{
  "id": 373734,
  "title": " AI really can detect cancers...? This problem looks really hard even for human.",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/373734",
  "author_name": "GarudA_Kai",
  "post_date": "2022-12-23T04:10:18.366000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I looked at many positive examples individually to see where the cancer was located, but thought it too difficult for me (I have nothing to do with medicine). Many of them looked just the same as the negative examples. This is probably because mammography is used as a <strong>screening test</strong> and can only tell us if it is a \"not problematic\" or a \"suspicious\". Even professionals sometimes cannot detect if it is definitely cancer… can AI really do that?!<br>\nHowever, some of the positive examples have <strong>a clear white and oval hue</strong> indicating cancer. So as a ML task, we could focus on these \"clearly suspicious examples\" . </p>\n<h1>Example Images</h1>\n<h2>\"A clearly suspicious example\"</h2>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fab8a3f5ebf86db23f0a0c31e7154a958%2F14941_303808170.png?generation=1671777342058831&amp;alt=media\" alt=\"img\"></p>\n<h2>\"a positive example that seems to be the same as negative\"</h2>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fcbc4fbd520c87de62931e1196134cebe%2F10589_605115808.png?generation=1671777227003084&amp;alt=media\" alt=\"img\"></p>\n<p>I am completely a beginner in medical science, so if you find some mistakes, please tell me quickly!</p>\n<p>This is my dataset which contains only positive examples. -&gt; <a href=\"https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only\" target=\"_blank\">https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only</a></p>",
  "messages": [
    {
      "id": 2073430,
      "postDate": "2022-12-23T04:10:18.367Z",
      "content": "<p>I looked at many positive examples individually to see where the cancer was located, but thought it too difficult for me (I have nothing to do with medicine). Many of them looked just the same as the negative examples. This is probably because mammography is used as a <strong>screening test</strong> and can only tell us if it is a \"not problematic\" or a \"suspicious\". Even professionals sometimes cannot detect if it is definitely cancer… can AI really do that?!<br>\nHowever, some of the positive examples have <strong>a clear white and oval hue</strong> indicating cancer. So as a ML task, we could focus on these \"clearly suspicious examples\" . </p>\n<h1>Example Images</h1>\n<h2>\"A clearly suspicious example\"</h2>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fab8a3f5ebf86db23f0a0c31e7154a958%2F14941_303808170.png?generation=1671777342058831&amp;alt=media\" alt=\"img\"></p>\n<h2>\"a positive example that seems to be the same as negative\"</h2>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fcbc4fbd520c87de62931e1196134cebe%2F10589_605115808.png?generation=1671777227003084&amp;alt=media\" alt=\"img\"></p>\n<p>I am completely a beginner in medical science, so if you find some mistakes, please tell me quickly!</p>\n<p>This is my dataset which contains only positive examples. -&gt; <a href=\"https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only\" target=\"_blank\">https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only</a></p>",
      "rawMarkdown": "I looked at many positive examples individually to see where the cancer was located, but thought it too difficult for me (I have nothing to do with medicine). Many of them looked just the same as the negative examples. This is probably because mammography is used as a **screening test** and can only tell us if it is a \"not problematic\" or a \"suspicious\". Even professionals sometimes cannot detect if it is definitely cancer... can AI really do that?!\nHowever, some of the positive examples have **a clear white and oval hue** indicating cancer. So as a ML task, we could focus on these \"clearly suspicious examples\" . \n\n#  Example Images\n## \"A clearly suspicious example\"\n\n![img](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fab8a3f5ebf86db23f0a0c31e7154a958%2F14941_303808170.png?generation=1671777342058831&alt=media)\n\n## \"a positive example that seems to be the same as negative\"\n![img](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fcbc4fbd520c87de62931e1196134cebe%2F10589_605115808.png?generation=1671777227003084&alt=media)\n\nI am completely a beginner in medical science, so if you find some mistakes, please tell me quickly!\n\nThis is my dataset which contains only positive examples. -> https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2073430": "I looked at many positive examples individually to see where the cancer was located, but thought it too difficult for me (I have nothing to do with medicine). Many of them looked just the same as the negative examples. This is probably because mammography is used as a **screening test** and can only tell us if it is a \"not problematic\" or a \"suspicious\". Even professionals sometimes cannot detect if it is definitely cancer... can AI really do that?!\nHowever, some of the positive examples have **a clear white and oval hue** indicating cancer. So as a ML task, we could focus on these \"clearly suspicious examples\" . \n\n#  Example Images\n## \"A clearly suspicious example\"\n\n![img](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fab8a3f5ebf86db23f0a0c31e7154a958%2F14941_303808170.png?generation=1671777342058831&alt=media)\n\n## \"a positive example that seems to be the same as negative\"\n![img](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8631933%2Fcbc4fbd520c87de62931e1196134cebe%2F10589_605115808.png?generation=1671777227003084&alt=media)\n\nI am completely a beginner in medical science, so if you find some mistakes, please tell me quickly!\n\nThis is my dataset which contains only positive examples. -> https://www.kaggle.com/datasets/garudakai/rsna-breast-cancer-1024-pngs-positive-only"
  }
}