First of all, thank you very much for your work, such a great project!
I am reading the annotations coming from the .csv files of the STAGES dataset: I realized the size of the signals (flow, ekg...) are not the same than the annotations file. For example, in the BOGN00003 exam, the signals have 640.98 minutes and the annotations have 625.5 minutes... Do you have an explanation?
Many thanks for your help!
Thanks for raising this question. We're happy you've found the NSRR useful.
My suspicion is that the BOGN site's sleep system "clips" the end of the annotations at the time of export. I opened up ~10 EDF/CSV pairs and found the following:
I will run this by a couple other folks on the NSRR team to see if they have any other suggestions. I may also reach out to the STAGES data submitters to see if they know more about the BOGN annotation export procedures.
If you find any recording pairs that don't meet the criteria I mentioned above, please bring those to our attention!
I ran this thread and my response by the STAGES team - they agreed with my response. I see that you have access to some other dataset (e.g., SHHS, MrOS) that were scored in Compumedics Profusion, which exports 30-second staging for the entire recording duration. Not all PSG collection/scoring systems will handle annotations in the same manner, as we are seeing here between the STAGES sites, which used a handful of different PSG systems.
Thank you for your prompt answer and for the tips !
I was assuming that the starts should coincide but I wanted to make sure before working on it for days... I will keep you update if I find some recordings not aligned.