How is the TRUE volume for the cases with number of slices <=3 are determined by the doctors? Thanks... I try to implement algorithms that completely mimic the process that a human figures out the volume, but there is not much detail about the process except that humans draw contour lines.
", "messages": [ { "id": "105147", "postDate": "01/20/2016 14:00:15", "content": "How is the TRUE volume for the cases with number of slices <=3 are determined by the doctors? Thanks... I try to implement algorithms that completely mimic the process that a human figures out the volume, but there is not much detail about the process except that humans draw contour lines.
", "rawMarkdown": "How is the TRUE volume for the cases with number of slices <=3 are determined by the doctors? Thanks... I try to implement algorithms that completely mimic the process that a human figures out the volume, but there is not much detail about the process except that humans draw contour lines.", "votes": null }, { "id": "105168", "postDate": "01/20/2016 16:10:51", "content": "@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!
", "rawMarkdown": "woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!", "votes": null }, { "id": "105169", "postDate": "01/20/2016 16:14:27", "content": "examples with slices <3: 595,597,599\nthere are some in the train set too.
\n\n[quote=Tim Hochberg;105168]
\n\n@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!
\n\n[/quote]
", "rawMarkdown": "examples with slices <3: 595,597,599\r\nthere are some in the train set too.\r\n\r\n[quote=Tim Hochberg;105168]\r\n\r\n@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!\r\n\r\n[/quote]", "votes": null }, { "id": "105173", "postDate": "01/20/2016 16:43:42", "content": "@woshialex, your right, I was looking at something else, sorry for the noise.
", "rawMarkdown": "woshialex, your right, I was looking at something else, sorry for the noise.", "votes": null } ], "comments": [ { "id": 105168, "author_name": "bitsofbits", "author_url": "", "post_date": "01/20/2016 16:10:51", "content": "@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!
", "votes": null, "replies": [] }, { "id": 105169, "author_name": "woshialex", "author_url": "", "post_date": "01/20/2016 16:14:27", "content": "examples with slices <3: 595,597,599\nthere are some in the train set too.
\n\n[quote=Tim Hochberg;105168]
\n\n@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!
\n\n[/quote]
", "votes": null, "replies": [] }, { "id": 105173, "author_name": "bitsofbits", "author_url": "", "post_date": "01/20/2016 16:43:42", "content": "@woshialex, your right, I was looking at something else, sorry for the noise.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "105147": "How is the TRUE volume for the cases with number of slices <=3 are determined by the doctors? Thanks... I try to implement algorithms that completely mimic the process that a human figures out the volume, but there is not much detail about the process except that humans draw contour lines.", "105168": "woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!", "105169": "examples with slices <3: 595,597,599\r\nthere are some in the train set too.\r\n\r\n[quote=Tim Hochberg;105168]\r\n\r\n@woshialex, I don't think there actually are any cases with slices <=3, there are just some cases where something is weird about either the directory layout or where the number of time-frames is less than 30 for some slices. For instance, there is one case where there are many slices with time-frame counts of 23, 24, etc, but only one slice where all 30 time frames are there. My guess is that in this case the frames taken capture both diastole and systole, so they could probably be segmented by hand without a problem. It does make painful for certain machine learning approaches though!\r\n\r\n[/quote]", "105173": "woshialex, your right, I was looking at something else, sorry for the noise." }, "source": "meta" }