5

I have raw data obtained from EMED barefoot scan containing a matrix of pressure sensors over about 70 frames. This totals 70 matrices that record a snapshot of the pressure over the duration of a person's natural walk over the pressure plate.

I wanted to ask if anyone knows the algorithm that is typically used to determine the Center of Pressure Line (indicated by the path on the heatmap below). One thing I have tried is to consider each row as an array and find the index of the maximum pressure across the row, and create a line by matching up all the points that correspond to these indices. My approach fails to produce a smoothed line (even when using Gaussian smoothing).

center of pressure line across foot heatmap

blaklaybul
  • 151
  • 4
  • What do you mean by "this path"? That is the first time that you mention it. – hazzey Oct 13 '15 at 16:13
  • @hazzey edited. I was referring to the path shown on the figure below – blaklaybul Oct 13 '15 at 16:15
  • I have no real knowledge in this matter so I won't post this as an answer, but I would get each row and get the center of gravity of that row (the weighed average of the horizontal position of each pixel and the pressure applied). This will not result in exactly the image you have, though, since you can see that in the arch of the foot the line is not within the pressure zone of the foot. This means the real method is non-local, taking into consideration more information than each individual row. – Wasabi Oct 13 '15 at 17:25
  • @Wasabi yea, it appears that the real method leverages some global knowledge of the horizontal pressure distribution. – blaklaybul Oct 13 '15 at 18:14
  • 2
    I'm no expert in the field, but I think that the image that you posted is misleading (and is perhaps the source of your confusion). I believe that the COP line is calculated for many different frames and then overlayed onto a single frame. If I am correct, then the COP of a single frame is simply the center of mass of the pressure image. The COP position moves forward with each frame as the person progresses through their gait, and these positions are overlayed on a frame near the center of the gait. – Chris Mueller Oct 13 '15 at 23:28
  • 1
    @ChrisMueller that's it. this view is actually a "peak pressure" view which takes the max for each sensor over the duration of the gait. Taking the center of mass for each frame produces the desired result. – blaklaybul Oct 14 '15 at 00:19

3 Answers3

0

Actually, scipy has a function that does exactly this. I used it on each of the 70 arrays that make up the time series, and points generate a smooth line.

https://docs.scipy.org/doc/scipy-0.16.1/reference/generated/scipy.ndimage.measurements.center_of_mass.html

blaklaybul
  • 151
  • 4
0

The way you mathematically describe your data should depend on how you plan to use the data. Just any smooth equation may not represent something meaningful. I am not familiar with practices in biomechanics but center of pressure is typically understood as a point, not a line.

That said, the simplest way I can see to process this data would be to rotate the plot 90 degrees (switch x and y). Then separate your data into the 6 levels or however many you have. The first level gets one data point, second level gets two data points and so on (assuming your scale is linear). Then you can model it with standard polynomial regression. This is kind of a crude way to go about it; depending on your implementation you can likely modify the standard polynomial regression to account for the "weight" of a data point instead making multiple points form single points.

grfrazee
  • 3,587
  • 1
  • 16
  • 33
ericnutsch
  • 7,689
  • 1
  • 9
  • 28
  • 1
    I'm not entirely clear on your response. My question was probing what exactly the biomechanical conventions are for calculating this center of pressure line. This is a standard metric that researchers, particularly in sports and diabetic research. – blaklaybul Oct 13 '15 at 16:36
  • I did some searching and I belive the vocabulary you are looking for is "CoPP (Centre of Pressure Progression)" Just skimming some texts, I belive this is equivalent to "center of pressure linear regression" http://www.podiatry-arena.com/podiatry-forum/showthread.php?t=77774 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3082780/ http://ir.lib.cyut.edu.tw:8080/bitstream/310901800/7343/1/B59.pdf http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4101275/ – ericnutsch Oct 20 '15 at 02:14
0

You might try looking at the k-means function in opencv for Python or C. Additionally, the k-means function in Matlab may be useful as well. To further clarify, I believe OP would describe the center line as a neutral axis nearly parallel to the sagittal plane.

piman
  • 41
  • 4