{"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_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"!pip install keras_cv_attention_models\n!pip install dicomsdl\n!pip install pylibjpeg\n!pip install python_gdcm","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-06-07T13:36:51.102353Z","iopub.execute_input":"2023-06-07T13:36:51.102727Z","iopub.status.idle":"2023-06-07T13:37:51.32535Z","shell.execute_reply.started":"2023-06-07T13:36:51.102695Z","shell.execute_reply":"2023-06-07T13:37:51.323947Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import pylibjpeg\nimport numpy as np\nimport pandas as pd\n\nimport pydicom\nimport matplotlib as mpl\nimport matplotlib.pyplot as plt\nimport tensorflow as tf\n\nfrom joblib import Parallel, delayed\nfrom tqdm.notebook import tqdm\nfrom multiprocessing import cpu_count\nfrom keras_cv_attention_models import convnext, efficientnet\n\nimport cv2\nimport glob\nimport importlib\nimport os\nimport joblib\nimport time\nimport dicomsdl\nimport gc\n","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:37:51.328152Z","iopub.execute_input":"2023-06-07T13:37:51.328703Z","iopub.status.idle":"2023-06-07T13:38:01.462496Z","shell.execute_reply.started":"2023-06-07T13:37:51.328643Z","shell.execute_reply":"2023-06-07T13:38:01.461423Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#import pylibjpegmodel","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.463683Z","iopub.execute_input":"2023-06-07T13:38:01.464378Z","iopub.status.idle":"2023-06-07T13:38:01.468688Z","shell.execute_reply.started":"2023-06-07T13:38:01.464344Z","shell.execute_reply":"2023-06-07T13:38:01.467818Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def crop(img , coord = False):\n\n    # Otsu's thresholding after Gaussian filtering\n    blur = cv2.GaussianBlur(img,(5,5),0)\n    non_nuls = img[img>=20]\n    h,w = img.shape\n    if(len(non_nuls)/(h*w) >0.7):\n        _, breast_mask = cv2.threshold(blur,np.percentile(img[img>=20], 35),255,cv2.THRESH_BINARY)\n    else:\n        _, breast_mask = cv2.threshold(blur,np.percentile(img[img>=20], 2),255,cv2.THRESH_BINARY)\n    \n    \n    \n    binarized = breast_mask\n    kernel = np.ones((20,20),np.uint8)\n    breast_mask = cv2.morphologyEx(breast_mask, cv2.MORPH_OPEN, kernel)\n    opened_binarized = breast_mask\n    kernel = np.ones((60,60),np.uint8)\n    breast_mask = cv2.morphologyEx(breast_mask, cv2.MORPH_CLOSE, kernel)\n    closed_binarized = breast_mask\n    cnts, _ = cv2.findContours(breast_mask.astype(np.uint8), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)\n    cnt = max(cnts, key = cv2.contourArea)\n    \n    countour = np.zeros(img.shape, np.uint8)\n    cv2.drawContours(countour, [cnt], -1, 255, cv2.FILLED)\n    breast_mask = countour\n    \n    \n    x, y, w, h = cv2.boundingRect(cnt)\n    if coord: \n        return img[y:y+h, x:x+w], breast_mask[y:y+h, x:x+w], blur, breast_mask, countour , y , x \n    return img[y:y+h, x:x+w], breast_mask[y:y+h, x:x+w], blur, breast_mask, countour ,binarized , opened_binarized , closed_binarized\ndef truncation_normalization(img , mask):\n    \"\"\"\n    Pixel clipped and normalized in breast ROI\n    \"\"\"\n    Pmin = np.percentile(img[img>=20], 10)\n    Pmax = np.percentile(img[img>=20], 99)\n    truncated = np.clip(img,Pmin, Pmax)  \n    normalized = (truncated - Pmin)/(Pmax - Pmin)\n    normalized[mask==0]=0\n    return normalized","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def clahe(img, clip):\n    #contrast enhancement\n    clahe = cv2.createCLAHE(clipLimit=clip)\n    cl = clahe.apply(np.array(img*255, dtype=np.uint8))\n    return cl\ndef synthesized_images(path , cmap = \"gray\" , display = False ):\n    \n    im = pydicom.dcmread(path)\n    