EEG acquisition hardware per site — documented anywhere?

[indie Cha]

Hi all,

One of the central difficulties in this task is generalizing across the five contributing sites, and differences in EEG acquisition hardware are a well-known source of inter-site signal variation.

From what I can tell, demographics.csv and the EDF headers don't contain any device-level information, presumably because it was removed during de-identification.

Is the EEG recording hardware used at each site documented anywhere — for example, the EEG amplifier / acquisition system manufacturer and model, or even just which sites share the same equipment? Anything at a coarse level would help in reasoning about and correcting for device-driven domain shift.

If this information can't be released, even confirmation of whether meaningful site-level EEG hardware differences exist would be useful. Thanks!

[PhysioNet Challenge]

Dear Junhao,

As you may know, EEGs commonly use either a bipolar montage or a referential montage. The "monopolar" referential montages have a common reference electrode, and we are working to share more information about the devices so make these details clearer. To preserve the realism of the problem, we did not re-reference the data to a common montage across all sites, but the example code and channel_table.csv file should help you if you decide to do so for your approach.

For the unofficial phase of the Challenge, we artificially balanced the positive and negative cases in the public training set (but not in the hidden validation or test sets) to reduce file sizes. For the official phase of the Challenge, we did not balance the data, and the numbers of positive and negative cases in the training, validation, and test sets reflect the prevalence of cognitive impairment diagnoses over the given time window in each data source. As a result, we have both a small version and a large version of the training set, and the small version of the training set is prevalence matched to the large version of the training set. Teams may decide to download either or both versions of the training set, and they may request that we train their models on either version of the training set.

Best,
Matt
(On behalf of the Challenge team.)

Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address. This email is maintained by a group. Please do not email us individually.