{
  "id": 441298,
  "title": "Acceleromter device removed - What is threshold of variation signal to determine this?",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/441298",
  "author_name": "",
  "post_date": "2023-09-18T09:55:04.898888200Z",
  "votes": 7,
  "comment_count": 4,
  "views": 0,
  "content": "<p>How do we identify these periods with suspiciously low variation in accelerometer signals. Is there any threshold?<br>\nHas anyone created notebook which handles this?</p>\n<p><a href=\"https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/data\" target=\"_blank\">Though each series is a continuous recording, there may be periods in the series when the accelerometer device was removed. These period are determined as those where suspiciously little variation in the accelerometer signals occur over an extended period of time, which is unrealistic for typical human participants. Events are not annotated for these periods, and you should attempt to refrain from making event predictions during these periods: an event prediction will be scored as false positive.</a></p>",
  "messages": [
    {
      "id": "2444499",
      "postDate": "09/18/2023 09:55:04",
      "content": "<p>How do we identify these periods with suspiciously low variation in accelerometer signals. Is there any threshold?<br>\nHas anyone created notebook which handles this?</p>\n<p><a href=\"https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/data\" target=\"_blank\">Though each series is a continuous recording, there may be periods in the series when the accelerometer device was removed. These period are determined as those where suspiciously little variation in the accelerometer signals occur over an extended period of time, which is unrealistic for typical human participants. Events are not annotated for these periods, and you should attempt to refrain from making event predictions during these periods: an event prediction will be scored as false positive.</a></p>",
      "rawMarkdown": "How do we identify these periods with suspiciously low variation in accelerometer signals. Is there any threshold?\nHas anyone created notebook which handles this?\n\n[Though each series is a continuous recording, there may be periods in the series when the accelerometer device was removed. These period are determined as those where suspiciously little variation in the accelerometer signals occur over an extended period of time, which is unrealistic for typical human participants. Events are not annotated for these periods, and you should attempt to refrain from making event predictions during these periods: an event prediction will be scored as false positive.](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/data)",
      "votes": null
    },
    {
      "id": "2445284",
      "postDate": "09/18/2023 17:54:07",
      "content": "<p>Hi Rohit,</p>\n<p>some time ago, my mother made a Holter exam and she removed the parts of the devices that were attached on her belly.  The doctors didn't understand the results. I had to clarify that she didn't \"die\" for some minutes then come back cause she removed the \"thing\".</p>\n<p>The point is, whenever we make Health studies, things aren't totally under control. And professionals should be prepared because it isn't like on exact subjects. </p>\n<p>Another point to keep attention is Patients lie about what happen during the observations (here, the sleeping hours).  </p>\n<p>And not to mention that we are dealing with children: \" the unique needs of each child.\"</p>\n<p>However, I don't know how to proceed to avoid making predictions during those periods, as it was pointed out on the competition.</p>",
      "rawMarkdown": "Hi Rohit,\n\nsome time ago, my mother made a Holter exam and she removed the parts of the devices that were attached on her belly.  The doctors didn't understand the results. I had to clarify that she didn't \"die\" for some minutes then come back cause she removed the \"thing\".\n\nThe point is, whenever we make Health studies, things aren't totally under control. And professionals should be prepared because it isn't like on exact subjects. \n\nAnother point to keep attention is Patients lie about what happen during the observations (here, the sleeping hours).  \n\nAnd not to mention that we are dealing with children: \" the unique needs of each child.\"\n\nHowever, I don't know how to proceed to avoid making predictions during those periods, as it was pointed out on the competition.",
      "votes": null
    },
    {
      "id": "2450663",
      "postDate": "09/22/2023 03:45:10",
      "content": "<p><a href=\"https://www.kaggle.com/ryanholbrook\" target=\"_blank\">@ryanholbrook</a> ,<br>\nCan you throw some light on this (How low is suspiciously low for variation in accelerometer reading ?)  and any example series in dataset where we have accelerator device removed?</p>",
