Previous competitions were about X Rays, like the RSNA Bone Age Challenge 2017, RSNA Pneumonia Challenge 2018 and the SIIM-ACR Pneumothorax Challenge 2019.
\n\nThis competition is about head CT, a type of cross-sectional imaging. Each study is made of stacked slices of the patient (not the patient, you got it). Each slice is stored in one DICOM file.
\n\nThe Data section states:
\n\n\n\n\nAll provided images are in DICOM format. DICOM images contain associated metadata. This will include PatientID, StudyInstanceUID, SeriesInstanceUID, and other features. You will notice some PatientIDs represented in both the stage 1 train and test sets. This is known and intentional. However, there will be no crossover of PatientIDs into stage 2 test. Additionally, per the rules, \"Submission predictions must be based entirely on the pixel data in the provided datasets.\" Therefore, you should not expect to use or gain advantage by use of this crossover in stage 1.
\n
My advice based on previous competitions is this: beware of this fact and take actions based on it to train your models.
", "messages": [ { "id": 629097, "postDate": "2019-09-18T10:52:33.637Z", "content": "Previous competitions were about X Rays, like the RSNA Bone Age Challenge 2017, RSNA Pneumonia Challenge 2018 and the SIIM-ACR Pneumothorax Challenge 2019.
\n\nThis competition is about head CT, a type of cross-sectional imaging. Each study is made of stacked slices of the patient (not the patient, you got it). Each slice is stored in one DICOM file.
\n\nThe Data section states:
\n\n\n\n\nAll provided images are in DICOM format. DICOM images contain associated metadata. This will include PatientID, StudyInstanceUID, SeriesInstanceUID, and other features. You will notice some PatientIDs represented in both the stage 1 train and test sets. This is known and intentional. However, there will be no crossover of PatientIDs into stage 2 test. Additionally, per the rules, \"Submission predictions must be based entirely on the pixel data in the provided datasets.\" Therefore, you should not expect to use or gain advantage by use of this crossover in stage 1.
\n
My advice based on previous competitions is this: beware of this fact and take actions based on it to train your models.
", "rawMarkdown": "Previous competitions were about X Rays, like the RSNA Bone Age Challenge 2017, RSNA Pneumonia Challenge 2018 and the SIIM-ACR Pneumothorax Challenge 2019.\n\nThis competition is about head CT, a type of cross-sectional imaging. Each study is made of stacked slices of the patient (not the patient, you got it). Each slice is stored in one DICOM file.\n\nThe Data section states:\n\n> All provided images are in DICOM format. DICOM images contain associated metadata. This will include PatientID, StudyInstanceUID, SeriesInstanceUID, and other features. You will notice some PatientIDs represented in both the stage 1 train and test sets. This is known and intentional. However, there will be no crossover of PatientIDs into stage 2 test. Additionally, per the rules, \"Submission predictions must be based entirely on the pixel data in the provided datasets.\" Therefore, you should not expect to use or gain advantage by use of this crossover in stage 1.\n\nMy advice based on previous competitions is this: beware of this fact and take actions based on it to train your models.", "votes": 5 }, { "id": 629629, "postDate": "2019-09-19T02:07:10.033Z", "content": "Even if the images come from the same patient, they may show different findings. Some images may be completely normal, even if other show a bleed.
", "rawMarkdown": "Even if the images come from the same patient, they may show different findings. Some images may be completely normal, even if other show a bleed.", "votes": 2, "replies": [ { "id": 630642, "postDate": "2019-09-20T14:33:39.053Z", "content": "And they will be labeled differently in this case? In other words: does label mean \"there is a bleed on this image\" or maybe \"this patient has a bleed\"?
", "rawMarkdown": "And they will be labeled differently in this case? In other words: does label mean \"there is a bleed on this image\" or maybe \"this patient has a bleed\"?", "votes": 3 }, { "id": 630663, "postDate": "2019-09-20T15:03:38.493Z", "content": "My understanding is that the labels mean \"there is a bleed on this image\"
", "rawMarkdown": "My understanding is that the labels mean \"there is a bleed on this image\" ", "votes": 3 } ] }, { "id": 629553, "postDate": "2019-09-18T22:56:46.813Z", "content": "\"Submission predictions must be based entirely on the pixel data in the provided datasets.\" does this mean we cannot use PatientID to identify patient and do what you suggest?
", "rawMarkdown": "\"Submission predictions must be based entirely on the pixel data in the provided datasets.\" does this mean we cannot use PatientID to identify patient and do what you suggest?", "replies": [ { "id": 629607, "postDate": "2019-09-19T01:09:31.480Z", "content": "I was suggesting to follow these rules.
", "rawMarkdown": "I was suggesting to follow these rules." } ] }, { "id": 629211, "postDate": "2019-09-18T14:08:03.107Z", "content": "Thanks
", "rawMarkdown": "Thanks" } ], "comments": [ { "id": 629629, "author_name": "Amil Gentili", "author_url": "", "post_date": "2019-09-19T02:07:10.033000", "content": "Even if the images come from the same patient, they may show different findings. Some images may be completely normal, even if other show a bleed.
", "votes": 2, "replies": [ { "id": 630642, "author_name": "Adams", "author_url": "", "post_date": "2019-09-20T14:33:39.053000", "content": "And they will be labeled differently in this case? In other words: does label mean \"there is a bleed on this image\" or maybe \"this patient has a bleed\"?
", "votes": 3, "replies": [] }, { "id": 630663, "author_name": "Amil Gentili", "author_url": "", "post_date": "2019-09-20T15:03:38.493000", "content": "My understanding is that the labels mean \"there is a bleed on this image\"
", "votes": 3, "replies": [] } ] }, { "id": 629553, "author_name": "Zhenlan", "author_url": "", "post_date": "2019-09-18T22:56:46.813000", "content": "\"Submission predictions must be based entirely on the pixel data in the provided datasets.\" does this mean we cannot use PatientID to identify patient and do what you suggest?
", "votes": 0, "replies": [ { "id": 629607, "author_name": "FelipeKitamura, MD, PhD", "author_url": "", "post_date": "2019-09-19T01:09:31.480000", "content": "I was suggesting to follow these rules.
", "votes": 0, "replies": [] } ] }, { "id": 629211, "author_name": "Aniket Dixit", "author_url": "", "post_date": "2019-09-18T14:08:03.107000", "content": "Thanks
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "629097": "Previous competitions were about X Rays, like the RSNA Bone Age Challenge 2017, RSNA Pneumonia Challenge 2018 and the SIIM-ACR Pneumothorax Challenge 2019.\n\nThis competition is about head CT, a type of cross-sectional imaging. Each study is made of stacked slices of the patient (not the patient, you got it). Each slice is stored in one DICOM file.\n\nThe Data section states:\n\n> All provided images are in DICOM format. DICOM images contain associated metadata. This will include PatientID, StudyInstanceUID, SeriesInstanceUID, and other features. You will notice some PatientIDs represented in both the stage 1 train and test sets. This is known and intentional. However, there will be no crossover of PatientIDs into stage 2 test. Additionally, per the rules, \"Submission predictions must be based entirely on the pixel data in the provided datasets.\" Therefore, you should not expect to use or gain advantage by use of this crossover in stage 1.\n\nMy advice based on previous competitions is this: beware of this fact and take actions based on it to train your models.", "629629": "Even if the images come from the same patient, they may show different findings. Some images may be completely normal, even if other show a bleed.", "629553": "\"Submission predictions must be based entirely on the pixel data in the provided datasets.\" does this mean we cannot use PatientID to identify patient and do what you suggest?", "629211": "Thanks" } }