Mark-Kramer / Spindle-Detector-Method

Spindle-Detector-Method

This repository contains the spindle detector method used in Kramer MA, Stoyell SM, Chinappen D, Ostrowski LM, Spencer ER, Morgan AK, Emerton BC, Jing J, Westover MB, Eden UT, Stickgold R, Manoach DS, Chu CJ Focal Sleep Spindle Deficits Reveal Focal Thalamocortical Dysfunction and Predict Cognitive Deficits in Sleep Activated Developmental Epilepsy. The Journal of Neuroscience 41, no. 8 (February 24, 2021): 1816–29.

Versions

• The branch master contains the most up-to-date, working version of the spinlde detector.

• For the original method applied in 2021 publication, see this branch j-neurosci-2021-publication-code.

Basic use

The latent state (LS) detector runs in two steps:

1. spindle_prob = LSM_spindle_probabilities(d, hdr);
2. spindle_det = LSM_spindle_detections(spindle_prob);

where

d = the data, [channels x time points]

hdr = structure that must have:

hdr.info.sfreq = sampling frequency (Hz).

hdr.info.ch_names = cell of channel names.

Step (1) is slow, step (2) is fast.

Advanced use

For narrowband analysis, include a third input to LSM_spindle_probabilities:

spindle_prob = LSM_spindle_probabilities(data, hdr, options)

where

options.StartFrequency = low frequency for narrowband analysis [Hz]. options.StopFrequency = high frequency for narrowband analysis [Hz].

Both options must be specified. For example, to focus analysis on 9-12 Hz,

options                = []
options.StartFrequency = 9;
options.StopFrequency  = 12;
spindle_prob = LSM_spindle_probabilities(data, hdr, options)

---

To visualize spindles

LSM_spindle_visualizer(data, hdr, spindle_det, channel)

---

When the code warns (Are your data in microvolts?)

If the code produces the warning:

Are your data in microvolts? If not, set options.MinPeakProminence

Then consider adding a third input to LSM_spindle_probabilities:

spindle_prob = LSM_spindle_probabilities(data, hdr, options)

where options.MinPeakProminence is a number indicating how much the peak must "stand out" to be identified. See here.

The code will run, but results might be meaningless.