Hello everyone. This is the first time I join a medical imaging competition. I want to ask a few questions about the data.
\nThank you.
", "messages": [ { "id": 967809, "postDate": "2020-08-12T14:19:02.870Z", "content": "Hello everyone. This is the first time I join a medical imaging competition. I want to ask a few questions about the data.
\nThank you.
", "rawMarkdown": "Hello everyone. This is the first time I join a medical imaging competition. I want to ask a few questions about the data.\n\n1. What is meant by \"baseline CT scan\"? Why is it possible for the \"week\" to be negative?\n2. The csv file has multiple rows of the same patient ID, and each folder in \"train\" has multiple DICOM files. Can anyone explain how to interpret the data? If the same patient undergoes multiple scans throughout a period of time, then the DICOM files correspond to which week? And also, are the DICOM files the slices of the CT scan, as in 3D image represented in multiple 2D images?\n\nThank you.", "votes": 9 }, { "id": 968121, "postDate": "2020-08-12T17:56:53.750Z", "content": "From what I can understand, the weeks are in reference to when the scan is taken, so negative weeks mean that the measurements were taken however many weeks before the week of that CT scan. From the data tab:
\n\n\nA patient has an image acquired at time Week = 0 and has numerous follow up visits over the course of approximately 1-2 years, at which time their FVC is measured.
\n
Basically saying that the weeks in the csv files are relative to the week of the CT scan.
\nEach row for each patient refers to a measurement taken on a given week. So if a patient has 8 measurements taken, there will be 8 rows for that patient corresponding to each week the patient had that measurement taken (recall that each week is in reference to when the CT scan was taken i.e. -3 means three weeks before, 6 means 6 weeks after, etc).
\nFor the images, each patient has only one scan; it's just that each of these scans have varying amounts of images (which I'm pretty sure are 2D slices of the entire 3D image). So: in the train folder you have a folder for each patient, in each of which is all the dicom files for that patient's CT scan.
\nAs far as I can tell, people are struggling to make effective use of the images, so good luck! Also, anyone please correct any misinformation I might be spreading here.
", "rawMarkdown": "From what I can understand, the weeks are in reference to when the scan is taken, so negative weeks mean that the measurements were taken however many weeks *before* the week of that CT scan. From the data tab: \n\n> A patient has an image acquired at time Week = 0 and has numerous follow up visits over the course of approximately 1-2 years, at which time their FVC is measured.\n\nBasically saying that the weeks in the csv files are relative to the week of the CT scan.\n\nEach row for each patient refers to a measurement taken on a given week. So if a patient has 8 measurements taken, there will be 8 rows for that patient corresponding to each week the patient had that measurement taken (recall that each week is in reference to when the CT scan was taken i.e. -3 means three weeks before, 6 means 6 weeks after, etc). \n\nFor the images, each patient has only one scan; it's just that each of these scans have varying amounts of images (which I'm pretty sure are 2D slices of the entire 3D image). So: in the train folder you have a folder for each patient, in each of which is all the dicom files for that patient's CT scan.\n\nAs far as I can tell, people are struggling to make effective use of the images, so good luck! Also, anyone please correct any misinformation I might be spreading here.", "votes": 3, "replies": [ { "id": 968180, "postDate": "2020-08-12T18:50:18.937Z", "content": "Thanks. But I wonder what the criteria of choosing the baseline CT scan is.
", "rawMarkdown": "Thanks. But I wonder what the criteria of choosing the baseline CT scan is. " }, { "id": 968203, "postDate": "2020-08-12T19:08:02.547Z", "content": "¯\\(ツ)/¯
\nI'm only assuming it's taken at the beginning/toward the beginning of the diagnosis. It does bring about an interesting question: Do the lungs visibly change over the course of the decline?
\nIf so, then the time of the CT scan is much more relevant, I think (assuming the time is what you're talking about). It also brings the question of how to incorporate that into the learning because I'm not sure if the model would be able to learn such correlations with the given data. If not though, I can't find too much of a reason to give it much attention.
