{
  "id": 155135,
  "title": "Bad input samples",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/155135",
  "author_name": "Artur Fattakhov (MIPT DIHT)",
  "post_date": "2020-05-31T13:49:45.956000",
  "votes": 8,
  "comment_count": 1,
  "views": 0,
  "content": "<p>The training data contains samples of poor quality, which can not accurately determine the cancer. For example, on the preparation 014006841b9807edc0ff277c4ab29b91 the cut is too thick, on 00d8a8c04886379e266406fdeff81c45 there is too little tissue with cores. It is likely that students who marked up the train did not have enough experience in real diagnostics and therefore did not miss such samples. It is very risky to diagnose a patient with a low quality input sample. As I understand it, more experienced pathologists marked out the test data, did they pass incomprehensible glasses? Or made an approximate and inaccurate diagnosis?</p>",
  "messages": [
    {
      "id": 868794,
      "postDate": "2020-05-31T13:49:45.957Z",
      "content": "<p>The training data contains samples of poor quality, which can not accurately determine the cancer. For example, on the preparation 014006841b9807edc0ff277c4ab29b91 the cut is too thick, on 00d8a8c04886379e266406fdeff81c45 there is too little tissue with cores. It is likely that students who marked up the train did not have enough experience in real diagnostics and therefore did not miss such samples. It is very risky to diagnose a patient with a low quality input sample. As I understand it, more experienced pathologists marked out the test data, did they pass incomprehensible glasses? Or made an approximate and inaccurate diagnosis?</p>",
      "rawMarkdown": "The training data contains samples of poor quality, which can not accurately determine the cancer. For example, on the preparation 014006841b9807edc0ff277c4ab29b91 the cut is too thick, on 00d8a8c04886379e266406fdeff81c45 there is too little tissue with cores. It is likely that students who marked up the train did not have enough experience in real diagnostics and therefore did not miss such samples. It is very risky to diagnose a patient with a low quality input sample. As I understand it, more experienced pathologists marked out the test data, did they pass incomprehensible glasses? Or made an approximate and inaccurate diagnosis?",
      "votes": 8
    },
    {
      "id": 871307,
      "postDate": "2020-06-02T09:00:19.217Z",
      "content": "<p>Good observation. The training set can indeed contain noise or low quality samples. This is not the case for the test set. The test set has been graded by multiple expert pathologists. If the experts decided that a diagnosis could not be given due to low image quality, that case was discarded. </p>",
      "rawMarkdown": "Good observation. The training set can indeed contain noise or low quality samples. This is not the case for the test set. The test set has been graded by multiple expert pathologists. If the experts decided that a diagnosis could not be given due to low image quality, that case was discarded. ",
      "votes": 3
    }
  ],
  "comments": [
    {
      "id": 871307,
      "author_name": "Wouter Bulten",
      "author_url": "",
      "post_date": "2020-06-02T09:00:19.217000",
      "content": "<p>Good observation. The training set can indeed contain noise or low quality samples. This is not the case for the test set. The test set has been graded by multiple expert pathologists. If the experts decided that a diagnosis could not be given due to low image quality, that case was discarded. </p>",
      "votes": 3,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "868794": "The training data contains samples of poor quality, which can not accurately determine the cancer. For example, on the preparation 014006841b9807edc0ff277c4ab29b91 the cut is too thick, on 00d8a8c04886379e266406fdeff81c45 there is too little tissue with cores. It is likely that students who marked up the train did not have enough experience in real diagnostics and therefore did not miss such samples. It is very risky to diagnose a patient with a low quality input sample. As I understand it, more experienced pathologists marked out the test data, did they pass incomprehensible glasses? Or made an approximate and inaccurate diagnosis?",
    "871307": "Good observation. The training set can indeed contain noise or low quality samples. This is not the case for the test set. The test set has been graded by multiple expert pathologists. If the experts decided that a diagnosis could not be given due to low image quality, that case was discarded. "
  }
}