{"cells":[{"metadata":{},"cell_type":"markdown","source":"# An Alternative Version of Creating Our TFRecords Datasets\nThere are already some notebooks about creating tfrecords in this competition: <a href='https://www.kaggle.com/teeyee314/pulmonary-embolism-create-tfrecords/'>by Tim Yee</a> and <a href='https://www.kaggle.com/marcosnovaes/building-a-tfrecord-dataset/'>by Marcos Novaes</a>. Refer to <a href='https://www.kaggle.com/cdeotte/how-to-create-tfrecords/'>Chris Deotte's introduction noteboook</a>, I also create a version where we can extract DICOM information by ourselves. For example, `Image Position (Patient)` helps us to sort the order of sliced images within a study so that we could use `Conv3D` or `Conv1D/RNN` in our models. We process the images in examwise. In each `*.tfrec` file, there are fixed number (my setting is 145) of exams.\nRefer to <a href='https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/discussion/182930'>Ian Pan's medical insights</a>, we use three windows as suggested.\n\n> RED channel / LUNG window / level=-600, width=1500  \n> GREEN channel / PE window / level=100, width=700  \n> BLUE channel / MEDIASTINAL window / level=40, width=400  \n\n**Note**: This version is only a QuickSave Version. Processing all training data will cost about 145 hours on Kaggle CPU. This output is just a peek and not enough to train on your model."},{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"import numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nimport os\nimport glob\nimport tensorflow as tf\nimport re, math\nimport matplotlib.pyplot as plt","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"os.system('cp ../input/gdcm-conda-install/gdcm.tar .')\nos.system('tar -xvzf gdcm.tar')\nos.system('conda install --offline ./gdcm/gdcm-2.8.9-py37h71b2a6d_0.tar.bz2')\nprint(\"GDCM Loaded!\")\nimport pydicom\nimport cv2","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"os.system('rm gdcm.tar')\nos.system('rm -r gdcm')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train = pd.read_csv(\"../input/rsna-str-pulmonary-embolism-detection/train.csv\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# PATHS TO IMAGES\nPATH_TRAIN = '../input/rsna-str-pulmonary-embolism-detection/train'\nPATH_TEST = '../input/rsna-str-pulmonary-embolism-detection/test'","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"case_train = glob.glob(PATH_TRAIN + '/*/*')\nprint('Total number of train cases: ', len(case_train))\ncase_test = glob.glob(PATH_TEST + '/*/*')\nprint('Total number of test cases: ', len(case_test))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def transform_to_hu(medical_image, image):\n    intercept = medical_image.RescaleIntercept\n    slope = medical_image.RescaleSlope\n    hu_image = image * slope + intercept\n    return hu_image\n\ndef window(img, WL=50, WW=350):\n    upper, lower = WL+WW//2, WL-WW//2\n    X = np.clip(img.copy(), lower, upper)\n    X = X - np.min(X)\n    X = X / np.max(X)\n    X = (X*255.0).astype('uint8')\n    return X\n\ndef convert_to_rgb(array):\n    shape_gray = array.shape\n    R_lung_window = window(array, -600, 1500)\n    G_pe_window = window(array, 100, 700)\n    B_mediastinal_window = window(array, 40, 400)\n    return np.stack([R_lung_window, G_pe_window, B_mediastinal_window], axis=2).reshape(shape_gray + (3,))","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-input":true,"trusted":true},"cell_type":"code","source":"# def get_imgs_by_case(path):\n#     img_path = glob.glob(path + '/*')\n    \n#     img_set = []\n#     z_set = []\n#     sop_set = []\n    \n#     for p in img_path:\n#         med_img = pydicom.dcmread(p)\n#         img = med_img.pixel_array\n#         img = transform_to_hu(med_img, img)\n#         img = convert_to_rgb(img)\n        \n#         img_set = np.append([img_set, img])\n#         z_set = np.append([z_set, float(med_img.ImagePositionPatient[-1])])\n#         sop_set = np.append([sop_set, p.split('/')[-1].split('.')[0]])\n    \n#     return img_set, z_set, sop_set","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def get_img(path):    \n    med_img = pydicom.dcmread(path)\n    img = med_img.pixel_array\n    img = transform_to_hu(med_img, img)\n    img = convert_to_rgb(img)\n\n    pos_z = float(med_img.ImagePositionPatient[-1])\n    sop = path.split('/')[-1].split('.')