pixels = im.pixel_array\n\n    if im.PhotometricInterpretation == \"MONOCHROME1\":\n        pixels = np.amax(pixels) - pixels\n    else:\n        pixels = pixels - np.min(pixels)\n\n    if np.max(pixels) != 0:\n        pixels = pixels / np.max(pixels)\n        pixels = (pixels * 255).astype(np.uint8)\n        \n    \n    breast, mask, blur, breast_mask, cnt , binarized_threshold , binarized_opened, binarized_closed = crop(pixels)\n\n    normalized = truncation_normalization(breast,mask)\n\n    cl1 = clahe(normalized, 1.0)\n    cl2 = clahe(normalized, 2.0)\n\n    synthetized = cv2.merge((np.array(normalized*255, dtype=np.uint8),cl1,cl2))\n    \n    \n    \n    if display:\n            fig, axes = plt.subplots(nrows = 5, ncols = 2, figsize = (24, 50),subplot_kw={'aspect': 1})\n\n            axes[0,0].imshow(pixels,cmap=cmap);\n            axes[0,0].set_title('Original Image');\n\n                #plt.gca().add_patch(circle2)\n\n            axes[0,1].imshow(blur,cmap=cmap);\n            axes[0,1].set_title('Blur Image');\n\n            axes[1,0].imshow(binarized_threshold,cmap=\"gray\");\n            axes[1,0].set_title('Binarized Image');\n            \n            axes[1,1].imshow(binarized_opened,cmap=\"gray\");\n            axes[1,1].set_title('Binarized Image Opened');\n            \n            axes[2,0].imshow(binarized_closed,cmap=\"gray\");\n            axes[2,0].set_title('Binarized Image Closed');\n\n            axes[2,1].imshow(cnt,cmap=\"gray\");\n            axes[2,1].set_title('Biggest Contour Image (mask)');\n\n            axes[3,0].imshow(breast,cmap=cmap);\n            axes[3,0].set_title('Crop based on mask boundaries');\n\n            axes[3,1].imshow(normalized,cmap=cmap);\n            axes[3,1].set_title('Normalised Image (Truncation Normalisation)');\n\n            axes[4,0].imshow(cl1,cmap=cmap);\n            axes[4,0].set_title('cl1');\n\n            axes[4,1].imshow(cl2,cmap=cmap);\n            axes[4,1].set_title('cl2');\n        \n        \n    \n    return breast, synthetized,normalized","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Tensorflow and CV2 set number of threads to 1 for speedup in parallell function mapping\ntf.config.threading.set_inter_op_parallelism_threads(num_threads=1)\ncv2.setNumThreads(1)\n\n# Pandas DataFrame Display Options\npd.options.display.max_colwidth = 99","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.471039Z","iopub.execute_input":"2023-06-07T13:38:01.472023Z","iopub.status.idle":"2023-06-07T13:38:01.490052Z","shell.execute_reply.started":"2023-06-07T13:38:01.471987Z","shell.execute_reply":"2023-06-07T13:38:01.488579Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"IS_INTERACTIVE = os.environ['KAGGLE_KERNEL_RUN_TYPE'] == 'Interactive'\nTARGET_HEIGHT = 1344\nTARGET_WIDTH = 768\nN_CHANNELS = 1\nINPUT_SHAPE = (TARGET_HEIGHT, TARGET_WIDTH, N_CHANNELS)\nTARGET_HEIGHT_WIDTH_RATIO = TARGET_HEIGHT / TARGET_WIDTH\nTHRESHOLD_BEST = 0.50\n\nCLAHE = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(32, 32))\n\nCROP_IMAGE = True\nAPPLY_CLAHE = False\nAPPLY_EQ_HIST = False\n\nIMAGE_FORMAT = 'jpg'","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.491508Z","iopub.execute_input":"2023-06-07T13:38:01.491888Z","iopub.status.idle":"2023-06-07T13:38:01.50623Z","shell.execute_reply.started":"2023-06-07T13:38:01.491857Z","shell.execute_reply":"2023-06-07T13:38:01.505022Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Source: https://www.kaggle.com/code/bobdegraaf/dicomsdl-voi-lut\ndef voi_lut(image, dicom):\n    # Additional Checks\n    if 'WindowWidth' not in dicom.getPixelDataInfo() or 'WindowWidth' not in dicom.getPixelDataInfo():\n        return image\n    \n    # Load only the variables we