Model_testing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Note: \n",
    "- This code is to perform image processing including registration and N4 bias correction via Nipype, skull-stripping, automated tumor segmentation, extraction of tumor loci and shape features, normalization and IDH status prediction which is the final step of hybrid model. \n",
    "\n",
    "<img src=\"workflow.png\">\n",
    "  \n",
    "- Python module requirements  : Nipype /  FSL /  ANTs / PyRadiomics / PyTorch\n",
    "- The process resquires GPU."
   ]
  },
  {
   "source": [
    "#### Use this to load a dumped session. (See last cell to dump session)"
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "#import dill\n",
    "#dill.load_session('notebook_env.db')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Image processing - skullstripping, resampling, registration and bias correction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from img_processing import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # file path of original T1C \n",
    "T1C_original_file = 'T1C.nii.gz'            \n",
    "T2_original_file = 'T2.nii.gz'\n",
    "FLAIR_original_file = 'FLAIR.nii.gz'\n",
    "    \n",
    "    # file path of skull-stripped images\n",
    "T1C_bet_file = 't1c_bet.nii.gz'            \n",
    "T2_bet_file = 't2_bet.nii.gz'\n",
    "FLAIR_bet_file = 'flair_bet.nii.gz'\n",
    "    \n",
    "    # file path of the mask for T1C skull stripping\n",
    "brainmask_T1C_file = 'T1C_bet0_temp_mask.nii.gz'  \n",
    "\n",
    "# filenames to save isovoxel images / brain mask\n",
    "T1C_iso_file = 't1c_isovoxel.nii.gz'\n",
    "T2_iso_file = 't2_isovoxel.nii.gz'\n",
    "FLAIR_iso_file = 'flair_isovoxel.nii.gz'\n",
    "brainmask_iso_file = 'mask_brain_isovoxel.nii.gz'    \n",
    "\n",
    "# filenames to save bias-corrected images   \n",
    "T1C_corrected_file = 't1c_corrected.nii.gz'\n",
    "T2_corrected_file = 't2_corrected.nii.gz'\n",
    "FLAIR_corrected_file = 'flair_corrected.nii.gz'\n",
    "\n",
    "## filename for preliminary skull-stripped T1C                                       \n",
    "T1C_bet_temp_file = 'T1C_bet0_temp.nii.gz'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "func_img_proc(T1C_original_file, T2_original_file, FLAIR_original_file,\n",
    "              T1C_bet_file, T2_bet_file, FLAIR_bet_file, brainmask_T1C_file,\n",
    "              T1C_iso_file, T2_iso_file, FLAIR_iso_file, brainmask_iso_file,\n",
    "              T1C_corrected_file, T2_corrected_file, FLAIR_corrected_file,\n",
    "              T1C_bet_temp_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model 1 : Automatic tumor segmentation  \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing for Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "(t1c_unet_arr, flair_unet_arr, cropdown_info) = func_norm_model1(T1C_corrected_file, FLAIR_corrected_file, brainmask_iso_file)\n",
    "\n",
    "# cropdown_info will be used for resmampling the predicted tumor mask to original isovoxel space, and preprcessing for Model2.\n",
    "cropdown_info"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Get tumor mask from Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "\n",
    "predmask_arr = func_get_predmask(t1c_unet_arr, flair_unet_arr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Resample the predicted mask back to original isovoxel space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "predmask_isovoxel_arr = func_mask_back2iso(predmask_arr, cropdown_info)\n",
    "predmask_isovoxel_arr_sitk = np.transpose(predmask_isovoxel_arr, (2,1,0))\n",
    "predmask_isovoxel_img = sitk.GetImageFromArray(predmask_isovoxel_arr_sitk)\n",
    "\n",
    "predmask_isovoxel_file = 'predmask_isovoxel.nii.gz' #filename for predicted mask of isovoxel resolution\n",
    "sitk.WriteImage(predmask_isovoxel_img, predmask_isovoxel_file)   # save the automatic segmentation of isovoxel resolution"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model 2 : CNN classifier for IDH status prediction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing for Model 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "t1c_corrected_img = nb.load(T1C_corrected_file)\n",
    "t1c_corrected_arr = t1c_corrected_img.get_data()\n",
    "t2_corrected_img = nb.load(T2_corrected_file)\n",
    "t2_corrected_arr = t2_corrected_img.get_data()\n",
    "brain_mask = nb.load(brainmask_iso_file)\n",
    "brain_mask_arr = brain_mask.get_data()\n",
    "\n",
    "t1c_resnet_arr = func_norm_resnet(t1c_corrected_arr, predmask_isovoxel_arr, brain_mask_arr, cropdown_info)\n",
    "t2_resnet_arr = func_norm_resnet(t2_corrected_arr, predmask_isovoxel_arr, brain_mask_arr, cropdown_info)\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Get shape and loci features from tumor mask of 1mm isovoxel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sla_features = func_shapeloci(T1C_iso_file, predmask_isovoxel_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Add patient's age"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sla_features['age'] = pd.Series(50)  ## change 50 to the patient's age."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df = sla_features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Normalize features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sla_features_mean = pd.read_csv('sla_features_mean.csv')\n",
    "sla_features_norm = sla_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#x-mean\n",
    "for i in range(len(sla_features.columns)):\n",
    "    sla_features_norm.iloc[0, i] = sla_features.iloc[0, i] - sla_features_mean.iloc[0, i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#show normalized features\n",
    "sla_features_norm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sla_features_norm = np.array(sla_features_norm)\n",
    "sla_arr = np.repeat(sla_features_norm, 5, axis=0)\n",
    "sla_arr.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Get probability of IDH mutation from Model 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#download Model 2\n",
    "from torch.hub import download_url_to_file\n",
    "\n",
    "url='https://github.com/yoonchoi-neuro/automated_hybrid_IDH/releases/download/final/MODEL2_CNNclassifier.pth'\n",
    "download_url_to_file(url, 'MODEL2_CNNclassifier.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_mean = get_IDH_pred(t1c_resnet_arr, t2_resnet_arr, predmask_arr, sla_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Dump current session so you can reload it later without needing to run everything again. func_img_proc() takes quite a lot of time. It might be wise to run this cell immediately after that function ends.\n",
    "import dill\n",
    "dill.dump_session('notebook_env.db')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}