[0]\n    \n    return img, pos_z, sop","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"def _bytes_feature(value):\n    \"\"\"Returns a bytes_list from a string / byte.\"\"\"\n    if isinstance(value, type(tf.constant(0))):\n        value = value.numpy() # BytesList won't unpack a string from an EagerTensor.\n    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))\n\ndef _float_feature(value):\n    \"\"\"Returns a float_list from a float / double.\"\"\"\n    return tf.train.Feature(float_list=tf.train.FloatList(value=[value]))\n\ndef _int64_feature(value):\n    \"\"\"Returns an int64_list from a bool / enum / int / uint.\"\"\"\n    return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def serialize_example(feature0, feature1, feature2, feature3):\n    feature = {\n      'image': _bytes_feature(feature0),\n      'image_id': _bytes_feature(feature1),\n      'position_z': _float_feature(feature2),\n      'target': _bytes_feature(feature3)\n    }\n    example_proto = tf.train.Example(features=tf.train.Features(feature=feature))\n    return example_proto.SerializeToString()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"selected_cols = ['pe_present_on_image',\n                 'negative_exam_for_pe', 'rv_lv_ratio_gte_1', 'rv_lv_ratio_lt_1',\n                 'leftsided_pe', 'chronic_pe', 'rightsided_pe',\n                 'acute_and_chronic_pe', 'central_pe', 'indeterminate']","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"BATCH = 51\nSIZE = len(case_train)//(BATCH-1)\nSIZE","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"PATH_SET = [[] for i in range(BATCH)]\nm = 0\nfor n, p in enumerate(case_train):\n    img_path = glob.glob(p + '/*')\n    if (n+1) % SIZE == 0:\n        m += 1\n    PATH_SET[m] = np.append(PATH_SET[m], img_path)\n\nPATH_SET = np.array(PATH_SET)\nprint(PATH_SET.shape, PATH_SET[0].shape)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### 5min/1000imgs => about 145 hours to process."},{"metadata":{"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"for j in range(BATCH):\n    print(); print('Writing TFRecord %i of %i...'%(j,BATCH-1))\n    with tf.io.TFRecordWriter('train%.2i_%i.tfrec'%(j, PATH_SET[j].shape[0])) as writer:\n        for k in range(PATH_SET[j].shape[0]):\n            img, pos_z, name = get_img(PATH_SET[j][k])\n            img = cv2.imencode('.jpg', img, (cv2.IMWRITE_JPEG_QUALITY, 94))[1].tostring()\n            target = train[train['SOPInstanceUID'] == name][selected_cols].values[0]\n            example = serialize_example(\n                img, str.encode(name), pos_z, \n                tf.io.serialize_tensor(np.array(target, dtype=np.uint8))\n            )\n            writer.write(example)\n            if k%1000==0: print(k,', ',end='')","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"np.set_printoptions(threshold=15, linewidth=80)\nCLASSES = [0,1]\n\ndef batch_to_numpy_images_and_labels(data):\n    images, labels = data\n    numpy_images = images.numpy()\n    numpy_labels = labels.numpy()\n    #if numpy_labels.dtype == object: # binary string in this case, these are image ID strings\n    #    numpy_labels = [None for _ in enumerate(numpy_images)]\n    # If no labels, only image IDs, return None for labels (this is the case for test data)\n    return numpy_images, numpy_labels\n\ndef title_from_label_and_target(label, correct_label):\n    if correct_label is None:\n        return CLASSES[label], True\n    correct = (label == correct_label)\n    return \"{} [{}{}{}]\".format(CLASSES[label], 'OK' if correct else 'NO', u\"\\u2192\" if not correct else '',\n                                CLASSES[correct_label] if not correct else ''), correct\n\ndef display_one_flower(image, title, subplot, red=False, titlesize=16):\n    plt.subplot(*subplot)\n    plt.axis('off')\n    plt.imshow(image)\n    if len(title) > 0:\n        plt.title(title, fontsize=int(titlesize) if not red else int(titlesize/1.2), color='red' if red else 'black', fontdict={'verticalalignment':'center'}, pad=int(titlesize/1.5))\n    return (subplot[0], subplot[1], subplot[2]+1)\n    \ndef display_batch_of_images(databatch, predictions=None):\n    \"\"\"This will work with:\n    display_batch_of_images(images)\n    display_batch_of_images(images, predictions)\n    display_batch_of_images((images, labels))\n    display_batch_of_images((images, labels), predictions)\n    \"\"\"\n    # data\n    images, labels = batch_to_numpy_images_and_labels(databatch)\n    if labels is None:\n        labels = [None for _ in enumerate(images)]\n        \n    # auto-squaring: this will drop data that does not fit into square or square-ish rectangle\n    rows = int(math.sqrt(len(images)))\n    cols = len(images)//rows\n        \n    # size and spacing\n    FIGSIZE = 13.0\n    SPACING = 0.1\n    subplot=(rows,cols,1)\n    if rows < cols:\n        plt.figure(figsize=(FIGSIZE,FIGSIZE/cols*rows))\n    else:\n        plt.figure(figsize=(FIGSIZE/rows*cols,FIGSIZE))\n    \n    # display\n    for i, (image, label) in enumerate(zip(images[:rows*cols], labels[:rows*cols])):\n        title = label\n        correct = True\n        if predictions is not None:\n            title, correct = title_from_label_and_target(predictions[i], label)\n        dynamic_titlesize = FIGSIZE*SPACING/max(rows,cols)*40+3 # magic formula tested to work from 1x1 to 10x10 images\n        subplot = display_one_flower(image, title, subplot, not correct, titlesize=dynamic_titlesize)\n    \n    #layout\n    plt.tight_layout()\n    if label is None and predictions is None:\n        plt.subplots_adjust(wspace=0, hspace=0)\n    else:\n        plt.subplots_adjust(wspace=SPACING, hspace=SPACING)\n    plt.show()","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"def decode_image(image_data):\n    image = tf.image.decode_jpeg(image_data, channels=3)\n    image = tf.cast(image, tf.float32) / 255.0  # convert image to floats in [0, 1] range\n    image = tf.reshape(image, [*IMAGE_SIZE, 3]) # explicit size needed for TPU\n    return image\n\ndef read_labeled_tfrecord(example):\n    LABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string means bytestring\n        \"image_id\": tf.io.FixedLenFeature([], tf.string),  # shape [] means single element\n    }\n    example = tf.io.parse_single_example(example, LABELED_TFREC_FORMAT)\n    image = decode_image(example['image'])\n    label = example['image_id']\n    return image, label # returns a dataset of (image, label) pairs\n\ndef load_dataset(filenames, labeled=True, ordered=False):\n    # Read from TFRecords. For optimal performance, reading from multiple files at once and\n    # disregarding data order. Order does not matter since we will be shuffling the data anyway.\n\n    ignore_order = tf.data.Options()\n    if not ordered:\n        ignore_order.experimental_deterministic = False # disable order, increase speed\n\n    dataset = tf.data.TFRecordDataset(filenames, num_parallel_reads=AUTO) # automatically interleaves reads from multiple files\n    dataset = dataset.with_options(ignore_order) # uses data as soon as it streams in, rather than in its original order\n    dataset = dataset.map(read_labeled_tfrecord)\n    # returns a dataset of (image, label) pairs if labeled=True or (image, id) pairs if labeled=False\n    return dataset\n\ndef get_training_dataset():\n    dataset = load_dataset(TRAINING_FILENAMES, labeled=True)\n    dataset = dataset.repeat() # the training dataset must repeat for several epochs\n    dataset = dataset.shuffle(2048)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef count_data_items(filenames):\n    # the number of data items is written in the name of the .tfrec files, i.e. flowers00-230.tfrec = 230 data items\n    n = [int(re.compile(r\"_([0-9]*)\\.\").search(filename).group(1)) for filename in filenames]\n    return np.sum(n)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"IMAGE_SIZE= [512,512]; BATCH_SIZE = 32\nAUTO = tf.data.experimental.AUTOTUNE\nTRAINING_FILENAMES = tf.io.gfile.glob('train*.tfrec')\nprint('There are %i train images'%count_data_items(TRAINING_FILENAMES))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"training_dataset = get_training_dataset()\ntraining_dataset = training_dataset.unbatch().batch(20)\ntrain_batch = iter(training_dataset)\n\ndisplay_batch_of_images(next(train_batch))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}