with current standards of clinical rehabilitation therapy. A typical intensive post-stroke treatment therapy consists of 36 sessions of 45 minutes over a period of 12 weeks [10]. Fitting IpsiHand is simple with Velcro® straps, and the EPOC headset self-fits. IpsiHand’s signal processing algorithm continually adapts to ensure full range of hand motion even with weak or moving signals. Each session will include of a 5 minute training period in which the algorithm adapts to the subject’s brainwaves.  The identified signal features are then used to control the orthosis during repetitive flexion-extension hand tasks. Patients that purchase IspiHand can benefit from increased therapy time with the device outside of normal therapy sessions.


IpsiHand was tested with three healthy subjects to verify the ability to use non-conventional signals from cortex on one side of the brain to control a hand on the same side of the body. We found that:

1.         Hand movement correlates with signals from the ipsilateral hemisphere (Figures 3 and 4).

2.         IpsiHand successfully uses EEG signals to move the hand (video above).

The identified frequencies and electrode locations used in the study are in Figures 3 and 4 and in our video. During the study, a cursor was controlled by the identified EEG signal feature, which was modulated when a subject moved, or imagined moving his hand. The modulated signal gave the subject control of 1D cursor movement, and the subject was tasked with moving the cursor to a target that randomly appeared on either side of a computer screen. Through 10 sets of trials with non-impaired individuals we were able to achieve an 81.3% success rate for this task. We expect that with optimization of our signal detection algorithms, we can achieve success rates upwards of 90% as it was not uncommon to see success rates upwards of 90%.

Left: EPOC headset electrode positions on head. Center: Correlation values between left hand movement condition and rest condition per channel and frequency bin.  Right: Frequency spectrum of signal of channel F3 showing changes in amplitude between left hand movement and rest

Figure 3- Left: Positions of EPOC headset electrodes. Center: Correlations between left hand movement condition and rest condition per electrode channel (y axis) and per frequency bin (x axis). Electrodes over the left hemisphere are on the lower half of the y axis and electrodes over the right hemisphere are on the upper half of the y axis. Bins with high correlation values can be used to predict whether the subject is moving the left hand or not. Right: Frequency spectrum of electrode channel F3 showing changes in amplitude between left hand movement and rest

Left: Colormap overlaid on head of correlation values between left hand movement condition and rest condition in 12Hz brainwaves. Shows high correlations around all electrodes over the frontal lobes. Right: Same colormap but of 22Hz brainwaves.  Shows high correlation values only around electrode F3.

Figure 4- Left: Colormap of correlation values between left hand movement and rest conditions in 12Hz brainwaves. Note high, bilateral correlations in frontal cortex electrodes. Right: Same colormap with 22Hz brainwaves. Note relatively high correlations seen only unilaterally in electrode F3. The strength of correlation at 12Hz allows IpsiHand to distinguish left hand movement versus rest, and the correlation at 22Hz in electrode F3 allows for distinction of left versus right hand movement.


Combining neurophysiology, electronics, and rehabilitation, IpsiHand offers more effective rehabilitation for stroke and TBI survivors even in severe cases of paralysis. In testing, IpsiHand was able to process EEG signals for real-time hand control with accuracy consistent with previous studies [8]. Recent evidence suggests that combining BCIs and orthotic devices induces neural plasticity and improves motor function [8]. IpsiHand combines this advance in rehabilitation with the discovery of motor related signals in the undamaged brain hemisphere. Furthermore, the potential for recovery is unhampered by the severity of neural pathway injury since IpsiHand circumvents the entire injured pathway and uses the brain’s plasticity to generate new ones.

If produced in volume, we estimate a cost of around $674 per unit (Table 1), which even when marked up for retail would be significantly cheaper than alternative devices.

Table showing estimated price of IpsiHand system per unit. Includes price accuracy and bulk discount.

Table 1 - Prototype cost and estimated cost for production including bulk discount. Bulk discounts determined with manufacturer correspondence. *Headset donated by Emotiv Systems; this is the retail price of a headset.

