{
  "id": 174712,
  "title": "Melanoma identification using \"contextual\" images",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/174712",
  "author_name": "PrasunMishra",
  "post_date": "2020-08-14T20:33:42.935000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Folks,<br>\nI was reading this sentence:<br>\n\"Dermatologists could enhance their diagnostic accuracy if detection algorithms take into account “contextual” images within the same patient to determine which images represent a melanoma.\"<br>\nGiven the data set has a group of images from a subset of patients (i think 482 unique patients) and each of these patients has multiple images. One of the images is\"malignant\".<br>\nOne approach could be to train within images of this 'sub-group' of patients (who have at least 1 malign image) and even upsample on the same 'sub-data'. Any thoughts? Any success?</p>",
  "messages": [
    {
      "id": 970812,
      "postDate": "2020-08-14T20:33:42.937Z",
      "content": "<p>Folks,<br>\nI was reading this sentence:<br>\n\"Dermatologists could enhance their diagnostic accuracy if detection algorithms take into account “contextual” images within the same patient to determine which images represent a melanoma.\"<br>\nGiven the data set has a group of images from a subset of patients (i think 482 unique patients) and each of these patients has multiple images. One of the images is\"malignant\".<br>\nOne approach could be to train within images of this 'sub-group' of patients (who have at least 1 malign image) and even upsample on the same 'sub-data'. Any thoughts? Any success?</p>",
      "rawMarkdown": "Folks,\nI was reading this sentence:\n\"Dermatologists could enhance their diagnostic accuracy if detection algorithms take into account “contextual” images within the same patient to determine which images represent a melanoma.\"\nGiven the data set has a group of images from a subset of patients (i think 482 unique patients) and each of these patients has multiple images. One of the images is\"malignant\".\nOne approach could be to train within images of this 'sub-group' of patients (who have at least 1 malign image) and even upsample on the same 'sub-data'. Any thoughts? Any success?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "970812": "Folks,\nI was reading this sentence:\n\"Dermatologists could enhance their diagnostic accuracy if detection algorithms take into account “contextual” images within the same patient to determine which images represent a melanoma.\"\nGiven the data set has a group of images from a subset of patients (i think 482 unique patients) and each of these patients has multiple images. One of the images is\"malignant\".\nOne approach could be to train within images of this 'sub-group' of patients (who have at least 1 malign image) and even upsample on the same 'sub-data'. Any thoughts? Any success?"
  }
}