", "rawMarkdown": "¯\\\\_(ツ)_/¯\n\nI'm only assuming it's taken at the beginning/toward the beginning of the diagnosis. It does bring about an interesting question: Do the lungs visibly change over the course of the decline? \n\nIf so, then the time of the CT scan is much more relevant, I think (assuming the time is what you're talking about). It also brings the question of how to incorporate that into the learning because I'm not sure if the model would be able to learn such correlations with the given data. If not though, I can't find too much of a reason to give it much attention." }, { "id": 968279, "postDate": "2020-08-12T20:23:12.490Z", "rawMarkdown": "", "isDeleted": true }, { "id": 973948, "postDate": "2020-08-17T16:14:29.373Z", "content": "@jtan2231 your statement Basically saying that the weeks in the csv files are relative to the week of the CT scan.
\nwhat you said was each week in the csv data is taken FVC and we will have images with same corresponding week which was not the case in the data , for example for patient(ID00007637202177411956430) we have weeks [-4 5 7 9 11 17 29 41 57] and images we have 30 images .
\nBy your explanation for it should have 9 images in train folder in ID00007637202177411956430 patient id but we have 30 images please have a kernal with EDA.
https://www.kaggle.com/kunduruanil/getstarted-kernal-eda
", "rawMarkdown": "@jtan2231 your statement Basically saying that the weeks in the csv files are relative to the week of the CT scan.\n\nwhat you said was each week in the csv data is taken FVC and we will have images with same corresponding week which was not the case in the data , for example for patient(ID00007637202177411956430) we have weeks [-4 5 7 9 11 17 29 41 57] and images we have 30 images .\nBy your explanation for it should have 9 images in train folder in ID00007637202177411956430 patient id but we have 30 images please have a kernal with EDA.\n\n\nhttps://www.kaggle.com/kunduruanil/getstarted-kernal-eda" }, { "id": 974158, "postDate": "2020-08-17T18:57:04.977Z", "content": "Apologies for the misunderstanding. By scan in this quote:
\n\n\nBasically saying that the weeks in the csv files are relative to the week of the CT scan.
\n
I was referring to the collection of slices for each patient that forms a 3D image. So, using your example of patient ID00007637202177411956430, each image in that folder corresponds to a slice in that patient's CT scan; i.e. the collection of the 30 2D dicom images in that folder make up the single 3D CT scan.
", "rawMarkdown": "Apologies for the misunderstanding. By scan in this quote:\n> Basically saying that the weeks in the csv files are relative to the week of the CT scan.\n\nI was referring to the collection of slices for each patient that forms a 3D image. So, using your example of patient ID00007637202177411956430, each image in that folder corresponds to a slice in that patient's CT scan; i.e. the collection of the 30 2D dicom images in that folder make up the single 3D CT scan." } ] }, { "id": 973337, "postDate": "2020-08-17T08:53:40.450Z", "content": "Hi there's already a thread for Questions about Data you can find some answers here
", "rawMarkdown": "Hi there's already a thread for Questions about Data you can find some answers [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/165149) ", "votes": 2 } ], "comments": [ { "id": 968121, "author_name": "Joseph Tan", "author_url": "", "post_date": "2020-08-12T17:56:53.750000", "content": "From what I can understand, the weeks are in reference to when the scan is taken, so negative weeks mean that the measurements were taken however many weeks before the week of that CT scan. From the data tab:
\n\n\nA patient has an image acquired at time Week = 0 and has numerous follow up visits over the course of approximately 1-2 years, at which time their FVC is measured.
\n
Basically saying that the weeks in the csv files are relative to the week of the CT scan.
\nEach row for each patient refers to a measurement taken on a given week. So if a patient has 8 measurements taken, there will be 8 rows for that patient corresponding to each week the patient had that measurement taken (recall that each week is in reference to when the CT scan was taken i.e. -3 means three weeks before, 6 means 6 weeks after, etc).
\nFor the images, each patient has only one scan; it's just that each of these scans have varying amounts of images (which I'm pretty sure are 2D slices of the entire 3D image). So: in the train folder you have a folder for each patient, in each of which is all the dicom files for that patient's CT scan.
\nAs far as I can tell, people are struggling to make effective use of the images, so good luck! Also, anyone please correct any misinformation I might be spreading here.
", "votes": 3, "replies": [ { "id": 968180, "author_name": "Stephen Lau", "author_url": "", "post_date": "2020-08-12T18:50:18.937000", "content": "Thanks. But I wonder what the criteria of choosing the baseline CT scan is.
