{ "id": 9636, "title": "latency number question", "url": "/competitions/seizure-detection/discussion/9636", "author_name": "", "post_date": "2014-07-02T12:21:43.753Z", "votes": null, "comment_count": 5, "views": 1613, "content": "

Need a clarification on the latency number provided in the ictal files. Sequentially the latency number for patients start at zero and increase in value, At some point the latency number becomes zero again (referred to as a "zero event" below) and continues to increase in value. Is this interpreted as the seizure 'starts' at a "zero event" -- to include the first segment if its' latency is zero? If so, the evaluation for within 15 seconds is successful by predicting the 15 segments leading up to a zero event? Or the 15 segments after a zero event? Or both sides of a zero event?

\n\n

Thank You

", "messages": [ { "id": "49955", "postDate": "07/02/2014 12:21:43", "content": "

Need a clarification on the latency number provided in the ictal files. Sequentially the latency number for patients start at zero and increase in value, At some point the latency number becomes zero again (referred to as a "zero event" below) and continues to increase in value. Is this interpreted as the seizure 'starts' at a "zero event" -- to include the first segment if its' latency is zero? If so, the evaluation for within 15 seconds is successful by predicting the 15 segments leading up to a zero event? Or the 15 segments after a zero event? Or both sides of a zero event?

\n\n

Thank You

", "rawMarkdown": "", "votes": null }, { "id": "49963", "postDate": "07/02/2014 16:27:43", "content": "

The goal is to identify the first 15 seconds of the seizure, i.e. clips with latency 0-14.

", "rawMarkdown": "", "votes": null }, { "id": "50001", "postDate": "07/03/2014 16:49:36", "content": "

Sorry just to re-confirm, this thread stated less-than-equal 15s which I accidentally took to mean latency less-than-equal 15. So it's actually latency less-than 15 aka 0-14 as you have just mentioned? I also re-submitted my results trained on 0-14 and got a worse score so I'm a bit unsure now.

\n

Thanks!

\n

edit: sorry not sure what's happening with my post formatting, I tried to use angle brackets for less-than/equals and everything breaks

", "rawMarkdown": "", "votes": null }, { "id": "51427", "postDate": "07/31/2014 17:27:35", "content": "

As Michael Hills points out, there is a contradiction between this thread and the other.  A response is needed. Can someone please clarify whether "early" corresponds to latency variable in the clips being strictly less-than or less-than-or-equal to 15?

", "rawMarkdown": "", "votes": null }, { "id": "51482", "postDate": "08/01/2014 14:19:00", "content": "

I apologize.... my post above is incorrect. The criteria is latency <=15. The latency field refers to the time elapsed between the epileptologist's seizure annotation and the first sample in the clip, hence the zero-based indexing.

\n

Thanks very much for the questions, and I apologize for the confusion on this.

", "rawMarkdown": "", "votes": null }, { "id": "51529", "postDate": "08/02/2014 19:34:23", "content": "

Hi Bbrink:

\n

I have a question related to the prediction ictal segments within 15 sec..I guess having 2 classifiers i.e(one for ictal vs interictal and the second classifer for prediciting less than 15 sec)  per subject is fine?

\n

Looking forward to your reply

\n

Thanks

\n

Navin

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 49963, "author_name": "bbrinkm", "author_url": "", "post_date": "07/02/2014 16:27:43", "content": "

The goal is to identify the first 15 seconds of the seizure, i.e. clips with latency 0-14.

", "votes": null, "replies": [] }, { "id": 50001, "author_name": "michaelhills", "author_url": "", "post_date": "07/03/2014 16:49:36", "content": "

Sorry just to re-confirm, this thread stated less-than-equal 15s which I accidentally took to mean latency less-than-equal 15. So it's actually latency less-than 15 aka 0-14 as you have just mentioned? I also re-submitted my results trained on 0-14 and got a worse score so I'm a bit unsure now.

\n

Thanks!

\n

edit: sorry not sure what's happening with my post formatting, I tried to use angle brackets for less-than/equals and everything breaks

", "votes": null, "replies": [] }, { "id": 51427, "author_name": "quarkmountain", "author_url": "", "post_date": "07/31/2014 17:27:35", "content": "

As Michael Hills points out, there is a contradiction between this thread and the other.  A response is needed. Can someone please clarify whether "early" corresponds to latency variable in the clips being strictly less-than or less-than-or-equal to 15?

", "votes": null, "replies": [] }, { "id": 51482, "author_name": "bbrinkm", "author_url": "", "post_date": "08/01/2014 14:19:00", "content": "

I apologize.... my post above is incorrect. The criteria is latency <=15. The latency field refers to the time elapsed between the epileptologist's seizure annotation and the first sample in the clip, hence the zero-based indexing.

\n

Thanks very much for the questions, and I apologize for the confusion on this.

", "votes": null, "replies": [] }, { "id": 51529, "author_name": "navincota", "author_url": "", "post_date": "08/02/2014 19:34:23", "content": "

Hi Bbrink:

\n

I have a question related to the prediction ictal segments within 15 sec..I guess having 2 classifiers i.e(one for ictal vs interictal and the second classifer for prediciting less than 15 sec)  per subject is fine?

\n

Looking forward to your reply

\n

Thanks

\n

Navin

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "49955": "", "49963": "", "50001": "", "51427": "", "51482": "", "51529": "" }, "source": "meta" }