Closed informaton closed 4 years ago
This is addressed in the beta branch (127683a031258d425ed9de3153949d9475f134a6)
The current solution replaces nan values with zero values prior to hypnodensity prediction and then retroactively updating the offending sections of the hypnodensity with nan values so that they are marked 'unknown' (7) instead of 'wake' (0) when producing the hypnogram staging file.
TODO: pull beta branch into master.
This was merged with pull #26 (commit 68d2bd1).
Flatline data, when excessive, produces a non-numeric result (nan) that, when fed into the hypnodensity models, will continue to produce nan results and prevent hypnodensity scoring for the remainder of the study. This impacts narcolepsy inferencing as well.