Since most top teams will be most likely using the tabular data to generate predictions, how will this aid in a real-world scenario for doctors who would most likely be using these?
", "messages": [ { "id": "1039233", "postDate": "10/06/2020 12:49:51", "content": "Since most top teams will be most likely using the tabular data to generate predictions, how will this aid in a real-world scenario for doctors who would most likely be using these?
", "rawMarkdown": "Since most top teams will be most likely using the tabular data to generate predictions, how will this aid in a real-world scenario for doctors who would most likely be using these?", "votes": null }, { "id": "1039367", "postDate": "10/06/2020 14:25:26", "content": "I would be really eager to know the winners' solutions. I would also love to know the hosts' views on utilizing these models in a real-life situation.
\nHow well would these models perform on real-life noisy data? Would there be a novel solution in predicting lung function decline after the end of the competition?
\nHopefully we get answers to these tomorrow! ;)
", "rawMarkdown": "I would be really eager to know the winners' solutions. I would also love to know the hosts' views on utilizing these models in a real-life situation.\n\nHow well would these models perform on real-life noisy data? Would there be a novel solution in predicting lung function decline after the end of the competition?\n\nHopefully we get answers to these tomorrow! ;)", "votes": null }, { "id": "1039372", "postDate": "10/06/2020 14:31:15", "content": "I don't think top teams will have only tabular data solutions.
", "rawMarkdown": "I don't think top teams will have only tabular data solutions.", "votes": null }, { "id": "1039380", "postDate": "10/06/2020 14:41:19", "content": "Given the way this competition is progressing and has progressed, it seems likely that at the end of shakeup top teams might have either a really novel way to use images or just pure tabular + image features (and if there is a very good solution with images taking a huge role, I believe it may be used in the real world scenarios if the hosts deem it so)
", "rawMarkdown": "Given the way this competition is progressing and has progressed, it seems likely that at the end of shakeup top teams might have either a really novel way to use images or just pure tabular + image features (and if there is a very good solution with images taking a huge role, I believe it may be used in the real world scenarios if the hosts deem it so)", "votes": null }, { "id": "1039393", "postDate": "10/06/2020 14:47:28", "content": "I'm also eager to know how the top-winning solutions work and if they make use of the images (and how, if so). However, I guess the winning model could be useful anyway, as it would be a scientifically proven method for estimating the progression with bounded uncertainty.
", "rawMarkdown": "I'm also eager to know how the top-winning solutions work and if they make use of the images (and how, if so). However, I guess the winning model could be useful anyway, as it would be a scientifically proven method for estimating the progression with bounded uncertainty.", "votes": null }, { "id": "1039412", "postDate": "10/06/2020 15:02:52", "content": "\nwe used several image derived features (with @gunesevitan, @authman and @sainatarajan7 ) already, and we did only scratch the tip of the iceberg what could be done… i will release a \"final insights\" soon, where i describe everything we used and segmented from the images and the ones, we had no time for anymore.
\nAbout the original question, in my opinion the models are going to used by some of the organizers (and other clinicians) on data that cannot be released. Many of us have hundreds when not thousands of high quality scans (along with clinical data) in house that were no consent given for such public analysis - i presume - but can be analysed in house retrospectively - so the research community will going to benefit from the results, that is sure. This alone is already benficial for the patients.
\nWe will see in the near future nice publications from these clinicians that use the models developed here. I can think on studies that question the role / modify the schedule of the follow up CT scans (why should we do if we can predict the decline based on the first one…) or compare the predicted course of the patient with the actual course after medication to validate the effect… but as a clinical decision tool won't be usable.
\nCurrently several clinical phase I–III trials are focusing on novel therapeutic agents for fibrosis, immunemodulatory agens are very promising, we hope for the best, that soon medicine is going to be developed that can completely stops the progression.
", "rawMarkdown": "nxrprime,\n\nwe used several image derived features (with @gunesevitan, @authman and @sainatarajan7 ) already, and we did only scratch the tip of the iceberg what could be done... i will release a \"final insights\" soon, where i describe everything we used and segmented from the images and the ones, we had no time for anymore.\n\nAbout the original question, in my opinion the models are going to used by some of the organizers (and other clinicians) on data that cannot be released. Many of us have hundreds when not thousands of high quality scans (along with clinical data) in house that were no consent given for such public analysis - i presume - but can be analysed in house retrospectively - so the research community will going to benefit from the results, that is sure. This alone is already benficial for the patients.\n\nWe will see in the near future nice publications from these clinicians that use the models developed here. I can think on studies that question the role / modify the schedule of the follow up CT scans (why should we do if we can predict the decline based on the first one...) or compare the predicted course of the patient with the actual course after medication to validate the effect... but as a clinical decision tool won't be usable.\n\nCurrently several clinical phase I–III trials are focusing on novel therapeutic agents for fibrosis, immunemodulatory agens are very promising, we hope for the best, that soon medicine is going to be developed that can completely stops the progression.", "votes": null } ], "comments": [ { "id": 1039367, "author_name": "aadhavvignesh", "author_url": "", "post_date": "10/06/2020 14:25:26", "content": "I would be really eager to know the winners' solutions. I would also love to know the hosts' views on utilizing these models in a real-life situation.
