best-cnn-segmentation-connected-components-384x384.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "d320f90f432c331afea02ef66fcddb59448ea200"
   },
   "source": [
    "# Approach\n",
    "\n",
    "* Firstly a convolutional neural network is used to segment the image, using the bounding boxes directly as a mask. \n",
    "* Secondly connected components is used to separate multiple areas of predicted pneumonia.\n",
    "* Finally a bounding box is simply drawn around every connected component.\n",
    "\n",
    "# Network\n",
    "\n",
    "* The network consists of a number of residual blocks with convolutions and downsampling blocks with max pooling.\n",
    "* At the end of the network a single upsampling layer converts the output to the same shape as the input.\n",
    "\n",
    "As the input to the network is 256 by 256 (instead of the original 1024 by 1024) and the network downsamples a number of times without any meaningful upsampling (the final upsampling is just to match in 256 by 256 mask) the final prediction is very crude. If the network downsamples 4 times the final bounding boxes can only change with at least 16 pixels.\n",
    "\n",
    "**Edit by EAS** - Change input image size to 320x320."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import csv\n",
    "import random\n",
    "import pydicom\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from skimage import measure\n",
    "from skimage.transform import resize\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "8d58cefae21c951e077c02c1a989d020f18465cc"
   },
   "source": [
    "# Load pneumonia locations\n",
    "\n",
    "Table contains [filename : pneumonia location] pairs per row. \n",
    "* If a filename contains multiple pneumonia, the table contains multiple rows with the same filename but different pneumonia locations. \n",
    "* If a filename contains no pneumonia it contains a single row with an empty pneumonia location.\n",
    "\n",
    "The code below loads the table and transforms it into a dictionary. \n",
    "* The dictionary uses the filename as key and a list of pneumonia locations in that filename as value. \n",
    "* If a filename is not present in the dictionary it means that it contains no pneumonia."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "e08496a85ef9b0823595c3745d2677c6e84b6a3a",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# empty dictionary\n",
    "pneumonia_locations = {}\n",
    "# load table\n",
    "with open(os.path.join('../input/stage_1_train_labels.csv'), mode='r') as infile:\n",
    "    # open reader\n",
    "    reader = csv.reader(infile)\n",
    "    # skip header\n",
    "    next(reader, None)\n",
    "    # loop through rows\n",
    "    for rows in reader:\n",
    "        # retrieve information\n",
    "        filename = rows[0]\n",
    "        location = rows[1:5]\n",
    "        pneumonia = rows[5]\n",
    "        # if row contains pneumonia add label to dictionary\n",
    "        # which contains a list of pneumonia locations per filename\n",
    "        if pneumonia == '1':\n",
    "            # convert string to float to int\n",
    "            location = [int(float(i)) for i in location]\n",
    "            # save pneumonia location in dictionary\n",
    "            if filename in pneumonia_locations:\n",
    "                pneumonia_locations[filename].append(location)\n",
    "            else:\n",
    "                pneumonia_locations[filename] = [location]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "5a4fa6da9833fd7cbd476d19584ba35695b38ddb"
   },
   "source": [
    "# Load filenames"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "ccd0b0d52cafd125558ed5560a9cc8fa15760bc5",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# load and shuffle filenames\n",
    "folder = '../input/stage_1_train_images'\n",
    "filenames = os.listdir(folder)\n",
    "random.shuffle(filenames)\n",
    "# split into train and validation filenames\n",
    "n_valid_samples = 2560\n",
    "train_filenames = filenames[n_valid_samples:]\n",
    "valid_filenames = filenames[:n_valid_samples]\n",
    "print('n train samples', len(train_filenames))\n",
    "print('n valid samples', len(valid_filenames))\n",
    "n_train_samples = len(filenames) - n_valid_samples"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "bd0633867d9d32180eb776703205767e8e891c50"
   },
