{
  "id": 597070,
  "title": "Patient-Level Splitting: Patients Contributing More than One Scan",
  "url": "/competitions/rsna-intracranial-aneurysm-detection/discussion/597070",
  "author_name": "adi_hatake",
  "post_date": "2025-08-06T08:51:56.788000",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>As a preface, I am a beginner and I just want to ask if patient IDs are provided. Have patients contributed more than one scan in the training set? </p>\n<p>I am a bit unfamiliar with medical data and I was wondering if patient-level splitting would be an appropriate approach here as I've seen in some other forums -- does this prevent data leakage? </p>\n<p>Thanks!</p>\n<p>Edit: <br>\nDid more reading and saw this paragraph: \"All three tags are included in the DICOMs for conformity to the DICOM standard, but only SeriesUID is meaningful. The PatientID and StudyUID tags do not identify anything.\"</p>\n<p>Does this mean each training series is unique? Thanks</p>",
  "messages": [
    {
      "id": 3265660,
      "postDate": "2025-08-07T20:05:44.010Z",
      "content": "<p>Currently we are not providing any information regarding presence or absence multiple scans per patient. We recommend that you treat each series as completely independent.</p>",
      "rawMarkdown": "Currently we are not providing any information regarding presence or absence multiple scans per patient. We recommend that you treat each series as completely independent.",
      "votes": 1
    },
    {
      "id": 3264155,
      "postDate": "2025-08-06T08:51:56.790Z",
      "content": "<p>Hi,</p>\n<p>As a preface, I am a beginner and I just want to ask if patient IDs are provided. Have patients contributed more than one scan in the training set? </p>\n<p>I am a bit unfamiliar with medical data and I was wondering if patient-level splitting would be an appropriate approach here as I've seen in some other forums -- does this prevent data leakage? </p>\n<p>Thanks!</p>\n<p>Edit: <br>\nDid more reading and saw this paragraph: \"All three tags are included in the DICOMs for conformity to the DICOM standard, but only SeriesUID is meaningful. The PatientID and StudyUID tags do not identify anything.\"</p>\n<p>Does this mean each training series is unique? Thanks</p>",
      "rawMarkdown": "Hi,\n\nAs a preface, I am a beginner and I just want to ask if patient IDs are provided. Have patients contributed more than one scan in the training set? \n\nI am a bit unfamiliar with medical data and I was wondering if patient-level splitting would be an appropriate approach here as I've seen in some other forums -- does this prevent data leakage? \n\nThanks!\n\nEdit: \nDid more reading and saw this paragraph: \"All three tags are included in the DICOMs for conformity to the DICOM standard, but only SeriesUID is meaningful. The PatientID and StudyUID tags do not identify anything.\"\n\nDoes this mean each training series is unique? Thanks",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 3265660,
      "author_name": "Evan Calabrese",
      "author_url": "",
      "post_date": "2025-08-07T20:05:44.010000",
      "content": "<p>Currently we are not providing any information regarding presence or absence multiple scans per patient. We recommend that you treat each series as completely independent.</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3265660": "Currently we are not providing any information regarding presence or absence multiple scans per patient. We recommend that you treat each series as completely independent.",
    "3264155": "Hi,\n\nAs a preface, I am a beginner and I just want to ask if patient IDs are provided. Have patients contributed more than one scan in the training set? \n\nI am a bit unfamiliar with medical data and I was wondering if patient-level splitting would be an appropriate approach here as I've seen in some other forums -- does this prevent data leakage? \n\nThanks!\n\nEdit: \nDid more reading and saw this paragraph: \"All three tags are included in the DICOMs for conformity to the DICOM standard, but only SeriesUID is meaningful. The PatientID and StudyUID tags do not identify anything.\"\n\nDoes this mean each training series is unique? Thanks"
  }
}