\nHow well would these models perform on real-life noisy data? Would there be a novel solution in predicting lung function decline after the end of the competition?
\nHopefully we get answers to these tomorrow! ;)
", "votes": null, "replies": [] }, { "id": 1039372, "author_name": "gunesevitan", "author_url": "", "post_date": "10/06/2020 14:31:15", "content": "I don't think top teams will have only tabular data solutions.
", "votes": null, "replies": [ { "id": 1039380, "author_name": "nxrprime", "author_url": "", "post_date": "10/06/2020 14:41:19", "content": "Given the way this competition is progressing and has progressed, it seems likely that at the end of shakeup top teams might have either a really novel way to use images or just pure tabular + image features (and if there is a very good solution with images taking a huge role, I believe it may be used in the real world scenarios if the hosts deem it so)
", "votes": null, "replies": [] }, { "id": 1039412, "author_name": "sandorkonya", "author_url": "", "post_date": "10/06/2020 15:02:52", "content": "\nwe used several image derived features (with @gunesevitan, @authman and @sainatarajan7 ) already, and we did only scratch the tip of the iceberg what could be done… i will release a \"final insights\" soon, where i describe everything we used and segmented from the images and the ones, we had no time for anymore.
\nAbout the original question, in my opinion the models are going to used by some of the organizers (and other clinicians) on data that cannot be released. Many of us have hundreds when not thousands of high quality scans (along with clinical data) in house that were no consent given for such public analysis - i presume - but can be analysed in house retrospectively - so the research community will going to benefit from the results, that is sure. This alone is already benficial for the patients.
\nWe will see in the near future nice publications from these clinicians that use the models developed here. I can think on studies that question the role / modify the schedule of the follow up CT scans (why should we do if we can predict the decline based on the first one…) or compare the predicted course of the patient with the actual course after medication to validate the effect… but as a clinical decision tool won't be usable.
\nCurrently several clinical phase I–III trials are focusing on novel therapeutic agents for fibrosis, immunemodulatory agens are very promising, we hope for the best, that soon medicine is going to be developed that can completely stops the progression.
", "votes": null, "replies": [] } ] }, { "id": 1039393, "author_name": "dcasbol", "author_url": "", "post_date": "10/06/2020 14:47:28", "content": "I'm also eager to know how the top-winning solutions work and if they make use of the images (and how, if so). However, I guess the winning model could be useful anyway, as it would be a scientifically proven method for estimating the progression with bounded uncertainty.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1039233": "Since most top teams will be most likely using the tabular data to generate predictions, how will this aid in a real-world scenario for doctors who would most likely be using these?", "1039367": "I would be really eager to know the winners' solutions. I would also love to know the hosts' views on utilizing these models in a real-life situation.\n\nHow well would these models perform on real-life noisy data? Would there be a novel solution in predicting lung function decline after the end of the competition?\n\nHopefully we get answers to these tomorrow! ;)", "1039372": "I don't think top teams will have only tabular data solutions.", "1039380": "Given the way this competition is progressing and has progressed, it seems likely that at the end of shakeup top teams might have either a really novel way to use images or just pure tabular + image features (and if there is a very good solution with images taking a huge role, I believe it may be used in the real world scenarios if the hosts deem it so)", "1039393": "I'm also eager to know how the top-winning solutions work and if they make use of the images (and how, if so). However, I guess the winning model could be useful anyway, as it would be a scientifically proven method for estimating the progression with bounded uncertainty.", "1039412": "nxrprime,\n\nwe used several image derived features (with @gunesevitan, @authman and @sainatarajan7 ) already, and we did only scratch the tip of the iceberg what could be done... i will release a \"final insights\" soon, where i describe everything we used and segmented from the images and the ones, we had no time for anymore.\n\nAbout the original question, in my opinion the models are going to used by some of the organizers (and other clinicians) on data that cannot be released. Many of us have hundreds when not thousands of high quality scans (along with clinical data) in house that were no consent given for such public analysis - i presume - but can be analysed in house retrospectively - so the research community will going to benefit from the results, that is sure. This alone is already benficial for the patients.\n\nWe will see in the near future nice publications from these clinicians that use the models developed here. I can think on studies that question the role / modify the schedule of the follow up CT scans (why should we do if we can predict the decline based on the first one...) or compare the predicted course of the patient with the actual course after medication to validate the effect... but as a clinical decision tool won't be usable.\n\nCurrently several clinical phase I–III trials are focusing on novel therapeutic agents for fibrosis, immunemodulatory agens are very promising, we hope for the best, that soon medicine is going to be developed that can completely stops the progression." }, "source": "meta" }