sw_detection, How to improve its accuracy?

[songhat]

When I use it, there will always be misidentification and missing some instance . How to improve it?

[raphaelvallat]

Hi @1358467720,

Are you using the default parameters? Have you tried changing the parameters of yasa.sw_detect? Did you apply any preprocessing to your data before running the SW detection?

As explained in the FAQ, I recommend manually annotating slow-waves or spindles events on a few representative recording of your own dataset, and then finding the parameters that give you the best agreement with your manual detection. This is the only way to get results that reflect your "own" scoring, especially for spindles detection.

Hope this helps,
Raphael