   "source": [
    "# Exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "daa7156380489227e510e8ef086b55c58b4b3ad8",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "print('Total train images:',len(filenames))\n",
    "print('Images with pneumonia:', len(pneumonia_locations))\n",
    "\n",
    "ns = [len(value) for value in pneumonia_locations.values()]\n",
    "plt.figure()\n",
    "plt.hist(ns)\n",
    "plt.xlabel('Pneumonia per image')\n",
    "plt.xticks(range(1, np.max(ns)+1))\n",
    "plt.show()\n",
    "\n",
    "heatmap = np.zeros((1024, 1024))\n",
    "ws = []\n",
    "hs = []\n",
    "for values in pneumonia_locations.values():\n",
    "    for value in values:\n",
    "        x, y, w, h = value\n",
    "        heatmap[y:y+h, x:x+w] += 1\n",
    "        ws.append(w)\n",
    "        hs.append(h)\n",
    "plt.figure()\n",
    "plt.title('Pneumonia location heatmap')\n",
    "plt.imshow(heatmap)\n",
    "plt.figure()\n",
    "plt.title('Pneumonia height lengths')\n",
    "plt.hist(hs, bins=np.linspace(0,1000,50))\n",
    "plt.show()\n",
    "plt.figure()\n",
    "plt.title('Pneumonia width lengths')\n",
    "plt.hist(ws, bins=np.linspace(0,1000,50))\n",
    "plt.show()\n",
    "print('Minimum pneumonia height:', np.min(hs))\n",
    "print('Minimum pneumonia width: ', np.min(ws))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "276b80f59fa9acdfd9307d74cee5f705cd6aa5b6",
    "collapsed": true
   },
   "source": [
    " # Data generator\n",
    "\n",
    "The dataset is too large to fit into memory, so we need to create a generator that loads data on the fly.\n",
    "\n",
    "* The generator takes in some filenames, batch_size and other parameters.\n",
    "\n",
    "* The generator outputs a random batch of numpy images and numpy masks.\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "86b3f780a03cddda78c6adfde461d6ff8dad5672",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class generator(keras.utils.Sequence):\n",
    "    \n",
    "    def __init__(self, folder, filenames, pneumonia_locations=None, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, augment=False, predict=False):\n",
    "        self.folder = folder\n",
    "        self.filenames = filenames\n",
    "        self.pneumonia_locations = pneumonia_locations\n",
    "        self.batch_size = batch_size\n",
    "        self.image_size = image_size\n",
    "        self.shuffle = shuffle\n",
    "        self.augment = augment\n",
    "        self.predict = predict\n",
    "        self.on_epoch_end()\n",
    "        \n",
    "    def __load__(self, filename):\n",
    "        # load dicom file as numpy array\n",
    "        img = pydicom.dcmread(os.path.join(self.folder, filename)).pixel_array\n",
    "        # create empty mask\n",
    "        msk = np.zeros(img.shape)\n",
    "        # get filename without extension\n",
    "        filename = filename.split('.')[0]\n",
    "        # if image contains pneumonia\n",
    "        if filename in pneumonia_locations:\n",
    "            # loop through pneumonia\n",
    "            for location in pneumonia_locations[filename]:\n",
    "                # add 1's at the location of the pneumonia\n",
    "                x, y, w, h = location\n",
    "                msk[y:y+h, x:x+w] = 1\n",
    "        # if augment then horizontal flip half the time\n",
    "        if self.augment and random.random() > 0.5:\n",
    "            img = np.fliplr(img)\n",
    "            msk = np.fliplr(msk)\n",
    "        # resize both image and mask\n",
    "        img = resize(img, (self.image_size, self.image_size), mode='reflect')\n",
    "        msk = resize(msk, (self.image_size, self.image_size), mode='reflect') > 0.5\n",
    "        # add trailing channel dimension\n",
    "        img = np.expand_dims(img, -1)\n",
    "        msk = np.expand_dims(msk, -1)\n",
    "        return img, msk\n",
    "    \n",
    "    def __loadpredict__(self, filename):\n",
    "        # load dicom file as numpy array\n",
    "        img = pydicom.dcmread(os.path.join(self.folder, filename)).pixel_array\n",
    "        # resize image\n",
    "        img = resize(img, (self.image_size, self.image_size), mode='reflect')\n",
    "        # add trailing channel dimension\n",
    "        img = np.expand_dims(img, -1)\n",
    "        return img\n",
    "        \n",
    "    def __getitem__(self, index):\n",
