Patient Sensor Evaluation - Humanos3D/ElMedalloBionicArm GitHub Wiki

Motivation

This document describes an initial evaluation of patient muscle funtion. The goals of the procedure include verifying that we have usable muscle signals for EMG triggering, finding the best location for the EMG sensor(s), and explaining the technology to the beneficiary. Note that it seems unavoidable that sensor placement will depend on the condition of the remaining limb, and so a significant level of patient-to-patient customization will be required. After performing this protocol, we should have a good idea of what the customization should look like.

General EMG and Patient Prep

  1. Do an equipment check beforehand. And make sure we can record raw data adequately before starting the session.

  2. Explain what we are doing to the recipient (some of this can be staggered with testing).

    • Introductions.
    • Establish parameters of the visit. "How long do you have available to visit today?" "Is it OK to take photos?"
    • Outline the general idea of EMG sensors.
    • Demonstate a prototype system on the table top. Have someone from e-NABLE perform the demo to start, and perhaps finish the session with the beneficiary doing it.
    • Describe the remaining timeline (e.g., this likely won't be ready in the next few months).
    • Describe how we'll use EMG stickers to perform a best-case evaluation.
    • And then describe how we'll move on to more representative sensors next

  3. Check whether they have any questions.

  4. Make sure all areas are clean and potentially shaved.

Baseline EMG Evaluation

We will need to understand how much muscle is left in the limb and to what extent it can be controlled. I.e., this is a little different from determining how our sensors might measure the signal. As a first step, we use the FlexVolt as it measures signals using medical grade stick-on sensors that can be arbitrarily placed.

  1. Test on the amputated limb using the FlexVolt (see here for instructions). Find a reasonable reference point (ideally directly over bone) and then look at various combinations of signal electrode placements. We will want to clearly label electrodes and take many photographs. Simultaneous multi-channel testing will determine whether locations are correlated (i.e., to do multiple grips, we'll need distinct muscle contractions). The FlexVolt Chrome app is good for both visualizing signals and recording data. A basic tool for reading FlexVolt files into Python is here

  2. If signal levels aren't good, it may be worth testing on the intact limb (assuming it exists). Place reference and various sticker sensors on equivalent areas on the intact limb. Record the signals using the FlexVolt and see if they're any better.

  3. It may be worth doing something similar with an “always available” subject from e-NABLE. This will allow us to make some inferences based on a staff member when the actual recipient isn’t available.

Representative EMG Evaluation

In the most promising areas evaluate the signal from our candidate dry-electrode EMG sensors. Again, labeling, photos and note taking are key. Currently our options are the Protesis Avanzadas sensor and the OYMotion sensor - these are both being run via Arduino. We have a basic algorithm in place to run these - the sensor type is changed with just a flag at the top of the file. Initial visualization can be done through the Arduino Serial Plotter, but for recording I suggest CoolTerm. CoolTerm is a cross-platform serial interface and the data stream can be easily saved to file. Note that the Serial Plotter and CoolTerm can not be run at the same time though - they both claim the serial data stream. We do not currently have a dedicated CoolTerm reader written (e.g., like the Python FlexVolt reader) but the output is just a space delimited text file, which can be read by many pieces of software (e.g., Excel).

Evaluation for Other Sensors

This is still largely TBD, but there is a possibility we get recipients with no usable muscle function. We may want to consider alternative triggering approaches. Switches may be preferable, or a good fallback, for some patients. Check whether they can move any muscles, in convenient locations, such that they could actuate a switch or touch pad.