{
  "id": 275627,
  "title": "3D R-CNN? Extend 2D R-CNN for 3D?",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/275627",
  "author_name": "",
  "post_date": "2021-10-01T03:54:11.340170600Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>How do we modify R-CNN or Mask R-CNN to account for 3D images?</p>\n<p>I see 3D efficientnet but is there something we can do for segmentation as well?</p>\n<p>Thanks much for your suggestions.</p>",
  "messages": [
    {
      "id": "1530247",
      "postDate": "10/01/2021 03:54:11",
      "content": "<p>How do we modify R-CNN or Mask R-CNN to account for 3D images?</p>\n<p>I see 3D efficientnet but is there something we can do for segmentation as well?</p>\n<p>Thanks much for your suggestions.</p>",
      "rawMarkdown": "How do we modify R-CNN or Mask R-CNN to account for 3D images?\n\nI see 3D efficientnet but is there something we can do for segmentation as well?\n\nThanks much for your suggestions.",
      "votes": null
    },
    {
      "id": "1532099",
      "postDate": "10/02/2021 16:09:03",
      "content": "<p>For segmentation, I think we need to get the corresponding mask and not Mask-RCNN precisely. That being said, it's possible to model a segmentation model (3D UNet, etc) for this task and its required external data set. </p>\n<p><a href=\"https://www.kaggle.com/andrewmvd/brain-tumor-segmentation-in-mri-brats-2015\" target=\"_blank\">- Brain Tumor Segmentation</a></p>\n<p>I personally think that this competition would be more interesting if the mask gt was provided. </p>",
      "rawMarkdown": "For segmentation, I think we need to get the corresponding mask and not Mask-RCNN precisely. That being said, it's possible to model a segmentation model (3D UNet, etc) for this task and its required external data set. \n\n[- Brain Tumor Segmentation](https://www.kaggle.com/andrewmvd/brain-tumor-segmentation-in-mri-brats-2015)\n\nI personally think that this competition would be more interesting if the mask gt was provided.",
      "votes": null
    },
    {
      "id": "1532104",
      "postDate": "10/02/2021 16:10:52",
      "content": "<p>For the 3D Segmentation model, you can check out these implementations, <a href=\"https://github.com/ZFTurbo/segmentation_models_3D\" target=\"_blank\">ZFTurbo/segmentation_models_3D. </a></p>",
      "rawMarkdown": "For the 3D Segmentation model, you can check out these implementations, [ZFTurbo/segmentation_models_3D. ](https://github.com/ZFTurbo/segmentation_models_3D)",
      "votes": null
    },
    {
      "id": "1532917",
      "postDate": "10/03/2021 13:45:29",
      "content": "<p>thanks for sharing! do you know how much difference the data between 2020 BraTS segmentation and our task 2 data? e.g. the data in 2021 task1, seems like all be registered against an atlas (SRI24) </p>",
      "rawMarkdown": "thanks for sharing! do you know how much difference the data between 2020 BraTS segmentation and our task 2 data? e.g. the data in 2021 task1, seems like all be registered against an atlas (SRI24)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1532099,
      "author_name": "ipythonx",
      "author_url": "",
      "post_date": "10/02/2021 16:09:03",
      "content": "<p>For segmentation, I think we need to get the corresponding mask and not Mask-RCNN precisely. That being said, it's possible to model a segmentation model (3D UNet, etc) for this task and its required external data set. </p>\n<p><a href=\"https://www.kaggle.com/andrewmvd/brain-tumor-segmentation-in-mri-brats-2015\" target=\"_blank\">- Brain Tumor Segmentation</a></p>\n<p>I personally think that this competition would be more interesting if the mask gt was provided. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1532104,
          "author_name": "ipythonx",
          "author_url": "",
          "post_date": "10/02/2021 16:10:52",
          "content": "<p>For the 3D Segmentation model, you can check out these implementations, <a href=\"https://github.com/ZFTurbo/segmentation_models_3D\" target=\"_blank\">ZFTurbo/segmentation_models_3D. </a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1532917,
          "author_name": "samshipengs",
          "author_url": "",
          "post_date": "10/03/2021 13:45:29",
          "content": "<p>thanks for sharing! do you know how much difference the data between 2020 BraTS segmentation and our task 2 data? e.g. the data in 2021 task1, seems like all be registered against an atlas (SRI24) </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1530247": "How do we modify R-CNN or Mask R-CNN to account for 3D images?\n\nI see 3D efficientnet but is there something we can do for segmentation as well?\n\nThanks much for your suggestions.",
    "1532099": "For segmentation, I think we need to get the corresponding mask and not Mask-RCNN precisely. That being said, it's possible to model a segmentation model (3D UNet, etc) for this task and its required external data set. \n\n[- Brain Tumor Segmentation](https://www.kaggle.com/andrewmvd/brain-tumor-segmentation-in-mri-brats-2015)\n\nI personally think that this competition would be more interesting if the mask gt was provided.",
    "1532104": "For the 3D Segmentation model, you can check out these implementations, [ZFTurbo/segmentation_models_3D. ](https://github.com/ZFTurbo/segmentation_models_3D)",
    "1532917": "thanks for sharing! do you know how much difference the data between 2020 BraTS segmentation and our task 2 data? e.g. the data in 2021 task1, seems like all be registered against an atlas (SRI24)"
  },
  "source": "meta"
}