{
  "id": 354898,
  "title": "How to speed up  loading dicom files in submission ?",
  "url": "/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/354898",
  "author_name": "Mithil Salunkhe",
  "post_date": "2022-09-24T12:15:20.957000",
  "votes": 11,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I am currently trying to a build a pipeline for comp. I have noticed that there is a large io bottleneck because of time it takes to dicom files. Are there any ways to make loading dicom files faster ?</p>",
  "messages": [
    {
      "id": 1953352,
      "postDate": "2022-09-24T12:15:20.957Z",
      "content": "<p>I am currently trying to a build a pipeline for comp. I have noticed that there is a large io bottleneck because of time it takes to dicom files. Are there any ways to make loading dicom files faster ?</p>",
      "rawMarkdown": "I am currently trying to a build a pipeline for comp. I have noticed that there is a large io bottleneck because of time it takes to dicom files. Are there any ways to make loading dicom files faster ?",
      "votes": 11
    },
    {
      "id": 1973914,
      "postDate": "2022-10-06T00:14:52.530Z",
      "content": "<p>I'm sure you know this already, but in my case, there were a lot of loading \"unused\" slices. Rather than loading whole slices -&gt; resize, I calculated stride and loaded those only.</p>",
      "rawMarkdown": "I'm sure you know this already, but in my case, there were a lot of loading \"unused\" slices. Rather than loading whole slices -> resize, I calculated stride and loaded those only."
    },
    {
      "id": 1965245,
      "postDate": "2022-10-01T07:53:03.383Z",
      "content": "<p>I couldn't find a way other than multiprocessing. I use torch DataLoader with num_workers=2  for submission.</p>",
      "rawMarkdown": "I couldn't find a way other than multiprocessing. I use torch DataLoader with num_workers=2  for submission."
    }
  ],
  "comments": [
    {
      "id": 1973914,
      "author_name": "JunHyeonKwon",
      "author_url": "",
      "post_date": "2022-10-06T00:14:52.530000",
      "content": "<p>I'm sure you know this already, but in my case, there were a lot of loading \"unused\" slices. Rather than loading whole slices -&gt; resize, I calculated stride and loaded those only.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1965245,
      "author_name": "Aruta Niimura",
      "author_url": "",
      "post_date": "2022-10-01T07:53:03.383000",
      "content": "<p>I couldn't find a way other than multiprocessing. I use torch DataLoader with num_workers=2  for submission.</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1953352": "I am currently trying to a build a pipeline for comp. I have noticed that there is a large io bottleneck because of time it takes to dicom files. Are there any ways to make loading dicom files faster ?",
    "1973914": "I'm sure you know this already, but in my case, there were a lot of loading \"unused\" slices. Rather than loading whole slices -> resize, I calculated stride and loaded those only.",
    "1965245": "I couldn't find a way other than multiprocessing. I use torch DataLoader with num_workers=2  for submission."
  }
}