lkorczowski / Tinnitus-n-Sleep

Detecting events in sleeping tinnitus patients
MIT License
1 stars 0 forks source link

Artifact Detection uses too strict criteria #90

Closed lkorczowski closed 4 years ago

lkorczowski commented 4 years ago

observation

too much brux artefacts makes bruxism artifact very badly detected (see figure below)

log report show that is_good is the reponsible, here more than 6% of epochs are removed for HZB_nuit_2:

                          bruxism           OMA       IMP
suppressed_is_good    6627.000000      0.000000       0.0
suppressed_amp_thr       0.000000    751.000000       0.0
suppressed_overall    6627.000000    751.000000       0.0
total_nb_epochs     104500.000000  26125.000000  104500.0
suppressed_ratio         0.063416      0.028746       0.0

solution

more permissive threshold (e.g. emg=1e-04 or emg=5e-04 instead of emg=5e-05)

read new commits on PR #85 for changes and rerun script compute_results


![Screenshot 2020-05-27 at 16 41 56](https://user-images.githubusercontent.com/15196605/83034867-0541cd80-a039-11ea-9de8-078624a2fef1.png)