Patent Landscape of Brain-Machine Interface Technology

Authors: Anastasia Greenberg, Alexis Cohen, and Monica Grewal

Like Comment
Read the paper

In 2006, a patient with tetraplegia was implanted with a microelectrode array in a part of the brain known as the primary motor cortex.[i]  The patient was trained to imagine a series of hand movements while a cursor on a computer screen was moved in coordination with the patient’s imagined movements. This allowed a brain-machine interface (BMI) system to be calibrated to the patient’s imagined movements.1 Eventually, the patient was able to use the BMI system to control a prosthetic hand, open simulated email, play video games, and control lights – all through imagined movements.1

BMI technology combines brain activity with computational processing.  A BMI is an input-output computing system: the input is brain activity, and the output is the manipulation of a device. BMIs allow for thought-controlled manipulation of devices such as computers, robots, prosthetic limbs, and wheelchairs.[ii],[iii]

In our recent paper published in Nature Biotechnology, we analyzed the intellectual property landscape, specifically patent applications/publications, from 1980 to 2020 in the BMI space. Our results show an exponential trend with growth from one patent publication in 1984 to 510 publications in 2020.  The majority of patents were filed in the United States (~47%), with China representing the second most common jurisdiction (~21%). Out of the top 20 assignees of BMI patent applications, about half were universities, suggesting that a large proportion of BMI technology is in early stages of development or commercialization. The top 20 assignees made up about a fifth of all filed patents, suggesting that development and/or patent protection in the BMI space is relatively spread out across many entities rather than concentrated among a few.

BMI Patent Publications over time

 The content of BMI patent publications included technological advancement in four main BMI process steps: 1) brain signal recording/acquisition, 2) brain signal feature extraction, 3) brain signal decoding and translation into a command for an output device, and 4) user feedback.

The first stage of implementing a BMI involves acquiring/recording brain activity.  Neurons communicate with one another via electrochemical signaling and this activity can be recorded.  Devices and methods used for recording brain activity fall into two main categories: invasive and non-invasive.[iv]  Invasive BMIs involve surgical implantation of recording devices (e.g., electrodes/electrode arrays) that are placed either on the surface of the cortex or penetrate deeper into brain tissue.4 Non-invasive BMIs record activity from the surface of the scalp or via sensors placed near the head and do not require any surgery.4 We found that patent publications were about 10 times more likely to be directed to non-invasive BMIs, particularly BMIs employing electroencephalography (EEG), than invasive BMIs.

The second step in a BMI system is extracting features from the acquired brain signals, which, when properly translated, can be used to drive specific BMI actions.  Among the features that we identified in the BMI patent landscape is the event-related desynchronization (ERD) which is a type of movement-related activity pattern that appears during movement, preparation for movement, and motor imagery.  Another feature that we identified in the patent landscape is generally categorized as visual-evoked potentials, which reflect specific brain signals detected in association with visual events. For example, the P300 evoked potential is a type of visual-evoked potential that can be recorded from EEG electrodes placed over the centro-parietal scalp areas after a person detects an unexpected or desired visual stimulus.[v]  The P300 has been used in BMI systems called “P300 spellers” allowing individuals to communicate without producing any movement or speech by focusing their attention on letters on a screen one-by-one to create words and sentences.5 Another, less common, feature that we identified in the BMI patent landscape is the error-related negative evoked potential (ERN) which can be detected over fronto-central scalp areas after an individual detects a mismatch between his/her intended action and the output response of a device.  The ERN could be used as a calibration and correction mechanism to precisely tailor BMI systems to an individual user.

Once brain activity is recorded and relevant features are extracted, the next step in a BMI system is to automatically interpret, or decode, the brain activity so that it can be translated into a device command.[vi]  The BMI patent landscape covers various methods of decoding brain activity including using a variety of machine learning algorithms to classify brain activity patterns into distinct groups which can then be mapped onto a specific device command.  

In some BMI technologies, after brain signals are recorded, decoded, and translated into device output, the final processing step is providing feedback to the user to allow the user to adjust their BMI control in response to the feedback.  Feedback can be of different forms such as visual feedback (e.g., an individual sees the action taken by a robot), haptic feedback (e.g., a usre feels vibration of an exoskeleton), and biofeedback (e.g., an individual is presented with information about their brain signals).  We found that about half of all BMI patent publications referred to at least one form of feedback.

BMIs have many possible applications including control of devices for therapeutic purposes, neurorehabilitation for direct repair of neurological damage, as well as for purely entertainment-based purposes such as video gaming.  We found that the BMI patent landscape is heavily focused on clinical or therapeutic applications rather than entertainment-based objectives.  This work demonstrates that BMI technology is at the cusp of commercial realization presenting promise for revolutionizing personalized therapeutic options for patients with physical constraints.

References

[i] Hochberg, Leigh R., et al. “Neuronal ensemble control of prosthetic devices by a human with tetraplegia.” Nature 442.7099 (2006): 164-171.

[ii] Prashant, Parmar, Anand Joshi, and Vaibhav Gandhi. “Brain computer interface: A review.” 2015 5th Nirma University International Conference on Engineering (NUiCONE). IEEE, 2015.

[iii] Abiri, Reza, et al. “A comprehensive review of EEG-based brain–computer interface paradigms.” Journal of neural engineering 16.1 (2019): 011001.

[iv] Rao, Rajesh PN. Brain-computer interfacing: an introduction. Cambridge University Press, 2013.

[v] Chaudhary, Ujwal, Niels Birbaumer, and Ander Ramos-Murguialday. “Brain–computer interfaces for communication and rehabilitation.” Nature Reviews Neurology 12.9 (2016): 513-525.

[vi] Nicolas-Alonso, Luis Fernando, and Jaime Gomez-Gil. “Brain computer interfaces, a review.” sensors 12.2 (2012): 1211-1279.

Anastasia Greenberg

IP Attorney, WilmerHale