I've combined all masks for each training image and put them into a pdf file (see attachment) so you can quickly get an idea of the variations in the images.
", "messages": [ { "id": "269985", "postDate": "01/17/2018 16:05:54", "content": "I've combined all masks for each training image and put them into a pdf file (see attachment) so you can quickly get an idea of the variations in the images.
", "rawMarkdown": "I've combined all masks for each training image and put them into a pdf file (see attachment) so you can quickly get an idea of the variations in the images.", "votes": null }, { "id": "270030", "postDate": "01/17/2018 17:09:29", "content": "Thanks for sharing!!!
", "rawMarkdown": "Thanks for sharing!!!", "votes": null }, { "id": "270056", "postDate": "01/17/2018 17:42:55", "content": "Thank you for doing this. It is an excellent starting point, especially because I am yet to download images. I will post few comments here regarding the images in your file and how well the masks represent the nuclei. It will probably be several posts as I don't have time to go through all the images.
\n\nYou may want to read some general considerations I posted here. Same disclaimer as in that post: when considering my observations, keep in mind that I am not a cell biologist - I only play one on TV.
\n\nFigure in page 1 is a relatively high magnification of cells - probably 400x. Most nuclei in it are masked fine, with the exception of two smaller dots at 3 o'clock.
\n\nPage 2 is lower magnification, ~50x most likely. Again, small dot at 4:30 next to a bigger one is not a cell. The same is true for dots just above and to the right from image center in page 4.
\n\nPage 5 is interesting. I would have to see it in higher magnification, but it seems here that whole cells are stained rather than nuclei. If so, the nucleus would be a relatively less stained oval in the middle.
\n\nPage 7: dots selected by image mask are some cytoplasmic structures. Nucleus is actually the less stained part in the middle.
\n\nPage 8: image masks are correct, but in this one you can actually see the faint outline of complete cells around nuclei.
\n\nDon't have much experience with histological staining that starts at page 10, but general principles are the same.
\n\nTough to interpret page 17. Some spots with strong signal are artefacts, but others could be cell structures. Almost none of brightest spots are actually nuclei.
\n\nSimilar argument for page 21 as for page 5 above. At least 4 clumped cells in the middle.
\n\nPage 26 is an example of a poorly focused image. Unfortunately, we may have to deal with few of those.
\n\nIt is relatively easy to say for page 41 that a group of dots at 1 o'clock is not a nucleus.
\n\nLots of connective tissue in page with only few elongated cells. Whole cells appear to be stained, so it is tougher to identify just nucleus in them.
\n\nSame observation for page 53 as for page 7.
\n\nIn page 56 nuclei are relatively unstained ovals in the middle, rather than the strong signal just outside of them. The signal outside of them is most likely Golgi aparatus.
\n\nTo be continued ...
", "rawMarkdown": "Thank you for doing this. It is an excellent starting point, especially because I am yet to download images. I will post few comments here regarding the images in your file and how well the masks represent the nuclei. It will probably be several posts as I don't have time to go through all the images.\n\nYou may want to read some general considerations I posted [__here__](https://www.kaggle.com/c/data-science-bowl-2018/discussion/47681). Same disclaimer as in that post: when considering my observations, keep in mind that I am not a cell biologist - I only play one on TV.\n\nFigure in page 1 is a relatively high magnification of cells - probably 400x. Most nuclei in it are masked fine, with the exception of two smaller dots at 3 o'clock.\n\nPage 2 is lower magnification, ~50x most likely. Again, small dot at 4:30 next to a bigger one is not a cell. The same is true for dots just above and to the right from image center in page 4.\n\nPage 5 is interesting. I would have to see it in higher magnification, but it seems here that whole cells are stained rather than nuclei. If so, the nucleus would be a relatively less stained oval in the middle.\n\nPage 7: dots selected by image mask are some cytoplasmic structures. Nucleus is actually the less stained part in the middle.\n\nPage 8: image masks are correct, but in this one you can actually see the faint outline of complete cells around nuclei.\n\nDon't have much experience with histological staining that starts at page 10, but general principles are the same.\n\nTough to interpret page 17. Some spots with strong signal are artefacts, but others could be cell structures. Almost none of brightest spots are actually nuclei.\n\nSimilar argument for page 21 as for page 5 above. At least 4 clumped cells in the middle.\n\nPage 26 is an example of a poorly focused image. Unfortunately, we may have to deal with few of those.\n\nIt is relatively easy to say for page 41 that a group of dots at 1 o'clock is not a nucleus.\n\nLots of connective tissue in page with only few elongated cells. Whole cells appear to be stained, so it is tougher to identify just nucleus in them.\n\nSame observation for page 53 as for page 7.\n\nIn page 56 nuclei are relatively unstained ovals in the middle, rather than the strong signal just outside of them. The signal outside of them is most likely Golgi aparatus.\n\nTo be continued ...", "votes": null }, { "id": "271179", "postDate": "01/19/2018 21:02:14", "content": "Awesome, it's really helpful to identify the poor performance on certain image, and get a general idea how the dataset looks like.
