I mean compared to a doctor, are our models any good? Besides, how will our models help doctors exactly? will they speed up the diagnostic process for example? Or are they more like a double-checking standard?
", "messages": [ { "id": 1058432, "postDate": "2020-10-23T17:20:57.053Z", "content": "To me, triage and reducing time-to-treatment are the most straightforward ways automatic PE detection can help.
\nThe ED can get incredibly busy, and radiologists deal with large volumes of ED imaging. It's unlikely that a radiologist would miss a potentially life-threatening PE on a CT scan, but it is possible that it may take a few hours for the scan to actually be read.
\nEdit: ED = emergency department
", "rawMarkdown": "To me, triage and reducing time-to-treatment are the most straightforward ways automatic PE detection can help.\n\nThe ED can get incredibly busy, and radiologists deal with large volumes of ED imaging. It's unlikely that a radiologist would miss a potentially life-threatening PE on a CT scan, but it is possible that it may take a few hours for the scan to actually be read. \n\nEdit: ED = emergency department", "votes": 4, "replies": [ { "id": 1058440, "postDate": "2020-10-23T17:36:31.773Z", "content": "I'm sorry but I just googled \"ED\" and it showed Erectile Dysfunction, so I don't think you meant that?
\nbtw thank you so much for answering Dr. Ian.
", "rawMarkdown": "I'm sorry but I just googled \"ED\" and it showed Erectile Dysfunction, so I don't think you meant that?\n\nbtw thank you so much for answering Dr. Ian.", "votes": 2 }, { "id": 1058441, "postDate": "2020-10-23T17:37:59.380Z", "content": "Haha I meant emergency department. Edited my original post for clarity.
", "rawMarkdown": "Haha I meant emergency department. Edited my original post for clarity.", "votes": 3 }, { "id": 1059092, "postDate": "2020-10-24T15:57:19.933Z", "content": "@vaillant (Doctor Ian) Thanks a lot for your data set. I joined competition. late after osic,may be some more time could have helped me get better in this competition.
\n. Few things that i struggled so far is minimizing the loss, i verified many times to see if make any mistake in building loss in line to competition metric ,but loss looked always fine .
\nIt could mean either
\n1) Some CT slices are not so good ,making model fumble every time
\n2) Or loss has still got an issue
any advise if you feel like giving at this time where could the issue lie and hence could help get better.
\nof course computing resources is another bottleneck for many including me..
I mean compared to a doctor, are our models any good? Besides, how will our models help doctors exactly? will they speed up the diagnostic process for example? Or are they more like a double-checking standard?
", "rawMarkdown": "I mean compared to a doctor, are our models any good? Besides, how will our models help doctors exactly? will they speed up the diagnostic process for example? Or are they more like a double-checking standard?", "votes": 1 }, { "id": 1058427, "postDate": "2020-10-23T17:11:46.310Z", "content": "I think organizers want to get some insights about how they can improve disease detection, so they can later implement them in their models. It will certainly speed up diagnostic, as well as increase it's accuracy.
", "rawMarkdown": "I think organizers want to get some insights about how they can improve disease detection, so they can later implement them in their models. It will certainly speed up diagnostic, as well as increase it's accuracy." } ], "comments": [ { "id": 1058432, "author_name": "Ian Pan", "author_url": "", "post_date": "2020-10-23T17:20:57.053000", "content": "To me, triage and reducing time-to-treatment are the most straightforward ways automatic PE detection can help.
\nThe ED can get incredibly busy, and radiologists deal with large volumes of ED imaging. It's unlikely that a radiologist would miss a potentially life-threatening PE on a CT scan, but it is possible that it may take a few hours for the scan to actually be read.
\nEdit: ED = emergency department
", "votes": 4, "replies": [ { "id": 1058440, "author_name": "DarkCube", "author_url": "", "post_date": "2020-10-23T17:36:31.773000", "content": "I'm sorry but I just googled \"ED\" and it showed Erectile Dysfunction, so I don't think you meant that?
\nbtw thank you so much for answering Dr. Ian.
", "votes": 2, "replies": [] }, { "id": 1058441, "author_name": "Ian Pan", "author_url": "", "post_date": "2020-10-23T17:37:59.380000", "content": "Haha I meant emergency department. Edited my original post for clarity.
", "votes": 3, "replies": [] }, { "id": 1059092, "author_name": "Jaideep", "author_url": "", "post_date": "2020-10-24T15:57:19.933000", "content": "@vaillant (Doctor Ian) Thanks a lot for your data set. I joined competition. late after osic,may be some more time could have helped me get better in this competition.
\n. Few things that i struggled so far is minimizing the loss, i verified many times to see if make any mistake in building loss in line to competition metric ,but loss looked always fine .
\nIt could mean either
\n1) Some CT slices are not so good ,making model fumble every time
\n2) Or loss has still got an issue
any advise if you feel like giving at this time where could the issue lie and hence could help get better.
\nof course computing resources is another bottleneck for many including me..
I think organizers want to get some insights about how they can improve disease detection, so they can later implement them in their models. It will certainly speed up diagnostic, as well as increase it's accuracy.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1058432": "To me, triage and reducing time-to-treatment are the most straightforward ways automatic PE detection can help.\n\nThe ED can get incredibly busy, and radiologists deal with large volumes of ED imaging. It's unlikely that a radiologist would miss a potentially life-threatening PE on a CT scan, but it is possible that it may take a few hours for the scan to actually be read. \n\nEdit: ED = emergency department", "1058323": "I mean compared to a doctor, are our models any good? Besides, how will our models help doctors exactly? will they speed up the diagnostic process for example? Or are they more like a double-checking standard?", "1058427": "I think organizers want to get some insights about how they can improve disease detection, so they can later implement them in their models. It will certainly speed up diagnostic, as well as increase it's accuracy." } }