need\n    center = dicom['WindowCenter']\n    width = dicom['WindowWidth']\n    bits_stored = dicom['BitsStored']\n    voi_lut_function = dicom['VOILUTFunction']\n\n    # For sigmoid it's a list, otherwise a single value\n    if isinstance(center, list):\n        center = center[0]\n    if isinstance(width, list):\n        width = width[0]\n\n    # Set y_min, max & range\n    y_min = 0\n    y_max = float(2**bits_stored - 1)\n    y_range = y_max\n\n    # Function with default LINEAR (so for Nan, it will use linear)\n    if voi_lut_function == 'SIGMOID':\n        image = y_range / (1 + np.exp(-4 * (image - center) / width)) + y_min\n    else:\n        # Checks width for < 1 (in our case not necessary, always >= 750)\n        center -= 0.5\n        width -= 1\n\n        below = image <= (center - width / 2)\n        above = image > (center + width / 2)\n        between = np.logical_and(~below, ~above)\n\n        image[below] = y_min\n        image[above] = y_max\n        if between.any():\n            image[between] = (\n                ((image[between] - center) / width + 0.5) * y_range + y_min\n            )\n\n    return image","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.507737Z","iopub.execute_input":"2023-06-07T13:38:01.508099Z","iopub.status.idle":"2023-06-07T13:38:01.520977Z","shell.execute_reply.started":"2023-06-07T13:38:01.508052Z","shell.execute_reply":"2023-06-07T13:38:01.519593Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Smooth vector used to smoothen sums/stds of axes\ndef smooth(l):\n    # kernel size is 1% of vector\n    kernel_size = int(len(l) * 0.01)\n    kernel = np.ones(kernel_size) / kernel_size\n    return np.convolve(l, kernel, mode='same')\n\n# X Crop offset based on first column with sum below 5% of maximum column sums*std\ndef get_x_offset(image, max_col_sum_ratio_threshold=0.05, debug=None):\n    # Image Dimensions\n    H, W = image.shape\n    # Percentual margin added to offset\n    margin = int(image.shape[1] * 0.00)\n    # Threshold values based on smoothed sum x std to capture varying intensity columns\n    vv = smooth(image.sum(axis=0).squeeze()) * smooth(image.std(axis=0).squeeze())\n    # Find maximum sum in first 75% of columns\n    vv_argmax = vv[:int(image.shape[1] * 0.75)].argmax()\n    # Threshold value\n    vv_threshold = vv.max() * max_col_sum_ratio_threshold\n    \n    # Find first column after maximum column below threshold value\n    for offset, v in enumerate(vv):\n        # Start searching from vv_argmax\n        if offset < vv_argmax:\n            continue\n        \n        # Column below threshold value found\n        if v < vv_threshold:\n            offset = min(W, offset + margin)\n            break\n            \n    if isinstance(debug, np.ndarray):\n        debug[1].imshow(image)\n        debug[1].set_title('X Offset')\n        vv_scale = H / vv.max() * 0.90\n        # Values\n        debug[1].plot(H - vv * vv_scale , c='red', label='vv')\n        # Threshold\n        debug[1].hlines(H - vv_threshold * vv_scale, 0, W -1, colors='orange', label='threshold')\n        # Max Value\n        debug[1].scatter(vv_argmax, H - vv[vv_argmax] * vv_scale, c='blue', s=100, label='Max', zorder=np.PINF)\n        # First Column Below Threshold\n        debug[1].scatter(offset, H - vv[offset] * vv_scale, c='purple', s=100, label='Offset', zorder=np.PINF)\n        debug[1].set_ylim(H, 0)\n        debug[1].legend()\n        debug[1].axis('off')\n        \n    return offset\n\n# Y Crop offset based on first bottom and top rows with sum below 10% of maximum row sum*std\ndef get_y_offsets(image, max_row_sum_ratio_threshold=0.10, debug=None):\n    # Image Dimensions\n    H, W = image.shape\n    # Margin to add to offsets\n    