      "rawMarkdown": "ryanholbrook ,\nCan you throw some light on this (How low is suspiciously low for variation in accelerometer reading ?)  and any example series in dataset where we have accelerator device removed?",
      "votes": null
    },
    {
      "id": "2450829",
      "postDate": "09/22/2023 06:28:24",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/rohitchaudhari25\" target=\"_blank\">@rohitchaudhari25</a>. I haven't submitted it yet, but I think it's not that difficult to detect clear trends that define events. However, it is possible that subtle trend judgments such as annotated events for which the event itself cannot be defined may lead to an improvement in the score. Good luck.</p>",
      "rawMarkdown": "Hi @rohitchaudhari25. I haven't submitted it yet, but I think it's not that difficult to detect clear trends that define events. However, it is possible that subtle trend judgments such as annotated events for which the event itself cannot be defined may lead to an improvement in the score. Good luck.",
      "votes": null
    },
    {
      "id": "2450834",
      "postDate": "09/22/2023 06:30:17",
      "content": "<p>Dear <a href=\"https://www.kaggle.com/rohitchaudhari25\" target=\"_blank\">@rohitchaudhari25</a> </p>\n<p>From the <a href=\"https://www.nature.com/articles/s41598-020-79217-x.pdf\" target=\"_blank\">paper</a>:</p>\n<blockquote>\n  <p>\"<strong><em>Nonwear detection</em></strong>. <em>Periods of nonwear less than 30 min will go undetected with available heuristic approaches. We investigated whether nonwear periods can be determined at a higher resolution with machine learning (random forests). The ground truth labels for our nonwear classification are defined based on two assumptions: The accelerometer is worn during the PSG recording as prescribed by the study protocol and supervised by the researcher, and the accelerometer is not worn outside the PSG recording, according to the study protocol. Only if the standard deviation in the acceleration signal per 15 min is larger than 13.0 mg (1 mg = 0.00981 m/s2 ) these 15 min outside the PSG recording are labelled as wear. Here, the threshold of 13.0 mg is borrowed from the Heuristic van Hees approach.</em>\"</p>\n</blockquote>\n<p>All the best,<br>\ncarl</p>",
      "rawMarkdown": "Dear @rohitchaudhari25 \n\nFrom the [paper](https://www.nature.com/articles/s41598-020-79217-x.pdf):\n\n> \"***Nonwear detection***. *Periods of nonwear less than 30 min will go undetected with available heuristic approaches. We investigated whether nonwear periods can be determined at a higher resolution with machine learning (random forests). The ground truth labels for our nonwear classification are defined based on two assumptions: The accelerometer is worn during the PSG recording as prescribed by the study protocol and supervised by the researcher, and the accelerometer is not worn outside the PSG recording, according to the study protocol. Only if the standard deviation in the acceleration signal per 15 min is larger than 13.0 mg (1 mg = 0.00981 m/s2 ) these 15 min outside the PSG recording are labelled as wear. Here, the threshold of 13.0 mg is borrowed from the Heuristic van Hees approach.*\"\n\nAll the best,\ncarl",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2445284,
      "author_name": "mpwolke",
      "author_url": "",
      "post_date": "09/18/2023 17:54:07",
      "content": "<p>Hi Rohit,</p>\n<p>some time ago, my mother made a Holter exam and she removed the parts of the devices that were attached on her belly.  The doctors didn't understand the results. I had to clarify that she didn't \"die\" for some minutes then come back cause she removed the \"thing\".</p>\n<p>The point is, whenever we make Health studies, things aren't totally under control. And professionals should be prepared because it isn't like on exact subjects. </p>\n<p>Another point to keep attention is Patients lie about what happen during the observations (here, the sleeping hours).  </p>\n<p>And not to mention that we are dealing with children: \" the unique needs of each child.\"</p>\n<p>However, I don't know how to proceed to avoid making predictions during those periods, as it was pointed out on the competition.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2450663,
      "author_name": "rohitchaudhari25",
      "author_url": "",
      "post_date": "09/22/2023 03:45:10",
      "content": "<p><a href=\"https://www.kaggle.com/ryanholbrook\" target=\"_blank\">@ryanholbrook</a> ,<br>\nCan you throw some light on this (How low is suspiciously low for variation in accelerometer reading ?)  and any example series in dataset where we have accelerator device removed?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2450829,
      "author_name": "mitsuyasuhoshino",
      "author_url": "",
