Good question - unfortunately the Profusion scoring software did not make this straightforward because it does not always output metadata about each respiratory event, such as "associated with a desaturation of X%" and/or "associated with an arousal". It would be up to you (the user) to try and piece this together.
Thanks for using the site.
Could you give me some advice about how to 'piece this together'?
The ahi_a0h3 definition is: (All apneas + hypopneas with >= 3% oxygen desaturation) / hour of sleep
I tried to screen events provided in “annotations-events-nsrr”:
Step1. delete the events in Wake status.
Step2. for every hypopnea event, detect the spo2 desaturation 30s later. If the desaturation < 3%, delete the event.
However, the ahi calculated in this way are still different from ahi_a0h3, and most ahi are smaller than ahi_a0h3.
How can I solve this problem? Thank you!
The "hard" part, I think, would be reverse engineering the linking of desaturations with hypopneas. I'm not sure what sort of time windows before/after the respiratory event that the scoring software used. You might play around with your detection.
You could add up the total number of events of each type and then compare them with the numerator of ahi_a0h3 - https://sleepdata.org/datasets/shhs/variables/ahi_a0h3
For instance, carbp + carop + canbp + canop should be the total number of central apneas. There is another set of variables for obstructive apneas, and another for hypopneas with >=3% desaturation.
Thank you for your advice!
I check the relationship between the number of events of each type and variables provided in CSV files, and I find that the ahi calculated by counting the events of each type is different from the variable: ahi_a0h3 and the difference is caused by the number of OSA events. Is there any error in the variables?
I also check the number of events calculated from annotations, the number of CSA events is larger than (carbp + carop + canbp + canop), and the number of OSA events differs from (oarbp + oarop + oanbp + oanop)(larger or smaller).
I'm very confused about how to use the dataset. Can you give me some advice about how to use it correctlly?
I'm very confused about how to use the dataset. Can you give me some advice about how to use it correctlly?
Note the "Known Issues" for SHHS, if you haven't already: https://sleepdata.org/datasets/shhs/pages/05-polysomnography-introduction.md
In addition to linking desaturations there may be other quirks with events that either 1) start in sleep and end in wake, or 2) start in wake and end in sleep.
You won't find as many of these discrepancies in newer datasets.