I am doing some EDA and I have noticed that there are quite a few incomplete sequences, i.e. image series where only part of the head is captured. I was trying to figure out a way of determining something like the absolute position of a slice in the head so that I could pad missing slices with a sensible value instead of stretching the existing 'half head' over a pre-defined volume size. How are others dealing with this? Am I missing something?
", "messages": [ { "id": 3288235, "postDate": "2025-09-13T12:46:40.707Z", "content": "This reflects real world data where we often only image a part of the head for angiographic imaging. So its not that the series are \"incomplete\" just that they were acquired that way.
\nHoping others can comment on your approach. To me, each person will be unique in head shape and vessel position, so it seems difficult to try to create a coordinate system that would fit for all scans in the dataset.
", "rawMarkdown": "This reflects real world data where we often only image a part of the head for angiographic imaging. So its not that the series are \"incomplete\" just that they were acquired that way. \n\nHoping others can comment on your approach. To me, each person will be unique in head shape and vessel position, so it seems difficult to try to create a coordinate system that would fit for all scans in the dataset.", "votes": 2, "replies": [ { "id": 3288708, "postDate": "2025-09-14T15:21:30.703Z", "content": "Thanks for your reply. I understand that there is a lot of variability in these scans. I was just looking for a way to automatically detect whether we are dealing with a whole head or just a portion. I have now found the ImagePositionPatient field which has a z-coordinate value for each slice. I guess that's what I was looking for.
", "rawMarkdown": "Thanks for your reply. I understand that there is a lot of variability in these scans. I was just looking for a way to automatically detect whether we are dealing with a whole head or just a portion. I have now found the ImagePositionPatient field which has a z-coordinate value for each slice. I guess that's what I was looking for.", "replies": [ { "id": 3290381, "postDate": "2025-09-17T17:53:54.997Z", "content": "OK great. I hope that works, but that z coordinate origin is defined manually by the technologist, so it may not be the same across patients. Good luck!
", "rawMarkdown": "OK great. I hope that works, but that z coordinate origin is defined manually by the technologist, so it may not be the same across patients. Good luck!" } ] } ] }, { "id": 3288008, "postDate": "2025-09-12T20:01:35.890Z", "content": "I am doing some EDA and I have noticed that there are quite a few incomplete sequences, i.e. image series where only part of the head is captured. I was trying to figure out a way of determining something like the absolute position of a slice in the head so that I could pad missing slices with a sensible value instead of stretching the existing 'half head' over a pre-defined volume size. How are others dealing with this? Am I missing something?
", "rawMarkdown": "I am doing some EDA and I have noticed that there are quite a few incomplete sequences, i.e. image series where only part of the head is captured. I was trying to figure out a way of determining something like the absolute position of a slice in the head so that I could pad missing slices with a sensible value instead of stretching the existing 'half head' over a pre-defined volume size. How are others dealing with this? Am I missing something?" } ], "comments": [ { "id": 3288235, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-09-13T12:46:40.707000", "content": "This reflects real world data where we often only image a part of the head for angiographic imaging. So its not that the series are \"incomplete\" just that they were acquired that way.
\nHoping others can comment on your approach. To me, each person will be unique in head shape and vessel position, so it seems difficult to try to create a coordinate system that would fit for all scans in the dataset.
", "votes": 2, "replies": [ { "id": 3288708, "author_name": "thomas rost", "author_url": "", "post_date": "2025-09-14T15:21:30.703000", "content": "Thanks for your reply. I understand that there is a lot of variability in these scans. I was just looking for a way to automatically detect whether we are dealing with a whole head or just a portion. I have now found the ImagePositionPatient field which has a z-coordinate value for each slice. I guess that's what I was looking for.
", "votes": 0, "replies": [ { "id": 3290381, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-09-17T17:53:54.997000", "content": "OK great. I hope that works, but that z coordinate origin is defined manually by the technologist, so it may not be the same across patients. Good luck!
", "votes": 0, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3288235": "This reflects real world data where we often only image a part of the head for angiographic imaging. So its not that the series are \"incomplete\" just that they were acquired that way. \n\nHoping others can comment on your approach. To me, each person will be unique in head shape and vessel position, so it seems difficult to try to create a coordinate system that would fit for all scans in the dataset.", "3288008": "I am doing some EDA and I have noticed that there are quite a few incomplete sequences, i.e. image series where only part of the head is captured. I was trying to figure out a way of determining something like the absolute position of a slice in the head so that I could pad missing slices with a sensible value instead of stretching the existing 'half head' over a pre-defined volume size. How are others dealing with this? Am I missing something?" } }