", "rawMarkdown": "Awesome, it's really helpful to identify the poor performance on certain image, and get a general idea how the dataset looks like.", "votes": null }, { "id": "277257", "postDate": "02/02/2018 16:50:30", "content": "Thanks for sharing
", "rawMarkdown": "Thanks for sharing", "votes": null }, { "id": "278095", "postDate": "02/05/2018 11:19:01", "content": "Good job and thanks for sharing.
\n\nAt a first look, most nuclei is easy to identify by human eye, most image have no disturbing shapes
", "rawMarkdown": "Good job and thanks for sharing.\n\nAt a first look, most nuclei is easy to identify by human eye, most image have no disturbing shapes", "votes": null }, { "id": "278166", "postDate": "02/05/2018 14:25:37", "content": "@Tilii. Just a FYI.... I love the disclaimer \" I am not a cell biologist - I only play one on TV\". I chuckle everytime I read your post. Thank you for all your insight. :)
", "rawMarkdown": "Tilii. Just a FYI.... I love the disclaimer \" I am not a cell biologist - I only play one on TV\". I chuckle everytime I read your post. Thank you for all your insight. :)", "votes": null } ], "comments": [ { "id": 270030, "author_name": "andrearapuzzi", "author_url": "", "post_date": "01/17/2018 17:09:29", "content": "Thanks for sharing!!!
", "votes": null, "replies": [] }, { "id": 270056, "author_name": "tilii7", "author_url": "", "post_date": "01/17/2018 17:42:55", "content": "Thank you for doing this. It is an excellent starting point, especially because I am yet to download images. I will post few comments here regarding the images in your file and how well the masks represent the nuclei. It will probably be several posts as I don't have time to go through all the images.
\n\nYou may want to read some general considerations I posted here. Same disclaimer as in that post: when considering my observations, keep in mind that I am not a cell biologist - I only play one on TV.
\n\nFigure in page 1 is a relatively high magnification of cells - probably 400x. Most nuclei in it are masked fine, with the exception of two smaller dots at 3 o'clock.
\n\nPage 2 is lower magnification, ~50x most likely. Again, small dot at 4:30 next to a bigger one is not a cell. The same is true for dots just above and to the right from image center in page 4.
\n\nPage 5 is interesting. I would have to see it in higher magnification, but it seems here that whole cells are stained rather than nuclei. If so, the nucleus would be a relatively less stained oval in the middle.
\n\nPage 7: dots selected by image mask are some cytoplasmic structures. Nucleus is actually the less stained part in the middle.
\n\nPage 8: image masks are correct, but in this one you can actually see the faint outline of complete cells around nuclei.
\n\nDon't have much experience with histological staining that starts at page 10, but general principles are the same.
\n\nTough to interpret page 17. Some spots with strong signal are artefacts, but others could be cell structures. Almost none of brightest spots are actually nuclei.
\n\nSimilar argument for page 21 as for page 5 above. At least 4 clumped cells in the middle.
\n\nPage 26 is an example of a poorly focused image. Unfortunately, we may have to deal with few of those.
\n\nIt is relatively easy to say for page 41 that a group of dots at 1 o'clock is not a nucleus.
