\n
![\"image\"](\"https://i.ibb.co/DYh5s5B/image.png\")
In one of the discussions it was pointed out that images taken from different centers have a small changes in color properties since, the slide preparation processes may differ across sites.
\n@barbaroserdal had mentioned that to overcome this we can use stain normalization methodologies
I found this package torchstain here at https://github.com/EIDOSLAB/torchstain
\n
Found another package StainTools here at https://github.com/Peter554/StainTools
\nThis also provides normalization and augmentation
\n(I guess this one is not well maintained so please check everything once before using it)
In one of the discussions it was pointed out that images taken from different centers have a small changes in color properties since, the slide preparation processes may differ across sites.
\n@barbaroserdal had mentioned that to overcome this we can use stain normalization methodologies
I found this package torchstain here at https://github.com/EIDOSLAB/torchstain
\n
Found another package StainTools here at https://github.com/Peter554/StainTools
\nThis also provides normalization and augmentation
\n(I guess this one is not well maintained so please check everything once before using it)
\n\nFound another package **StainTools** here at https://github.com/Peter554/StainTools\nThis also provides normalization and augmentation\n(I guess this one is not well maintained so please check everything once before using it)",
"votes": 13
},
{
"id": 1871231,
"postDate": "2022-07-26T06:47:55.717Z",
"content": "While stain normalization will likely be helpful for the competition, this package in particular likely will not be since it is only for H&E normalization, and I don't think the stains used for the images in this dataset are H&E stains.
", "rawMarkdown": "While stain normalization will likely be helpful for the competition, this package in particular likely will not be since it is only for H&E normalization, and I don't think the stains used for the images in this dataset are H&E stains.", "votes": 4, "replies": [ { "id": 1871317, "postDate": "2022-07-26T07:59:32.427Z", "content": "Oh okay
\nThanks for pointing this out @tanlikesmath !
So do you suggest any methods for normalization for this competition in particular?
", "rawMarkdown": "Oh okay\nThanks for pointing this out @tanlikesmath !\n\nSo do you suggest any methods for normalization for this competition in particular?" }, { "id": 1933494, "postDate": "2022-09-10T14:49:57.120Z", "content": "@ilovescience how do you know that the stains in this dataset are not H&E stains?
", "rawMarkdown": "@ilovescience how do you know that the stains in this dataset are not H&E stains?" }, { "id": 1944517, "postDate": "2022-09-18T10:52:17.927Z", "content": "I think it is because the images contain yellow.😄
", "rawMarkdown": "I think it is because the images contain yellow.😄", "votes": 2 } ] }, { "id": 1871372, "postDate": "2022-07-26T08:37:22.297Z", "content": "Looks really promising. Great find!
", "rawMarkdown": "Looks really promising. Great find!\n\n" }, { "id": 1868392, "postDate": "2022-07-24T00:15:06.613Z", "content": "Great material. Thanks for sharing!
", "rawMarkdown": "Great material. Thanks for sharing!", "votes": 1 } ], "comments": [ { "id": 1871231, "author_name": "ilovescience", "author_url": "", "post_date": "2022-07-26T06:47:55.717000", "content": "While stain normalization will likely be helpful for the competition, this package in particular likely will not be since it is only for H&E normalization, and I don't think the stains used for the images in this dataset are H&E stains.
", "votes": 4, "replies": [ { "id": 1871317, "author_name": "Pathik Ghugare", "author_url": "", "post_date": "2022-07-26T07:59:32.427000", "content": "Oh okay
\nThanks for pointing this out @tanlikesmath !
So do you suggest any methods for normalization for this competition in particular?
", "votes": 0, "replies": [] }, { "id": 1933494, "author_name": "cosmosaa", "author_url": "", "post_date": "2022-09-10T14:49:57.120000", "content": "@ilovescience how do you know that the stains in this dataset are not H&E stains?
", "votes": 0, "replies": [] }, { "id": 1944517, "author_name": "Hiroshi Sakiyama", "author_url": "", "post_date": "2022-09-18T10:52:17.927000", "content": "I think it is because the images contain yellow.😄
", "votes": 2, "replies": [] } ] }, { "id": 1871372, "author_name": "The Devastator", "author_url": "", "post_date": "2022-07-26T08:37:22.297000", "content": "Looks really promising. Great find!
", "votes": 0, "replies": [] }, { "id": 1868392, "author_name": "Iuryck Santos", "author_url": "", "post_date": "2022-07-24T00:15:06.613000", "content": "Great material. Thanks for sharing!
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "1867724": "In one of the [discussions](https://www.kaggle.com/competitions/mayo-clinic-strip-ai/discussion/337811) it was pointed out that images taken from different centers have a small changes in color properties since, the slide preparation processes may differ across sites. \n@barbaroserdal had mentioned that to overcome this we can use **stain normalization methodologies**\n\nI found this package **torchstain** here at https://github.com/EIDOSLAB/torchstain\n\n\nFound another package **StainTools** here at https://github.com/Peter554/StainTools\nThis also provides normalization and augmentation\n(I guess this one is not well maintained so please check everything once before using it)",
"1871231": "While stain normalization will likely be helpful for the competition, this package in particular likely will not be since it is only for H&E normalization, and I don't think the stains used for the images in this dataset are H&E stains.",
"1871372": "Looks really promising. Great find!\n\n",
"1868392": "Great material. Thanks for sharing!"
}
}