Clinical Grade Sleep Tracking from Heart Rate using Z3Score-HRV API

z3score-hrv.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "import requests\n",
    "import patoolib\n",
    "import pyedflib\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from requests import post\n",
    "from pyndf import create_ndf_from_ecg\n",
    "from sklearn.metrics import cohen_kappa_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# lets download the ISRUC subgroup 3 dataset, comprises of 10 healthy subjects\n",
    "# You can use multi-threading to speed this up, but I had all night to download!\n",
    "for i in range(10):\n",
    "    url = \"http://dataset.isr.uc.pt/ISRUC_Sleep/subgroupIII/%d.rar\" %(i+1)\n",
    "    response = requests.get(url, stream=True)\n",
    "    filename = str(i+1) + \".rar\"\n",
    "    print(\"Downloading %s\" %(filename))\n",
    "    with open(filename, \"wb\") as handle:\n",
    "        for data in tqdm(response.iter_content()):\n",
    "            handle.write(data)\n",
    "    print(\"unzipping...\")\n",
    "    patoolib.extract_archive(filename, outdir=\".\")\n",
    "    print('removing the zip file')\n",
    "    os.remove(str(i+1) + \".rar\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# license key\n",
    "serverURL = 'https://z3score.com/api/v4'\n",
    "email = r'[email protected]'\n",
    "key = r'yourAccessKey'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": "Scoring: 1\\1.rec\nScoring: 2\\2.rec\nScoring: 3\\3.rec\nScoring: 4\\4.rec\nScoring: 5\\5.rec\nScoring: 6\\6.rec\nScoring: 7\\7.rec\nScoring: 8\\8.rec\nScoring: 9\\9.rec\nScoring: 10\\10.rec\n"
    }
   ],
   "source": [
    "# Now lets score all the files\n",
    "for i in range(1,11):\n",
    "    filename = '%d\\\\%d.rec' %(i,i)\n",
    "    print('Scoring: %s' %(filename))\n",
    "    edf_file = pyedflib.EdfReader(filename)\n",
    "    labels = edf_file.getSignalLabels()\n",
    "    try:\n",
    "        idx = labels.index('X2')\n",
    "    except ValueError:\n",
    "        print('ECG channel not found')\n",
    "        continue\n",
    "\n",
    "    fs = edf_file.getSampleFrequency(idx)\n",
    "    ecg = np.asarray(edf_file.readSignal(idx))*10\n",
    "    edf_file._close()\n",
    "    stream = create_ndf_from_ecg(ecg, fs)\n",
    "    response = post(serverURL+'/hrv', files={'file': ('stream.ndf', stream)}, data={'email':email, 'key':key})\n",
    "    data = response.json()\n",
    "    auto_scores = np.array(data['message'])\n",
    "    savefile = '%d\\\\%d_auto.csv' %(i,i)\n",
    "    np.savetxt(savefile, auto_scores, fmt=\"%d\", delimiter=',')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compare accuracy\n",
    "# Wake, Light Sleep, Deep Sleep & REM\n",
    "df = pd.DataFrame(columns=['Accuracy 1','Accuracy 2','kappa 1', 'kappa 2'])\n",
    "for i in range(1,11):\n",
    "    expert1 = '%d\\\\%d_1.txt' %(i,i)\n",
    "    expert2 = '%d\\\\%d_2.txt' %(i,i)\n",
    "    auto = '%d\\\\%d_auto.csv' %(i,i)\n",
    "\n",
    "    exp_score1 = np.genfromtxt(expert1, delimiter=',')\n",
    "    exp_score2 = np.genfromtxt(expert2, delimiter=',')\n",
    "    auto_score = np.genfromtxt(auto, delimiter=',')\n",
    "\n",
    "    num_epochs = min(auto_score.size, exp_score1.size, exp_score2.size)\n",
    "\n",
    "    auto_score = auto_score[0:num_epochs,0]\n",
    "    exp_score1 = exp_score1[0:num_epochs]\n",
    "    exp_score2 = exp_score2[0:num_epochs]\n",
    "    exp_score1[exp_score1==1] = 2\n",
    "    exp_score2[exp_score2==1] = 2\n",
    "\n",
    "    accuracy1 = np.sum(exp_score1 == auto_score)*100.0/num_epochs\n",
    "    accuracy2 = np.sum(exp_score2 == auto_score)*100.0/num_epochs\n",
    "    kappa1 = cohen_kappa_score(exp_score1, auto_score)\n",
    "    kappa2 = cohen_kappa_score(exp_score2, auto_score)\n",
    "    df = df.append({'Accuracy 1':accuracy1, 'Accuracy 2':accuracy2, 'kappa 1':kappa1, 'kappa 2':kappa2}, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Accuracy 1    70.967779\nAccuracy 2    72.601184\nkappa 1        0.556185\nkappa 2        0.579426\ndtype: float64"
     },
     "metadata": {},
     "execution_count": 120
    }
   ],
   "source": [
    "df.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compare accuracy\n",
    "# Wake, NREM & REM\n",
    "df = pd.DataFrame(columns=['Accuracy 1','Accuracy 2','kappa 1', 'kappa 2'])\n",
    "for i in range(1,11):\n",
    "    expert1 = '%d\\\\%d_1.txt' %(i,i)\n",
    "    expert2 = '%d\\\\%d_2.txt' %(i,i)\n",
    "    auto = '%d\\\\%d_auto.csv' %(i,i)\n",
    "\n",
    "    exp_score1 = np.genfromtxt(expert1, delimiter=',')\n",
    "    exp_score2 = np.genfromtxt(expert2, delimiter=',')\n",
    "    auto_score = np.genfromtxt(auto, delimiter=',')\n",
    "\n",
    "    num_epochs = min(auto_score.size, exp_score1.size, exp_score2.size)\n",
    "\n",
    "    auto_score = auto_score[0:num_epochs,0]\n",
    "    exp_score1 = exp_score1[0:num_epochs]\n",
    "    exp_score2 = exp_score2[0:num_epochs]\n",
    "    exp_score1[exp_score1==1] = 2\n",
    "    exp_score2[exp_score2==1] = 2\n",
    "    exp_score1[exp_score1==3] = 2\n",
    "    exp_score2[exp_score2==3] = 2\n",
    "    auto_score[auto_score==3] = 2\n",
    "\n",
    "    accuracy1 = np.sum(exp_score1 == auto_score)*100.0/num_epochs\n",
    "    accuracy2 = np.sum(exp_score2 == auto_score)*100.0/num_epochs\n",
    "    kappa1 = cohen_kappa_score(exp_score1, auto_score)\n",
    "    kappa2 = cohen_kappa_score(exp_score2, auto_score)\n",
    "    df = df.append({'Accuracy 1':accuracy1, 'Accuracy 2':accuracy2, 'kappa 1':kappa1, 'kappa 2':kappa2}, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Accuracy 1    82.520431\nAccuracy 2    83.576178\nkappa 1        0.608654\nkappa 2        0.631983\ndtype: float64"
     },
     "metadata": {},
     "execution_count": 122
    }
   ],
   "source": [
    "df.mean()"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.10-final"
  },
  "orig_nbformat": 2,
  "kernelspec": {
   "name": "python361064bitdevenvvenvcd022a983f794dd1941c299f2f977cbf",
   "display_name": "Python 3.6.10 64-bit ('devenv': venv)"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}