    "        # select batch\n",
    "        filenames = self.filenames[index*self.batch_size:(index+1)*self.batch_size]\n",
    "        # predict mode: return images and filenames\n",
    "        if self.predict:\n",
    "            # load files\n",
    "            imgs = [self.__loadpredict__(filename) for filename in filenames]\n",
    "            # create numpy batch\n",
    "            imgs = np.array(imgs)\n",
    "            return imgs, filenames\n",
    "        # train mode: return images and masks\n",
    "        else:\n",
    "            # load files\n",
    "            items = [self.__load__(filename) for filename in filenames]\n",
    "            # unzip images and masks\n",
    "            imgs, msks = zip(*items)\n",
    "            # create numpy batch\n",
    "            imgs = np.array(imgs)\n",
    "            msks = np.array(msks)\n",
    "            return imgs, msks\n",
    "        \n",
    "    def on_epoch_end(self):\n",
    "        if self.shuffle:\n",
    "            random.shuffle(self.filenames)\n",
    "        \n",
    "    def __len__(self):\n",
    "        if self.predict:\n",
    "            # return everything\n",
    "            return int(np.ceil(len(self.filenames) / self.batch_size))\n",
    "        else:\n",
    "            # return full batches only\n",
    "            return int(len(self.filenames) / self.batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "ebc622a4b406354cc4ef28801eab72346a724d8b"
   },
   "source": [
    "# Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGE_SIZE=320\n",
    "BATCH_SIZE=16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "9fc2b108689637a6037b48ebab3f7659b8704bf9",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_downsample(channels, inputs):\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(inputs)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(channels, 1, padding='same', use_bias=False)(x)\n",
    "    x = keras.layers.MaxPool2D(2)(x)\n",
    "    return x\n",
    "\n",
    "def create_resblock(channels, inputs):\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(inputs)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(x)\n",
    "    return keras.layers.add([x, inputs])\n",
    "\n",
    "def create_network(input_size, channels, n_blocks=2, depth=4):\n",
    "    # input\n",
    "    inputs = keras.Input(shape=(input_size, input_size, 1))\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(inputs)\n",
    "    # residual blocks\n",
    "    for d in range(depth):\n",
    "        channels = channels * 2\n",
    "        x = create_downsample(channels, x)\n",
    "        for b in range(n_blocks):\n",
    "            x = create_resblock(channels, x)\n",
    "    # output\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(1, 1, activation='sigmoid')(x)\n",
    "    outputs = keras.layers.UpSampling2D(2**depth)(x)\n",
    "    model = keras.Model(inputs=inputs, outputs=outputs)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "bee3bc9363e9c1829eadf17da7a67fbd1a6a369e"
   },
   "source": [
    "# Train network\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "4369be30f61440eb6858d57829fa541c4ee893bf",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# define iou or jaccard loss function\n",
    "def iou_loss(y_true, y_pred):\n",
    "    y_true = tf.reshape(y_true, [-1])\n",
    "    y_pred = tf.reshape(y_pred, [-1])\n",
    "    intersection = tf.reduce_sum(y_true * y_pred)\n",
    "    score = (intersection + 1.) / (tf.reduce_sum(y_true) + tf.reduce_sum(y_pred) - intersection + 1.)\n",
    "    return 1 - score\n",
    "\n",
    "# combine bce loss and iou loss\n",
    "def iou_bce_loss(y_true, y_pred):\n",
    "    return 0.5 * keras.losses.binary_crossentropy(y_true, y_pred) + 0.5 * iou_loss(y_true, y_pred)\n",
    "\n",
    "# mean iou as a metric\n",
    "def mean_iou(y_true, y_pred):\n",
    "    y_pred = tf.round(y_pred)\n",
    "    intersect = tf.reduce_sum(y_true * y_pred, axis=[1, 2, 3])\n",
    "    union = tf.reduce_sum(y_true, axis=[1, 2, 3]) + tf.reduce_sum(y_pred, axis=[1, 2, 3])\n",
    "    smooth = tf.ones(tf.shape(intersect))\n",
    "    return tf.reduce_mean((intersect + smooth) / (union - intersect + smooth))\n",
    "\n",
    "# create network and compiler\n",
    "model = create_network(input_size=IMAGE_SIZE, channels=32, n_blocks=2, depth=4)\n",
