dwongh/teetacsi

Code for acquiring and analysing data for the TEETACSI project (Tracking Expert Eyes to Train AI for Clinical Signal Interpretation)

teetacsi's Issues

Tracking Signals

Currently, signal Y-Axis data is tracked using the channel baseline coordinates. This might result in poorer accuracy if the signals are particularly skewed from the baselines, or if the user has zoomed in close.

The --signals option enables signal reprojection by re-reading the edf file and using the tracking data to reconstruct the position of the signals. This would result in more accurate tracking - however, it is not stable for several reasons.

First, when zooming out past the whole recording (so there is blank space either side of the eeg signals), the reprojected signals appear skewed to the left - a problem with filling the signal buffers. This should be fixed by rolling the buffer to the right in the update_channel_data function.

Second, the edfreader module cannot load less than 1 sample, which becomes a problem when zooming in below 1msec/page. Will need to find a way to handle this, possibly by just using the baselines instead.

Thirdly, when calculating the signal image coordinates, the coordinates appear to invert either side of 100 volts per cm. I.e., when zooming out the signals get bigger instead of getting smaller and vice versa. Some other scale information would likely need to be considered in the calculation in draw_signal.

Finally, to improve accuracy the tracking would need to handle the application of different filters when reprojecting and calculating the position of the signals