12k patients with multiple mammographs in train and 8k patients in test set. This looks really good! I expect stable training and stable CV/LB. This looks like one of the most promising medical computer vision competitions. Finally! :)
\nAnyone ever attended the RSNA conference?
", "messages": [ { "id": 2047658, "postDate": "2022-11-28T23:35:01.790Z", "content": "12k patients with multiple mammographs in train and 8k patients in test set. This looks really good! I expect stable training and stable CV/LB. This looks like one of the most promising medical computer vision competitions. Finally! :)
\nAnyone ever attended the RSNA conference?
", "rawMarkdown": "12k patients with multiple mammographs in train and 8k patients in test set. This looks really good! I expect stable training and stable CV/LB. This looks like one of the most promising medical computer vision competitions. Finally! :)\n\nAnyone ever attended the RSNA conference?", "votes": 35 }, { "id": 2047732, "postDate": "2022-11-29T02:24:18.937Z", "content": "I agree @crodoc, this comp looks promising, but I am afraid it might be much like last year's RSNA Brain Tumor comp .. where it's not possible to actually detect cancer from images alone. We've been using CAD to find abnormal densities in mammograms for 30 years but we still cannot definitively diagnose cancer without biopsy or lab tests. I am excited to see if this competition produces any clinically useful solutions.
\nI have gone to RSNA several times .. in fact, it's happening this week. I didn't go this year though.
", "rawMarkdown": "I agree @crodoc, this comp looks promising, but I am afraid it might be much like last year's RSNA Brain Tumor comp .. where it's not possible to actually detect cancer from images alone. We've been using CAD to find abnormal densities in mammograms for 30 years but we still cannot definitively diagnose cancer without biopsy or lab tests. I am excited to see if this competition produces any clinically useful solutions. \n\nI have gone to RSNA several times .. in fact, it's happening this week. I didn't go this year though.", "votes": 12, "replies": [ { "id": 2048042, "postDate": "2022-11-29T07:44:17.017Z", "content": "I really hope this is not the case. I've been waiting for a CV comp and I am exctied about this one.
", "rawMarkdown": "I really hope this is not the case. I've been waiting for a CV comp and I am exctied about this one." }, { "id": 2048046, "postDate": "2022-11-29T07:45:08.680Z", "content": "Plus it is a medical one ❤️
", "rawMarkdown": "Plus it is a medical one ❤️" } ] }, { "id": 2049188, "postDate": "2022-11-30T01:37:38.080Z", "content": "I dunno I'm having a bad feeling after seeing my baseline result. Even train pF1 can't get higher than 0. val_loss gets constant after some epochs. Imbalance and Small object size may be a deal-breaker in this competition.
I'm currently at RSNA, and have gone many times over the past decade, one of my favorite conferences to see all the technical exhibits and new tech showcased. I highly recommend checking it out if you can, especially if you're a rad.
", "rawMarkdown": "I'm currently at RSNA, and have gone many times over the past decade, one of my favorite conferences to see all the technical exhibits and new tech showcased. I highly recommend checking it out if you can, especially if you're a rad.", "votes": 2, "replies": [ { "id": 2053372, "postDate": "2022-12-03T07:33:40.410Z", "content": "I'd love to attend one of those :) although I am not a radiologist.
", "rawMarkdown": "I'd love to attend one of those :) although I am not a radiologist." } ] }, { "id": 2049340, "postDate": "2022-11-30T04:15:59.397Z", "content": "I checked the leaderboard and the F1 score is only 0.05, are there any signals from the data?
", "rawMarkdown": "I checked the leaderboard and the F1 score is only 0.05, are there any signals from the data?" }, { "id": 2047741, "postDate": "2022-11-29T02:44:23.523Z", "rawMarkdown": "", "votes": -24, "isDeleted": true, "replies": [ { "id": 2047751, "postDate": "2022-11-29T02:53:30.483Z", "content": "Please don't go there. Keep it professional okay?
", "rawMarkdown": "Please don't go there. Keep it professional okay?", "votes": 11 }, { "id": 2048519, "postDate": "2022-11-29T14:07:50.147Z", "content": "Please also keep in mind that this is about breast CANCER, a very serious issue as the 2nd leading cancer cause of death in women, and for many of us (including myself), something personal since we will very likely know someone who has/had it (1 in 8 women develop breast cancer).
