Non-Invasive Motor Imagery Decoding for Assistive Control in Spinal Cord Injury

Definition Statement

There are 15.4 million people with spinal cord injury, which disrupts the connection between their brain and muscles. Current technology does not reliably restore intuitive voluntary movement. A system that would be able to understand what the user wants to do and stimulate their muscles would restore voluntary movement.

I am solving this problem for the 15.4 million people who deal with spinal cord injury today. There are two main types of SCI: Tetraplegia which is paralysis of all 4 limbs and some of the torso, and paraplegia where the legs cannot move but the arms still work. I will. specifically focus on helping people with Cervical SCI to help restore upper-limb movement.

Possible Solutions

1. Noninvasive BCIs connected to electrodes to externally stimulate muscle contraction
   • This idea is the best for my current position. I do not have access to a lab to implant electrodes and BCIs. Noninvasive BCIs are not as powerful because of the noise that they pick up alongside the brain activity. However, using AI I will be able to remove some noise to detect brain intent.
2. Invasive BCIs connected to electrodes to stimulate muscle contraction
   • Although this idea would be much more powerful than the previous, I do not have access to a lab where I could implant an invasive BCI, nor do I have the license to. 
3. Electrodes on the spine to restore connection between brain and muscles
   • This idea would also require a lab to insert the electrodes on the spine, which is not feasible in my position. However, this idea could potentially be more powerful than idea #2 because instead of getting signals directly from the brain and externally stimulating the muscles, this would theoretically restore the original movement because it is the link between the brain and the muscles. 
4. Implant electrodes inside muscle tissue to make precise movement connected with BCIs
   • This is not so much of a standalone idea as something to add on to another idea because implanting electrodes in the muscles to get more precise movement would be much better when retraining motor control to help restore fine movements.
5. BCI controlled exoskeleton
   • A BCI controlled exoskeleton although it would be simpler than stimulating the muscles, it is bulky and the goal of the BCI solution is to make something discreet to make the people who suffer from SCIs feel more normal.

Possible Prototype for #1

The prototype would be a non-invasive BCI to get signals from the brain. Because the brain still sends the same signals to the body even when they can’t move the body parts, those signals can be read to send signals to different electrodes placed on the body. When the electrodes send a small electrical pulse, the muscle contracts.

What I hope to have ready in May at the minimum is a noninvasive BCI that I can read the signals from to determine some sort of action the user would like to do. Depending on how much progress I make I could either show the results on the computer and make the user do something such as play a simple game like flappy bird/dinosaur jump,  or I could connect it to electrodes on my muscles to stimulate them, if I am allowed.

Plan for this Project

The first step of this project is understanding software. I will use an EEG dataset to make a classification algorithm to classify motor intent. Then I will simulate it to see if it works and iterate. 

Once I have a working algorithm, I will purchase a noninvasive EEG and use the algorithm to decode the signals from my brain. I will attempt decoding starting at simple left and right imagery, then go into more complex commands and see how good my algorithm is. I will iterate with the algorithm until it is strong enough to use.

After this I will purchase an electrode to attempt stimulating my muscle. I will need to be very careful at this stage because doing it wrong could cause harm to the wearer, which would be me. I may decide not to use an electrode but instead use the signal to send signal to a robot arm, which would move based on the commands.