{
  "id": 145204,
  "title": "How to handle WSI?",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/145204",
  "author_name": "seefun",
  "post_date": "2020-04-22T09:08:51.297000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Whole Slide Image (WSI) has quite large image size. Could you share your insight about how to handle the huge input size? Like patch classification?</p>\n\n<hr>\n\n<p>update:</p>\n\n<p>In this <a href=\"https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/146836\">discussion</a>, we down sample the image to make it small.</p>",
  "messages": [
    {
      "id": 816331,
      "postDate": "2020-04-22T09:32:44.763Z",
      "content": "<p>This is a good question. A good baseline would be to honestly resize and directly use the WSI image, as shown <a href=\"https://www.kaggle.com/tanlikesmath/fastai2-training-baseline\">here</a>. Another approach, as you mentioned, is patch classification. I am considering selecting a few large patches (ex: ~10) and resizing them and using that. Hopefully, that is a technique that works well.</p>\n\n<p>It seems patch-based classification is the most commonly used method.</p>",
      "rawMarkdown": "This is a good question. A good baseline would be to honestly resize and directly use the WSI image, as shown [here](https://www.kaggle.com/tanlikesmath/fastai2-training-baseline). Another approach, as you mentioned, is patch classification. I am considering selecting a few large patches (ex: ~10) and resizing them and using that. Hopefully, that is a technique that works well.\n\nIt seems patch-based classification is the most commonly used method.",
      "votes": 1
    },
    {
      "id": 816315,
      "postDate": "2020-04-22T09:08:51.297Z",
      "content": "<p>Whole Slide Image (WSI) has quite large image size. Could you share your insight about how to handle the huge input size? Like patch classification?</p>\n\n<hr>\n\n<p>update:</p>\n\n<p>In this <a href=\"https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/146836\">discussion</a>, we down sample the image to make it small.</p>",
      "rawMarkdown": "Whole Slide Image (WSI) has quite large image size. Could you share your insight about how to handle the huge input size? Like patch classification?\n\n----------\n\nupdate:\n\nIn this [discussion](https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/146836), we down sample the image to make it small."
    }
  ],
  "comments": [
    {
      "id": 816331,
      "author_name": "ilovescience",
      "author_url": "",
      "post_date": "2020-04-22T09:32:44.763000",
      "content": "<p>This is a good question. A good baseline would be to honestly resize and directly use the WSI image, as shown <a href=\"https://www.kaggle.com/tanlikesmath/fastai2-training-baseline\">here</a>. Another approach, as you mentioned, is patch classification. I am considering selecting a few large patches (ex: ~10) and resizing them and using that. Hopefully, that is a technique that works well.</p>\n\n<p>It seems patch-based classification is the most commonly used method.</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "816331": "This is a good question. A good baseline would be to honestly resize and directly use the WSI image, as shown [here](https://www.kaggle.com/tanlikesmath/fastai2-training-baseline). Another approach, as you mentioned, is patch classification. I am considering selecting a few large patches (ex: ~10) and resizing them and using that. Hopefully, that is a technique that works well.\n\nIt seems patch-based classification is the most commonly used method.",
    "816315": "Whole Slide Image (WSI) has quite large image size. Could you share your insight about how to handle the huge input size? Like patch classification?\n\n----------\n\nupdate:\n\nIn this [discussion](https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/146836), we down sample the image to make it small."
  }
}