{
  "id": 108023,
  "title": "Models compared to expert doctor/team of doctors",
  "url": "/competitions/aptos2019-blindness-detection/discussion/108023",
  "author_name": "",
  "post_date": "2019-09-08T14:16:19.469553600Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>If there is any way we can know how top models performed in comparison to expert doctors. The main aim of this competition was <strong>to gain the ability to automatically screen images for disease and provide information on how severe the condition may be</strong>. If the organizers can share some insights this will be great.</p>",
  "messages": [
    {
      "id": "621441",
      "postDate": "09/08/2019 14:16:19",
      "content": "<p>If there is any way we can know how top models performed in comparison to expert doctors. The main aim of this competition was <strong>to gain the ability to automatically screen images for disease and provide information on how severe the condition may be</strong>. If the organizers can share some insights this will be great.</p>",
      "rawMarkdown": "If there is any way we can know how top models performed in comparison to expert doctors. The main aim of this competition was **to gain the ability to automatically screen images for disease and provide information on how severe the condition may be**. If the organizers can share some insights this will be great.",
      "votes": null
    },
    {
      "id": "621448",
      "postDate": "09/08/2019 14:20:04",
      "content": "<p>You can use dataset of 2015 kaggle competition to indirectly sense how well doctors or a single doctor agree by measuring quadratic weighted kappa or drawing confusion matrix with the grades of right and left eyes. It is well known that two eyes have strong correlations in diabetic retinopathy. FYI, I remember it was ~0.87. But beware that the quadratic weighted kappa is vulnerable to the target distribution (distribution of the ground truth)</p>",
      "rawMarkdown": "You can use dataset of 2015 kaggle competition to indirectly sense how well doctors or a single doctor agree by measuring quadratic weighted kappa or drawing confusion matrix with the grades of right and left eyes. It is well known that two eyes have strong correlations in diabetic retinopathy. FYI, I remember it was ~0.87. But beware that the quadratic weighted kappa is vulnerable to the target distribution (distribution of the ground truth)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 621448,
      "author_name": "jaeminkkk",
      "author_url": "",
      "post_date": "09/08/2019 14:20:04",
      "content": "<p>You can use dataset of 2015 kaggle competition to indirectly sense how well doctors or a single doctor agree by measuring quadratic weighted kappa or drawing confusion matrix with the grades of right and left eyes. It is well known that two eyes have strong correlations in diabetic retinopathy. FYI, I remember it was ~0.87. But beware that the quadratic weighted kappa is vulnerable to the target distribution (distribution of the ground truth)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "621441": "If there is any way we can know how top models performed in comparison to expert doctors. The main aim of this competition was **to gain the ability to automatically screen images for disease and provide information on how severe the condition may be**. If the organizers can share some insights this will be great.",
    "621448": "You can use dataset of 2015 kaggle competition to indirectly sense how well doctors or a single doctor agree by measuring quadratic weighted kappa or drawing confusion matrix with the grades of right and left eyes. It is well known that two eyes have strong correlations in diabetic retinopathy. FYI, I remember it was ~0.87. But beware that the quadratic weighted kappa is vulnerable to the target distribution (distribution of the ground truth)"
  },
  "source": "meta"
}