Data folder

[Ján Pavlus]

Dear Physionet Challenge team,

what should I expect to be in data folder? Would there be signal, header and image? Because in the github repository preparing steps, the generated images are generated to the new folder and they are moved there only with the header files no signals.

Best,
Jan

[PhysioNet Challenge]

Dear Jan,

Good question. For the training set, you should expect a WFDB header file, a WFDB signal file, and one or more images for each record. For the validation and test sets, you should expect an incomplete WFDB header file (no labels, no initial values or checksums), no WFDB signal file, and one or more images for each record.

This tables summarizes what you should expect in each case:
https://moody-challenge.physionet.org/2024/#data-formats

The last step of the data preparation steps removes the signal files and parts of the header file, including the labels, because we're asking teams to digitize or classify the ECGs from the images, so their code should expect for the signal files and labels to be available during inference like they are in for training.

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.

[Ján Pavlus]

Dear Matt,

In the option with more than one image, should the signal be concatenated, or should we return a list of digitalized signals? How would more images be generated? Would the single be cut to the 2.5s sections, or will it be the start of every 10s part?

Best,

Jan

Dne čtvrtek 18. července 2024 v 19:04:51 UTC+2 uživatel PhysioNet Challenge napsal:

[PhysioNet Challenge]

Dear Jan,

If there is more than one image for a record, e.g., the record is longer than 10 seconds, then you should still return one digitized signal, i.e., one WFDB header file and WFDB signal file, for the record. You can try plotting a longer signal, e.g., a 30-second signal, to see what the ECG image generation code produces. We would compare your digitized signal with the original signal that you used to generate the ECG images.

In practice, I wouldn't worry about records with multiple images -- there is enough to do with records with a single image!

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.