", "votes": 0, "replies": [] }, { "id": 968203, "author_name": "Joseph Tan", "author_url": "", "post_date": "2020-08-12T19:08:02.547000", "content": "¯\\(ツ)/¯
\nI'm only assuming it's taken at the beginning/toward the beginning of the diagnosis. It does bring about an interesting question: Do the lungs visibly change over the course of the decline?
\nIf so, then the time of the CT scan is much more relevant, I think (assuming the time is what you're talking about). It also brings the question of how to incorporate that into the learning because I'm not sure if the model would be able to learn such correlations with the given data. If not though, I can't find too much of a reason to give it much attention.
", "votes": 0, "replies": [] }, { "id": 968279, "author_name": "", "author_url": "", "post_date": "2020-08-12T20:23:12.490000", "content": "", "votes": 0, "replies": [] }, { "id": 973948, "author_name": "Anil Kumar Reddy", "author_url": "", "post_date": "2020-08-17T16:14:29.373000", "content": "@jtan2231 your statement Basically saying that the weeks in the csv files are relative to the week of the CT scan.
\nwhat you said was each week in the csv data is taken FVC and we will have images with same corresponding week which was not the case in the data , for example for patient(ID00007637202177411956430) we have weeks [-4 5 7 9 11 17 29 41 57] and images we have 30 images .
\nBy your explanation for it should have 9 images in train folder in ID00007637202177411956430 patient id but we have 30 images please have a kernal with EDA.
https://www.kaggle.com/kunduruanil/getstarted-kernal-eda
", "votes": 0, "replies": [] }, { "id": 974158, "author_name": "Joseph Tan", "author_url": "", "post_date": "2020-08-17T18:57:04.977000", "content": "Apologies for the misunderstanding. By scan in this quote:
\n\n\nBasically saying that the weeks in the csv files are relative to the week of the CT scan.
\n
I was referring to the collection of slices for each patient that forms a 3D image. So, using your example of patient ID00007637202177411956430, each image in that folder corresponds to a slice in that patient's CT scan; i.e. the collection of the 30 2D dicom images in that folder make up the single 3D CT scan.
", "votes": 0, "replies": [] } ] }, { "id": 973337, "author_name": "Amed", "author_url": "", "post_date": "2020-08-17T08:53:40.450000", "content": "Hi there's already a thread for Questions about Data you can find some answers here
", "votes": 2, "replies": [] } ], "raw_markdown_by_id": { "967809": "Hello everyone. This is the first time I join a medical imaging competition. I want to ask a few questions about the data.\n\n1. What is meant by \"baseline CT scan\"? Why is it possible for the \"week\" to be negative?\n2. The csv file has multiple rows of the same patient ID, and each folder in \"train\" has multiple DICOM files. Can anyone explain how to interpret the data? If the same patient undergoes multiple scans throughout a period of time, then the DICOM files correspond to which week? And also, are the DICOM files the slices of the CT scan, as in 3D image represented in multiple 2D images?\n\nThank you.", "968121": "From what I can understand, the weeks are in reference to when the scan is taken, so negative weeks mean that the measurements were taken however many weeks *before* the week of that CT scan. From the data tab: \n\n> A patient has an image acquired at time Week = 0 and has numerous follow up visits over the course of approximately 1-2 years, at which time their FVC is measured.\n\nBasically saying that the weeks in the csv files are relative to the week of the CT scan.\n\nEach row for each patient refers to a measurement taken on a given week. So if a patient has 8 measurements taken, there will be 8 rows for that patient corresponding to each week the patient had that measurement taken (recall that each week is in reference to when the CT scan was taken i.e. -3 means three weeks before, 6 means 6 weeks after, etc). \n\nFor the images, each patient has only one scan; it's just that each of these scans have varying amounts of images (which I'm pretty sure are 2D slices of the entire 3D image). So: in the train folder you have a folder for each patient, in each of which is all the dicom files for that patient's CT scan.\n\nAs far as I can tell, people are struggling to make effective use of the images, so good luck! Also, anyone please correct any misinformation I might be spreading here.", "973337": "Hi there's already a thread for Questions about Data you can find some answers [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/165149) " } }