\n\nLots of connective tissue in page with only few elongated cells. Whole cells appear to be stained, so it is tougher to identify just nucleus in them.
\n\nSame observation for page 53 as for page 7.
\n\nIn page 56 nuclei are relatively unstained ovals in the middle, rather than the strong signal just outside of them. The signal outside of them is most likely Golgi aparatus.
\n\nTo be continued ...
", "votes": null, "replies": [ { "id": 278166, "author_name": "dskswu", "author_url": "", "post_date": "02/05/2018 14:25:37", "content": "@Tilii. Just a FYI.... I love the disclaimer \" I am not a cell biologist - I only play one on TV\". I chuckle everytime I read your post. Thank you for all your insight. :)
", "votes": null, "replies": [] } ] }, { "id": 271179, "author_name": "abnerchou", "author_url": "", "post_date": "01/19/2018 21:02:14", "content": "Awesome, it's really helpful to identify the poor performance on certain image, and get a general idea how the dataset looks like.
", "votes": null, "replies": [] }, { "id": 277257, "author_name": "omoekan09", "author_url": "", "post_date": "02/02/2018 16:50:30", "content": "Thanks for sharing
", "votes": null, "replies": [] }, { "id": 278095, "author_name": "amxineohp", "author_url": "", "post_date": "02/05/2018 11:19:01", "content": "Good job and thanks for sharing.
\n\nAt a first look, most nuclei is easy to identify by human eye, most image have no disturbing shapes
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "269985": "I've combined all masks for each training image and put them into a pdf file (see attachment) so you can quickly get an idea of the variations in the images.", "270030": "Thanks for sharing!!!", "270056": "Thank you for doing this. It is an excellent starting point, especially because I am yet to download images. I will post few comments here regarding the images in your file and how well the masks represent the nuclei. It will probably be several posts as I don't have time to go through all the images.\n\nYou may want to read some general considerations I posted [__here__](https://www.kaggle.com/c/data-science-bowl-2018/discussion/47681). Same disclaimer as in that post: when considering my observations, keep in mind that I am not a cell biologist - I only play one on TV.\n\nFigure in page 1 is a relatively high magnification of cells - probably 400x. Most nuclei in it are masked fine, with the exception of two smaller dots at 3 o'clock.\n\nPage 2 is lower magnification, ~50x most likely. Again, small dot at 4:30 next to a bigger one is not a cell. The same is true for dots just above and to the right from image center in page 4.\n\nPage 5 is interesting. I would have to see it in higher magnification, but it seems here that whole cells are stained rather than nuclei. If so, the nucleus would be a relatively less stained oval in the middle.\n\nPage 7: dots selected by image mask are some cytoplasmic structures. Nucleus is actually the less stained part in the middle.\n\nPage 8: image masks are correct, but in this one you can actually see the faint outline of complete cells around nuclei.\n\nDon't have much experience with histological staining that starts at page 10, but general principles are the same.\n\nTough to interpret page 17. Some spots with strong signal are artefacts, but others could be cell structures. Almost none of brightest spots are actually nuclei.\n\nSimilar argument for page 21 as for page 5 above. At least 4 clumped cells in the middle.\n\nPage 26 is an example of a poorly focused image. Unfortunately, we may have to deal with few of those.\n\nIt is relatively easy to say for page 41 that a group of dots at 1 o'clock is not a nucleus.\n\nLots of connective tissue in page with only few elongated cells. Whole cells appear to be stained, so it is tougher to identify just nucleus in them.\n\nSame observation for page 53 as for page 7.\n\nIn page 56 nuclei are relatively unstained ovals in the middle, rather than the strong signal just outside of them. The signal outside of them is most likely Golgi aparatus.\n\nTo be continued ...", "271179": "Awesome, it's really helpful to identify the poor performance on certain image, and get a general idea how the dataset looks like.", "277257": "Thanks for sharing", "278095": "Good job and thanks for sharing.\n\nAt a first look, most nuclei is easy to identify by human eye, most image have no disturbing shapes", "278166": "Tilii. Just a FYI.... I love the disclaimer \" I am not a cell biologist - I only play one on TV\". I chuckle everytime I read your post. Thank you for all your insight. :)" }, "source": "meta" }