Hi guys, I just start to join this competition and I found paper which is 2nd place solution in CheXpert competition.
\nhttps://arxiv.org/pdf/1911.06475
\nThis paper is written by Vingroup Big Data Institute(Competition Host)
\nIt's seems like that there is hierarchy of Classes(ex Lung Opacity --> Atelectasis) And They also tried to train model depend on hierarchy of diseases and ensemble.
\nSo it's might helpful consider Hierarchy of Diseases when you are doing pre/post processing.
\nHope to be helpful this information. Thanks
", "messages": [ { "id": 1193271, "postDate": "2021-02-09T14:45:12.717Z", "content": "Hi guys, I just start to join this competition and I found paper which is 2nd place solution in CheXpert competition.
\nhttps://arxiv.org/pdf/1911.06475
\nThis paper is written by Vingroup Big Data Institute(Competition Host)
\nIt's seems like that there is hierarchy of Classes(ex Lung Opacity --> Atelectasis) And They also tried to train model depend on hierarchy of diseases and ensemble.
\nSo it's might helpful consider Hierarchy of Diseases when you are doing pre/post processing.
\nHope to be helpful this information. Thanks
", "rawMarkdown": "Hi guys, I just start to join this competition and I found paper which is 2nd place solution in [CheXpert](https://stanfordmlgroup.github.io/competitions/chexpert/) competition.\n\n[https://arxiv.org/pdf/1911.06475](https://arxiv.org/pdf/1911.06475)\n\nThis paper is written by Vingroup Big Data Institute(Competition Host)\n\nIt's seems like that there is hierarchy of Classes(ex Lung Opacity --> Atelectasis) And They also tried to train model depend on hierarchy of diseases and ensemble.\n\n![](https://drive.google.com/uc?export=view&id=1mN4SNZrToK62H6vgkkKM-dIdOPaxepLj\")\n\nSo it's might helpful consider Hierarchy of Diseases when you are doing pre/post processing.\n\nHope to be helpful this information. Thanks", "votes": 14 }, { "id": 1200598, "postDate": "2021-02-14T19:15:58.800Z", "content": "Hi @jinssaa,
\nlung opacity and atelectasis are not diseases, but certain features on imaging. This figure is ambiguous, what is the purpose of consolidation? A lesion is also a collective term - it could mean pneumonia, cancer, etc. There are also many other diseases that needed to be classified. For ex. we may classify diseases as:
Thank you for mentioning @Medaxone.
\nYeah I might say combination of classes as well.
\nNow I'm starting to read below NB and Discuss to understand medical domain knowledge.
https://www.kaggle.com/zahaviguy/what-are-lung-opacities
\nhttps://www.kaggle.com/sakuraandblackcat/chest-x-ray-knowledges-for-the-14-abnormalities
\nhttps://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/212859
It might help to understand I want to know what classes mean is
", "rawMarkdown": "Thank you for mentioning @Medaxone.\nYeah I might say combination of classes as well.\nNow I'm starting to read below NB and Discuss to understand medical domain knowledge. \n\n[https://www.kaggle.com/zahaviguy/what-are-lung-opacities](https://www.kaggle.com/zahaviguy/what-are-lung-opacities)\n[https://www.kaggle.com/sakuraandblackcat/chest-x-ray-knowledges-for-the-14-abnormalities](https://www.kaggle.com/sakuraandblackcat/chest-x-ray-knowledges-for-the-14-abnormalities)\n[https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/212859](https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/212859)\n\nIt might help to understand I want to know what classes mean is" } ] }, { "id": 1252890, "postDate": "2021-03-26T06:39:03.177Z", "content": "thank you very much
", "rawMarkdown": "thank you very much" } ], "comments": [ { "id": 1200598, "author_name": "Medaxone", "author_url": "", "post_date": "2021-02-14T19:15:58.800000", "content": "Hi @jinssaa,
\nlung opacity and atelectasis are not diseases, but certain features on imaging. This figure is ambiguous, what is the purpose of consolidation? A lesion is also a collective term - it could mean pneumonia, cancer, etc. There are also many other diseases that needed to be classified. For ex. we may classify diseases as:
Thank you for mentioning @Medaxone.
\nYeah I might say combination of classes as well.
\nNow I'm starting to read below NB and Discuss to understand medical domain knowledge.
https://www.kaggle.com/zahaviguy/what-are-lung-opacities
\nhttps://www.kaggle.com/sakuraandblackcat/chest-x-ray-knowledges-for-the-14-abnormalities
\nhttps://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/212859
It might help to understand I want to know what classes mean is
", "votes": 0, "replies": [] } ] }, { "id": 1252890, "author_name": "EunilJung", "author_url": "", "post_date": "2021-03-26T06:39:03.177000", "content": "thank you very much
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1193271": "Hi guys, I just start to join this competition and I found paper which is 2nd place solution in [CheXpert](https://stanfordmlgroup.github.io/competitions/chexpert/) competition.\n\n[https://arxiv.org/pdf/1911.06475](https://arxiv.org/pdf/1911.06475)\n\nThis paper is written by Vingroup Big Data Institute(Competition Host)\n\nIt's seems like that there is hierarchy of Classes(ex Lung Opacity --> Atelectasis) And They also tried to train model depend on hierarchy of diseases and ensemble.\n\n![](https://drive.google.com/uc?export=view&id=1mN4SNZrToK62H6vgkkKM-dIdOPaxepLj\")\n\nSo it's might helpful consider Hierarchy of Diseases when you are doing pre/post processing.\n\nHope to be helpful this information. Thanks", "1200598": "Hi @jinssaa,\nlung opacity and atelectasis are not diseases, but certain features on imaging. This figure is ambiguous, what is the purpose of consolidation? A lesion is also a collective term - it could mean pneumonia, cancer, etc. There are also many other diseases that needed to be classified. For ex. we may classify diseases as: \n1. Lung/parenchymal diseases\n2. Airways diseases\n3. Pleural diseases\n4. Mediastinal diseases\n5. Blood vessel diseases \n5. Bone diseases", "1252890": "thank you very much" } }