Signal acquisition, signal processing, and mechanical control methods are established, but synthesizing them with the new technique of ipsilateral cortex recording is potentially new intellectual property. Based on discussions with therapists at the Rehabilitation Institute of St. Louis, IpsiHand is unique and an improvement over existing robotic assist and plasticity-facilitating devices.  Using muscle signals rather than neural signals, Myomo of Neuro-robotic Systems® uses facilitates plasticity less directly. The Bioness H200 from Ness®, an electrical muscle stimulator, has an unwieldy number of parts, is expensive (~$6200), and is a passive-assist-device not designed to promote active attempts at hand control. Hand Mentor from Kinetic Muscles Inc. does emphasize treatment through interactive movement, but is also expensive, non-portable, and not an option for those without residual motor control. IpsiHand fulfills these shortcomings by tapping into plasticity directly to create new neural pathways in portable package applicable even in severe cases [9]. Allowing patients to regain hand control with their thoughts will provide tremendous encouragement to continue with a therapy. Combined with IpsiHand’s affordability and minimal requirements for therapist supervision, IpsiHand also makes in-home treatment a very practical possibility.

Based on therapist discussions, we will make improvements upon the prototype. Currently, a laptop processes the EEG signals to be used for orthosis movement. We plan to use a miniature single-board Gumstix® computer to provide a portable data processing on a package smaller than a stick of gum. Mounting the micro-computer and a battery pack to the orthosis would give our system complete portability and allow patients to go beyond rehabilitation and use IpsiHand as a replacement of daily hand function.


We would like to especially thank Dr. Eric Leuthardt, our faculty mentor, and David Bundy, our graduate student mentor, for their guidance. We would also like to thank Joanne Rasch and the Rehabilitation Institute of St. Louis for providing consumer feedback and expert opinion on current rehabilitation, and Professors Robert Morley and Joseph Klaesner for instruction during senior design. This work is supported in part by The National Collegiate Inventors and Innovators Alliance, the Washington University School of Engineering, and Emotiv Systems.

First Author: Sam Fok:  13196 Strawberry Way, St. Louis MO 63146, sbf3@cec.wustl.edu


[1] HS-Jorgensen, et_al,-“Outcome_and_time_course_of-recovery_in_stroke: Part_II:_Timecourse_of_recovery. The_Copenhagen_Stroke_Study,”-Archives-of-Physical-Medicine-and-Rehabilitation,-1995.
[2] D_Lloyd-Jones, et_al., “Heart_disease_and_stroke_statistics–2010 update: a_report_from_the_American_Heart_Association_Statistics_Committee and_Stroke_Statistics_Subcommittee,” Circulation, 2009.
[3] National_Institute_of_Neurolgical_Disorders_and_Stroke, “Interagency_head_injury_task_force_report,” National_Institute_of_Neurolgical_Disorders_and_Stroke, 1989.
[4] CE_Levy, et_al, “Functional_MRI_evidence_of_cortical_reorganization_in_upper-limb_stroke_hemiplegia treated_with_constraint-induced_movement_therapy,” Am_J_Phys_Med_Rehabil, 2001.
[5] KJ_Wineski et_al., “Unique_cortical_physiology_associated_with_ipsilateral_hand_movements and_neuroprosthetic_implications,” Stroke, 2009.
[6] DJ_McFarland, et_al, “Spatial_filter_selection_for_EEG-based_communication,” Electroencephalography_and_Clinical_Neurophysiology, vol. 103, 1997.
[7] G_Schalk, et_al, “BCI2000:_a_general-purpose_brain-computer_interface_(BCI)_system,” IEEE_Transactions_on_Biomedical_Engineering, 2004.
[8] DJ_McFarland, et_al, “Electroencephalographic_(EEG)_control_of_three-dimensional_movement,” J._Neural_Eng., vol. 7, 2010.
[8] D_Broetz, et_al, “Combination_of_brain-computer_interface_training_and_goal_directed_physical_therapy in_chronic_stroke: A_case_report,” J_Neurorehab_and_Neural_Repair, 2010.
[9] HI_Krebs, et_al., “A_Paradigm-Shift:_Rehabilitation_Robotics,” IEEE_Engineering_in_Medicine_and_Biology, vol. 7, 2008.
[10] A_Lo, et_al. “Robot-assisted_therapy_for_long-term_upper_limb_impairment_after_stroke,” The_New_England_Journal_of_Medicine, 2010.


, , , , , ,

Powered by WordPress. Designed by WooThemes

Skip to toolbar