{
  "id": 145194,
  "title": "Mislabeled Data?",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/145194",
  "author_name": "Leonie",
  "post_date": "2020-04-22T08:19:21.446000",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>During my EDA, I found a suspicious image id where the <code>isup_grade</code> seems to be not aligned correctly with the <code>gleason_score</code>. I'm not sure whether the <code>isup_grade</code> should be 3 or the <code>gleason_score</code> should be 3+4 instead.\nindex | image_id | data_provider | isup_grade | gleason_score|\n| -- | -- | |\n7273 | b0a92a74cb53899311acc30b7405e101 | karolinska | 2 | 4+3|</p>",
  "messages": [
    {
      "id": 816268,
      "postDate": "2020-04-22T08:19:21.447Z",
      "content": "<p>During my EDA, I found a suspicious image id where the <code>isup_grade</code> seems to be not aligned correctly with the <code>gleason_score</code>. I'm not sure whether the <code>isup_grade</code> should be 3 or the <code>gleason_score</code> should be 3+4 instead.\nindex | image_id | data_provider | isup_grade | gleason_score|\n| -- | -- | |\n7273 | b0a92a74cb53899311acc30b7405e101 | karolinska | 2 | 4+3|</p>",
      "rawMarkdown": "During my EDA, I found a suspicious image id where the `isup_grade` seems to be not aligned correctly with the `gleason_score`. I'm not sure whether the `isup_grade` should be 3 or the `gleason_score` should be 3+4 instead.\nindex | image_id | data_provider | isup_grade | gleason_score|\n| -- | -- | |\n7273 | b0a92a74cb53899311acc30b7405e101 | karolinska | 2 | 4+3|\n",
      "votes": 3
    },
    {
      "id": 817432,
      "postDate": "2020-04-23T06:29:46.763Z",
      "content": "<p>The short answer is that both are correct. </p>\n\n<p>As you may have understood, the Gleason/ISUP labels are somewhat subjective and different pathologists may choose different labels. All the Karolinska images in the training data is graded by the same pathologist. However, for the test set we used several pathologists who each labeled the images using ISUP (not Gleason) and derived a consensus label. The mislabeled image was one of those images but was later moved to the training set. So, the Gleason score is the original label by the study pathologist and the ISUP is the consensus label by the pathologists who graded the test set. We should have used the original ISUP label to avoid confusion, sorry for this.</p>",
      "rawMarkdown": "The short answer is that both are correct. \n\nAs you may have understood, the Gleason/ISUP labels are somewhat subjective and different pathologists may choose different labels. All the Karolinska images in the training data is graded by the same pathologist. However, for the test set we used several pathologists who each labeled the images using ISUP (not Gleason) and derived a consensus label. The mislabeled image was one of those images but was later moved to the training set. So, the Gleason score is the original label by the study pathologist and the ISUP is the consensus label by the pathologists who graded the test set. We should have used the original ISUP label to avoid confusion, sorry for this.\n",
      "votes": 4,
      "replies": [
        {
          "id": 817474,
          "postDate": "2020-04-23T07:30:46.293Z",
          "content": "<p>Thank you for the thorough explanation. This helped a lot!</p>",
          "rawMarkdown": "Thank you for the thorough explanation. This helped a lot!"
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 817432,
      "author_name": "PeterStröm",
      "author_url": "",
      "post_date": "2020-04-23T06:29:46.763000",
      "content": "<p>The short answer is that both are correct. </p>\n\n<p>As you may have understood, the Gleason/ISUP labels are somewhat subjective and different pathologists may choose different labels. All the Karolinska images in the training data is graded by the same pathologist. However, for the test set we used several pathologists who each labeled the images using ISUP (not Gleason) and derived a consensus label. The mislabeled image was one of those images but was later moved to the training set. So, the Gleason score is the original label by the study pathologist and the ISUP is the consensus label by the pathologists who graded the test set. We should have used the original ISUP label to avoid confusion, sorry for this.</p>",
      "votes": 4,
      "replies": [
        {
          "id": 817474,
          "author_name": "Leonie",
          "author_url": "",
          "post_date": "2020-04-23T07:30:46.293000",
          "content": "<p>Thank you for the thorough explanation. This helped a lot!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "816268": "During my EDA, I found a suspicious image id where the `isup_grade` seems to be not aligned correctly with the `gleason_score`. I'm not sure whether the `isup_grade` should be 3 or the `gleason_score` should be 3+4 instead.\nindex | image_id | data_provider | isup_grade | gleason_score|\n| -- | -- | |\n7273 | b0a92a74cb53899311acc30b7405e101 | karolinska | 2 | 4+3|\n",
    "817432": "The short answer is that both are correct. \n\nAs you may have understood, the Gleason/ISUP labels are somewhat subjective and different pathologists may choose different labels. All the Karolinska images in the training data is graded by the same pathologist. However, for the test set we used several pathologists who each labeled the images using ISUP (not Gleason) and derived a consensus label. The mislabeled image was one of those images but was later moved to the training set. So, the Gleason score is the original label by the study pathologist and the ISUP is the consensus label by the pathologists who graded the test set. We should have used the original ISUP label to avoid confusion, sorry for this.\n"
  }
}