{
  "competition": "rsna-knee-abnormality-detection",
  "topic_id": "740718",
  "comments": [
    {
      "id": 3523638,
      "authorName": "ParsaAshrafi",
      "votes": 0,
      "postDate": "2026-09-12T08:23:32.370000",
      "content": "<p>Based on Section 3.4(b), the restriction applies to hand-labeling or human prediction of validation/test records. It does not prohibit manually reviewing the provided training MRIs. Since those images are part of the Competition Data, participant-created labels for the training set would generally be derived labels rather than External Data.</p>\n<p>I would still recommend documenting the labeling protocol and asking the host for explicit confirmation before relying on it, especially if the labels are used in a final submission.</p>"
    },
    {
      "id": 3523299,
      "authorName": "PC Jimmmy",
      "votes": 0,
      "postDate": "2026-09-11T03:17:23.667000",
      "content": "<p>I believe the rules state that you cannot hand label test images - that's not applicable to this competition as we cannot see the test images.  It was applicable in competitions in the past when the test images were available for download/review.</p>\n<p>If you want to manually examine and label the train images - go for it.  Modification of train data has NEVER been considered as \"external data\".</p>"
    }
  ],
  "messages": [],
  "raw_show": {
    "topic": {
      "id": 740718,
      "title": "Clarification on Manual Annotation of Unlabeled Training MRIs",
      "authorName": "oshiko0450",
      "commentCount": 2,
      "votes": 0,
      "postDate": "2026-09-10T21:40:23.352000"
    },
    "comments": [
      {
        "id": 3523638,
        "authorName": "ParsaAshrafi",
        "votes": 0,
        "postDate": "2026-09-12T08:23:32.370000",
        "content": "<p>Based on Section 3.4(b), the restriction applies to hand-labeling or human prediction of validation/test records. It does not prohibit manually reviewing the provided training MRIs. Since those images are part of the Competition Data, participant-created labels for the training set would generally be derived labels rather than External Data.</p>\n<p>I would still recommend documenting the labeling protocol and asking the host for explicit confirmation before relying on it, especially if the labels are used in a final submission.</p>"
      },
      {
        "id": 3523299,
        "authorName": "PC Jimmmy",
        "votes": 0,
        "postDate": "2026-09-11T03:17:23.667000",
        "content": "<p>I believe the rules state that you cannot hand label test images - that's not applicable to this competition as we cannot see the test images.  It was applicable in competitions in the past when the test images were available for download/review.</p>\n<p>If you want to manually examine and label the train images - go for it.  Modification of train data has NEVER been considered as \"external data\".</p>"
      }
    ]
  },
  "topic": {
    "id": 740718,
    "title": "Clarification on Manual Annotation of Unlabeled Training MRIs",
    "authorName": "oshiko0450",
    "commentCount": 2,
    "votes": 0,
    "postDate": "2026-09-10T21:40:23.352000"
  },
  "index": {
    "id": "740718",
    "title": "Clarification on Manual Annotation of Unlabeled Training MRIs",
    "authorName": "",
    "commentCount": "2",
    "votes": "0",
    "postDate": "2026-09-10 21:40:23.352000"
  }
}