    "model.compile(optimizer='adam',\n",
    "              loss=iou_bce_loss,\n",
    "              metrics=['accuracy', mean_iou])\n",
    "\n",
    "# cosine learning rate annealing\n",
    "def cosine_annealing(x):\n",
    "    lr = 0.001\n",
    "    epochs = 20\n",
    "    return lr*(np.cos(np.pi*x/epochs)+1.)/2\n",
    "learning_rate = tf.keras.callbacks.LearningRateScheduler(cosine_annealing)\n",
    "\n",
    "# create train and validation generators\n",
    "folder = '../input/stage_1_train_images'\n",
    "train_gen = generator(folder, train_filenames, pneumonia_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, augment=True, predict=False)\n",
    "valid_gen = generator(folder, valid_filenames, pneumonia_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=False, predict=False)\n",
    "\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "f6bb70654592bc3b47f50f09982d42f4c6971463",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate], epochs=15, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "3666ba4cac9ed2c3029b824af220404bfcc16f23",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "plt.figure(figsize=(12,4))\n",
    "plt.subplot(131)\n",
    "plt.plot(history.epoch, history.history[\"loss\"], label=\"Train loss\")\n",
    "plt.plot(history.epoch, history.history[\"val_loss\"], label=\"Valid loss\")\n",
    "plt.legend()\n",
    "plt.subplot(132)\n",
    "plt.plot(history.epoch, history.history[\"acc\"], label=\"Train accuracy\")\n",
    "plt.plot(history.epoch, history.history[\"val_acc\"], label=\"Valid accuracy\")\n",
    "plt.legend()\n",
    "plt.subplot(133)\n",
    "plt.plot(history.epoch, history.history[\"mean_iou\"], label=\"Train iou\")\n",
    "plt.plot(history.epoch, history.history[\"val_mean_iou\"], label=\"Valid iou\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "25dc7ad848190bb37349a80f96ed2bdb5c3821b0"
   },
   "source": [
    "# Predict test images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "2c9277e4ec9f12712dd690002c540b396278c504",
    "collapsed": true,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# load and shuffle filenames\n",
    "folder = '../input/stage_1_test_images'\n",
    "test_filenames = os.listdir(folder)\n",
    "print('n test samples:', len(test_filenames))\n",
    "\n",
    "# create test generator with predict flag set to True\n",
    "test_gen = generator(folder, test_filenames, None, batch_size=25, image_size=IMAGE_SIZE, shuffle=False, predict=True)\n",
    "\n",
    "# create submission dictionary\n",
    "submission_dict = {}\n",
    "# loop through testset\n",
    "for imgs, filenames in test_gen:\n",
    "    # predict batch of images\n",
    "    preds = model.predict(imgs)\n",
    "    # loop through batch\n",
    "    for pred, filename in zip(preds, filenames):\n",
    "        # resize predicted mask\n",
    "        pred = resize(pred, (1024, 1024), mode='reflect')\n",
    "        # threshold predicted mask\n",
    "        comp = pred[:, :, 0] > 0.5\n",
    "        # apply connected components\n",
    "        comp = measure.label(comp)\n",
    "        # apply bounding boxes\n",
    "        predictionString = ''\n",
    "        for region in measure.regionprops(comp):\n",
    "            # retrieve x, y, height and width\n",
    "            y, x, y2, x2 = region.bbox\n",
    "            height = y2 - y\n",
    "            width = x2 - x\n",
    "            # proxy for confidence score\n",
    "            conf = np.mean(pred[y:y+height, x:x+width])\n",
    "            # add to predictionString\n",
    "            predictionString += str(conf) + ' ' + str(x) + ' ' + str(y) + ' ' + str(width) + ' ' + str(height) + ' '\n",
    "        # add filename and predictionString to dictionary\n",
    "        filename = filename.split('.')[0]\n",
    "        submission_dict[filename] = predictionString\n",
    "    # stop if we've got them all\n",
    "    if len(submission_dict) >= len(test_filenames):\n",
    "        break\n",
    "\n",
    "# save dictionary as csv file\n",
    "sub = pd.DataFrame.from_dict(submission_dict,orient='index')\n",
    "sub.index.names = ['patientId']\n",
    "sub.columns = ['PredictionString']\n",
    "sub.to_csv('submission.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_uuid": "6b7bbc372fddde87f95660b79482d58137ff4ff3",
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}