      "post_date": "09/22/2023 06:28:24",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/rohitchaudhari25\" target=\"_blank\">@rohitchaudhari25</a>. I haven't submitted it yet, but I think it's not that difficult to detect clear trends that define events. However, it is possible that subtle trend judgments such as annotated events for which the event itself cannot be defined may lead to an improvement in the score. Good luck.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2450834,
      "author_name": "carlmcbrideellis",
      "author_url": "",
      "post_date": "09/22/2023 06:30:17",
      "content": "<p>Dear <a href=\"https://www.kaggle.com/rohitchaudhari25\" target=\"_blank\">@rohitchaudhari25</a> </p>\n<p>From the <a href=\"https://www.nature.com/articles/s41598-020-79217-x.pdf\" target=\"_blank\">paper</a>:</p>\n<blockquote>\n  <p>\"<strong><em>Nonwear detection</em></strong>. <em>Periods of nonwear less than 30 min will go undetected with available heuristic approaches. We investigated whether nonwear periods can be determined at a higher resolution with machine learning (random forests). The ground truth labels for our nonwear classification are defined based on two assumptions: The accelerometer is worn during the PSG recording as prescribed by the study protocol and supervised by the researcher, and the accelerometer is not worn outside the PSG recording, according to the study protocol. Only if the standard deviation in the acceleration signal per 15 min is larger than 13.0 mg (1 mg = 0.00981 m/s2 ) these 15 min outside the PSG recording are labelled as wear. Here, the threshold of 13.0 mg is borrowed from the Heuristic van Hees approach.</em>\"</p>\n</blockquote>\n<p>All the best,<br>\ncarl</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2444499": "How do we identify these periods with suspiciously low variation in accelerometer signals. Is there any threshold?\nHas anyone created notebook which handles this?\n\n[Though each series is a continuous recording, there may be periods in the series when the accelerometer device was removed. These period are determined as those where suspiciously little variation in the accelerometer signals occur over an extended period of time, which is unrealistic for typical human participants. Events are not annotated for these periods, and you should attempt to refrain from making event predictions during these periods: an event prediction will be scored as false positive.](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/data)",
    "2445284": "Hi Rohit,\n\nsome time ago, my mother made a Holter exam and she removed the parts of the devices that were attached on her belly.  The doctors didn't understand the results. I had to clarify that she didn't \"die\" for some minutes then come back cause she removed the \"thing\".\n\nThe point is, whenever we make Health studies, things aren't totally under control. And professionals should be prepared because it isn't like on exact subjects. \n\nAnother point to keep attention is Patients lie about what happen during the observations (here, the sleeping hours).  \n\nAnd not to mention that we are dealing with children: \" the unique needs of each child.\"\n\nHowever, I don't know how to proceed to avoid making predictions during those periods, as it was pointed out on the competition.",
    "2450663": "ryanholbrook ,\nCan you throw some light on this (How low is suspiciously low for variation in accelerometer reading ?)  and any example series in dataset where we have accelerator device removed?",
    "2450829": "Hi @rohitchaudhari25. I haven't submitted it yet, but I think it's not that difficult to detect clear trends that define events. However, it is possible that subtle trend judgments such as annotated events for which the event itself cannot be defined may lead to an improvement in the score. Good luck.",
    "2450834": "Dear @rohitchaudhari25 \n\nFrom the [paper](https://www.nature.com/articles/s41598-020-79217-x.pdf):\n\n> \"***Nonwear detection***. *Periods of nonwear less than 30 min will go undetected with available heuristic approaches. We investigated whether nonwear periods can be determined at a higher resolution with machine learning (random forests). The ground truth labels for our nonwear classification are defined based on two assumptions: The accelerometer is worn during the PSG recording as prescribed by the study protocol and supervised by the researcher, and the accelerometer is not worn outside the PSG recording, according to the study protocol. Only if the standard deviation in the acceleration signal per 15 min is larger than 13.0 mg (1 mg = 0.00981 m/s2 ) these 15 min outside the PSG recording are labelled as wear. Here, the threshold of 13.0 mg is borrowed from the Heuristic van Hees approach.*\"\n\nAll the best,\ncarl"
  },
  "source": "meta"
}