24/7 measurement of tremor and bradykinesia
 

I've built a system using an Arduino microcontroller and an accelerometer that can be used, 24/7, to measure some of the key symptoms of PD. In itself, this approach is not new, such equipment already exists, and similar functionality can be got by using tablet apps. But my approach has a number of nice features:

- it can take readings as frequently as every 0.02 sec;
- it can collect and hold data 24/7 (the data is stored on a micro-SD card);
- it is small enough (68X45x18 mm) to wear on the arm;
- it can be built easily - no more than 20 solders;
- the components cost only about £25;
- once it is on, it does not require the engagement of the person;
- the software is in the public domain and is free;
- the box can be repurposed by changing the software;
- the technology can be easily extended to be used to proactively reduce some of the symptoms of PD, e.g. good posture.

To show what it can do, I attach two graphs showing the results from a 60 second run with samples being taken of the magnitude of the acceleration every 0.02s (i.e. 1/50 sec). The first graph shows the trace of a "tremor attack". The second graph shows what frequencies are important: the peak is just below 5hz, common for PD. (The graphs and analysis are done using the free statistical programming language r.)

Attachment 8642

Attachment 8643

The next thing to do is to validate a measure of bradykinesia based on the different traces between boxes at different parts of the body. It will vary from person to person, but consider three boxes placed near:
A: the worst tremor (in my case the left hand);
B: the equivalent lesser affected place (the right hand);
C: the most static position (the upper chest).
A-B gives an estimate of the tremor;
C-B gives an estimate of the non-arm swing.
C-B will vary over the time of the effectiveness of a dose. The gap between the highs and the lows provide an estimate of the bradykinesia.

If anyone wants to get into this area, please get in touch.

John

Just a quick note to say that I'm still looking at this, but progress is slow.

Here's a graph of my head movements over a 5hr period.


Attachment 9006


The head was chosen as the position of the accelerometer because I have no tremor there. This means that more movement is good. Whereas if the measurement is done on a place that tremors it confuses the issue.

The test was done using the Arduino/accelerometer system with my software as described in the previous post.

Time 0 is:
- 6 hours after my last medication;
- the time at which I take a new dose (Stalevo 75mg);
- 4hr 40min after eating.

I spent the 5hrs reading and dozing.

Each dot represents the average acceleration magnitude over one minute of readings taken every 0.1 second.

Although there's a lot of noise, I think the following features can be seen:
0-40 minutes: increasing acceleration (=symptom relief);
40-120 minutes: plateau;
120-180: decreasing acceleration;
180-280: there's an increase followed by a decline in movement that I can't explain.

John

That is fantastic! I love real data.

What device are you using for accelerometer?

The microprocessor is an Ardulog. The accelerometer is a MMA8452.

I'm looking for a proxy measure of bradykinesia in Parkinson's. This is made difficult because:
- more movement which is good where it is desired needs to be differentiated from tremor where it is not;
- strongly intentional movements (such as walking a mile) need to be differentiated from weakly intentional movements (such as changing your position on a chair).

To judge whether an approach is successful I expect to get a graph of movement against time following a levodopa dose that is similar to those you get from tap testing.

For this run I positioned the accelerometer on the back of my head, resting in a turned up ski hat.

Has anyone on the forum played with a smart phone app that stores accelerometer data and graphs it?

This may all seem to be arcane, but if we had better ways to measure PD, we could advance faster the search for better therapies.

John

For those of us in the USA, there is a MJFF study that will pair a Android Smart Phone and Pebble Watch. While I don't wear a watch anymore, I will be wearing one as part of this study!

Cool tear down on the watch is posted by eevblog Dave. Cool tech stuff.

One of the biggest issues when designing with accelerometers is drift. First part I played with drove me crazy as the data was great for rapid deceleration (air bag trigger) yet when tracking across time it seemed to really get odd data.

Sensor Fusion on Android Devices - TED talk 2010 on Youtube - give a good insight into the port processing math required when using an accelerometer.

The MM7150 Module has the Bosch Sensors on it, in addition to a processor to handle sensor fusion. This is what I was planning on using, but the pebble watch provides the same functions and is a cool watch.

http://www.microchip.com/_images/sen...odule-Lrge.jpg

Thank you for highlighting the MJFF initiative. I find it very exciting.

Theoretically you should be able to take (integrate) accelerations to get velocities and you should be able to take velocities to get positions. But what you find in practise is that small errors build up and soon you have meaningless results. This is the "drift" that TexasTom refers to. You can improve the situation by using better algorithms or by collecting sanity checking data from magnetometers or gyroscopes.

I spent a lot of time on this problem before realizing that in some situations, and I think we have one in measuring bradykinesia, it is unnecessary to derive the velocity or position from the accelerations. We can use as our proxy measure of PD the accelerations directly, and to make it simpler still, rather than work with the individual components of the acceleration vector, we can work with the magnitude of the accelerations.

There is great advantage using a smart phone app for data collection. However, it is important to remember that the real aim is therapy. So, as we go forward, I think that whatever approach to measurement is used it should be adaptable enough to form part of intelligent, personalized, dynamic therapies. As Marx wrote:

"The philosophers have only interpreted the world, in various ways. The point, however, is to change it."

John