{ "id": 388989, "title": "Dataset exploration and observations", "url": "/competitions/rsna-breast-cancer-detection/discussion/388989", "author_name": "Koubouratou IDJATON", "post_date": "2023-02-20T13:29:04.639000", "votes": 0, "comment_count": 1, "views": 0, "content": "

Hello, kaggler!
\nI started this notebook about the dataset exploration and understanding.
\nI make these observations so far:
\nOn cancer:
\n97.9% of the train images are cancer free. A big class imbalance.
\nOn Laterality:
\nThe number of images in the train set are approximately equal for each kind laterality (L or R).
\nThey relatively equal number of cancer image in each kind of laterality.
\nOn Site:
\nThey are 5% more image from site 1 than from site 2.
\nIt also show more image with cancer from site 1 than from site 2.

\n

Please share any others interesting thing you have noted on the dataset, and state
\nAlso any suggestions are very welcome to improve and oriented me on some others things to look for to do a better exploration.
\nThank you

", "messages": [ { "id": 2158142, "postDate": "2023-02-24T16:50:28.390Z", "content": "

I find that machine 29 have the data that with large pixel values than other machines

\n

here are the visulization and the pixel value counting results]

\n

\"\"

\n

\"pixel

", "rawMarkdown": "I find that machine 29 have the data that with large pixel values than other machines\n\nhere are the visulization and the pixel value counting results]\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F998389%2F1597285fd4d4ac36f32d954ffe62f8d7%2FPicture3.png?generation=1677257403229737&alt=media)\n\n![pixel value very large](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F998389%2Fa444f4435d129b753a6fa73fdefec961%2FPicture2.png?generation=1677257320145302&alt=media)\n\n", "votes": 1 }, { "id": 2151940, "postDate": "2023-02-20T13:29:04.640Z", "content": "

Hello, kaggler!
\nI started this notebook about the dataset exploration and understanding.
\nI make these observations so far:
\nOn cancer:
\n97.9% of the train images are cancer free. A big class imbalance.
\nOn Laterality:
\nThe number of images in the train set are approximately equal for each kind laterality (L or R).
\nThey relatively equal number of cancer image in each kind of laterality.
\nOn Site:
\nThey are 5% more image from site 1 than from site 2.
\nIt also show more image with cancer from site 1 than from site 2.

\n

Please share any others interesting thing you have noted on the dataset, and state
\nAlso any suggestions are very welcome to improve and oriented me on some others things to look for to do a better exploration.
\nThank you

", "rawMarkdown": "Hello, kaggler!\nI started this [notebook](https://www.kaggle.com/code/koubouratouidjaton/rsna-breast-cancer-dataset-exploration) about the dataset exploration and understanding. \nI make these observations so far:\nOn cancer:\n97.9% of the train images are cancer free. A big class imbalance.\nOn Laterality:\nThe number of images in the train set are approximately equal for each kind laterality (L or R).\nThey relatively equal number of cancer image in each kind of laterality.\nOn Site:\nThey are 5% more image from site 1 than from site 2.\nIt also show more image with cancer from site 1 than from site 2.\n\nPlease share any others interesting thing you have noted on the dataset, and state\nAlso any suggestions are very welcome to improve and oriented me on some others things to look for to do a better exploration.\nThank you", "votes": -1 } ], "comments": [ { "id": 2158142, "author_name": "HongCheng", "author_url": "", "post_date": "2023-02-24T16:50:28.390000", "content": "

I find that machine 29 have the data that with large pixel values than other machines

\n

here are the visulization and the pixel value counting results]

\n

\"\"

\n

\"pixel

", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "2158142": "I find that machine 29 have the data that with large pixel values than other machines\n\nhere are the visulization and the pixel value counting results]\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F998389%2F1597285fd4d4ac36f32d954ffe62f8d7%2FPicture3.png?generation=1677257403229737&alt=media)\n\n![pixel value very large](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F998389%2Fa444f4435d129b753a6fa73fdefec961%2FPicture2.png?generation=1677257320145302&alt=media)\n\n", "2151940": "Hello, kaggler!\nI started this [notebook](https://www.kaggle.com/code/koubouratouidjaton/rsna-breast-cancer-dataset-exploration) about the dataset exploration and understanding. \nI make these observations so far:\nOn cancer:\n97.9% of the train images are cancer free. A big class imbalance.\nOn Laterality:\nThe number of images in the train set are approximately equal for each kind laterality (L or R).\nThey relatively equal number of cancer image in each kind of laterality.\nOn Site:\nThey are 5% more image from site 1 than from site 2.\nIt also show more image with cancer from site 1 than from site 2.\n\nPlease share any others interesting thing you have noted on the dataset, and state\nAlso any suggestions are very welcome to improve and oriented me on some others things to look for to do a better exploration.\nThank you" } }