", "rawMarkdown": "Please also keep in mind that this is about breast CANCER, a very serious issue as the 2nd leading cancer cause of death in women, and for many of us (including myself), something personal since we will very likely know someone who has/had it (1 in 8 women develop breast cancer).", "votes": 6 } ] } ], "comments": [ { "id": 2047732, "author_name": "David Roberts", "author_url": "", "post_date": "2022-11-29T02:24:18.937000", "content": "I agree @crodoc, this comp looks promising, but I am afraid it might be much like last year's RSNA Brain Tumor comp .. where it's not possible to actually detect cancer from images alone. We've been using CAD to find abnormal densities in mammograms for 30 years but we still cannot definitively diagnose cancer without biopsy or lab tests. I am excited to see if this competition produces any clinically useful solutions.
\nI have gone to RSNA several times .. in fact, it's happening this week. I didn't go this year though.
", "votes": 12, "replies": [ { "id": 2048042, "author_name": "CroDoc", "author_url": "", "post_date": "2022-11-29T07:44:17.017000", "content": "I really hope this is not the case. I've been waiting for a CV comp and I am exctied about this one.
", "votes": 0, "replies": [] }, { "id": 2048046, "author_name": "CroDoc", "author_url": "", "post_date": "2022-11-29T07:45:08.680000", "content": "Plus it is a medical one ❤️
", "votes": 0, "replies": [] } ] }, { "id": 2049188, "author_name": "Awsaf", "author_url": "", "post_date": "2022-11-30T01:37:38.080000", "content": "I dunno I'm having a bad feeling after seeing my baseline result. Even train pF1 can't get higher than 0. val_loss gets constant after some epochs. Imbalance and Small object size may be a deal-breaker in this competition.
I'm currently at RSNA, and have gone many times over the past decade, one of my favorite conferences to see all the technical exhibits and new tech showcased. I highly recommend checking it out if you can, especially if you're a rad.
", "votes": 2, "replies": [ { "id": 2053372, "author_name": "CroDoc", "author_url": "", "post_date": "2022-12-03T07:33:40.410000", "content": "I'd love to attend one of those :) although I am not a radiologist.
", "votes": 0, "replies": [] } ] }, { "id": 2049340, "author_name": "Phat Tran", "author_url": "", "post_date": "2022-11-30T04:15:59.397000", "content": "I checked the leaderboard and the F1 score is only 0.05, are there any signals from the data?
", "votes": 0, "replies": [] }, { "id": 2047741, "author_name": "", "author_url": "", "post_date": "2022-11-29T02:44:23.523000", "content": "", "votes": -24, "replies": [ { "id": 2047751, "author_name": "inversion", "author_url": "", "post_date": "2022-11-29T02:53:30.483000", "content": "Please don't go there. Keep it professional okay?
", "votes": 11, "replies": [] }, { "id": 2048519, "author_name": "Jesse", "author_url": "", "post_date": "2022-11-29T14:07:50.147000", "content": "Please also keep in mind that this is about breast CANCER, a very serious issue as the 2nd leading cancer cause of death in women, and for many of us (including myself), something personal since we will very likely know someone who has/had it (1 in 8 women develop breast cancer).
", "votes": 6, "replies": [] } ] } ], "raw_markdown_by_id": { "2047658": "12k patients with multiple mammographs in train and 8k patients in test set. This looks really good! I expect stable training and stable CV/LB. This looks like one of the most promising medical computer vision competitions. Finally! :)\n\nAnyone ever attended the RSNA conference?", "2047732": "I agree @crodoc, this comp looks promising, but I am afraid it might be much like last year's RSNA Brain Tumor comp .. where it's not possible to actually detect cancer from images alone. We've been using CAD to find abnormal densities in mammograms for 30 years but we still cannot definitively diagnose cancer without biopsy or lab tests. I am excited to see if this competition produces any clinically useful solutions. \n\nI have gone to RSNA several times .. in fact, it's happening this week. I didn't go this year though.", "2049188": "I dunno I'm having a bad feeling after seeing my baseline result. Even train `pF1` can't get higher than `0`. `val_loss` gets constant after some epochs. **Imbalance and Small object size may be a deal-breaker** in this competition.", "2048524": "I'm currently at RSNA, and have gone many times over the past decade, one of my favorite conferences to see all the technical exhibits and new tech showcased. I highly recommend checking it out if you can, especially if you're a rad.", "2049340": "I checked the leaderboard and the F1 score is only 0.05, are there any signals from the data?", "2047741": "" } }