margin = 0\n    # Threshold values based on smoothed sum x std to capture varying intensity columns\n    vv = smooth(image.sum(axis=1).squeeze()) * smooth(image.std(axis=1).squeeze())\n    # Find maximum sum * std row in inter quartile rows\n    vv_argmax = int(image.shape[0] * 0.25) + vv[int(image.shape[0] * 0.25):int(image.shape[0] * 0.75)].argmax()\n    # Threshold value\n    vv_threshold = vv.max() * max_row_sum_ratio_threshold\n    # Default crop offsets\n    offset_bottom = 0\n    offset_top = H\n\n    # Bottom offset, search from argmax to bottom\n    for offset in reversed(range(0, vv_argmax)):\n        v = vv[offset]\n        if v < vv_threshold:\n            offset_bottom = offset\n            break\n    \n    if isinstance(debug, np.ndarray):\n        debug[2].imshow(image)\n        debug[2].set_title('Y Bottom Offset')\n        vv_scale = W / vv.max() * 0.90\n        # Values\n        debug[2].plot(vv * vv_scale, np.arange(H), c='red', label='vv')\n        # Threshold\n        debug[2].vlines(vv_threshold * vv_scale, 0, H -1, colors='orange', label='threshold')\n        # Max Value\n        debug[2].scatter(vv[vv_argmax] * vv_scale, vv_argmax, c='blue', s=100, label='Max', zorder=np.PINF)\n        # First Column Below Threshold\n        debug[2].scatter(vv[offset_bottom] * vv_scale, offset_bottom, c='purple', s=100, label='Offset', zorder=np.PINF)\n        debug[2].set_ylim(H, 0)\n        debug[2].legend()\n        debug[2].axis('off')\n            \n    # Top offset, search from argmax to top\n    for offset in range(vv_argmax, H):\n        v = vv[offset]\n        if v < vv_threshold:\n            offset_top = offset\n            break\n            \n    if isinstance(debug, np.ndarray):\n        debug[3].imshow(image)\n        debug[3].set_title('Y Top Offset')\n        vv_scale = W / vv.max() * 0.90\n        # Values\n        debug[3].plot(vv * vv_scale, np.arange(H) , c='red', label='vv')\n        # Threshold\n        debug[3].vlines(vv_threshold * vv_scale, 0, H -1, colors='orange', label='threshold')\n        # Max Value\n        debug[3].scatter(vv[vv_argmax] * vv_scale, vv_argmax, c='blue', s=100, label='Max', zorder=np.PINF)\n        # First Column Below Threshold\n        debug[3].scatter(vv[offset_top] * vv_scale, offset_top, c='purple', s=100, label='Offset', zorder=np.PINF)\n        debug[2].set_ylim(H, 0)\n        debug[3].legend()\n        debug[3].axis('off')\n            \n    return max(0, offset_bottom - margin), min(image.shape[0], offset_top + margin)\n\n# Crop image and pad offsets to target image height/width ratio to preserve information\ndef crop(image, size=None, debug=False):\n    # Image dimensions\n    H, W = image.shape\n    # Compute x/bottom/top offsets\n    x_offset = get_x_offset(image, debug=debug)\n    offset_bottom, offset_top = get_y_offsets(image[:,:x_offset], debug=debug)\n    # Crop Height and Width\n    h_crop = offset_top - offset_bottom\n    w_crop = x_offset\n    \n    # Pad crop offsets to target aspect ratio\n    if size is not None:\n        # Height too large, pad x offset\n        if (h_crop / w_crop) > TARGET_HEIGHT_WIDTH_RATIO:\n            x_offset += int(h_crop / TARGET_HEIGHT_WIDTH_RATIO - w_crop)\n        else:\n            # Height too small, pad bottom/top offsets\n            offset_bottom -= int(0.50 * (w_crop * TARGET_HEIGHT_WIDTH_RATIO - h_crop))\n            offset_bottom_correction = max(0, -offset_bottom)\n            offset_bottom += offset_bottom_correction\n\n            offset_top += int(0.50 * (w_crop * TARGET_HEIGHT_WIDTH_RATIO - h_crop))\n            offset_top += offset_bottom_correction\n        \n    # Crop Image\n    image = image[offset_bottom:offset_top:,:x_offset]\n        \n    return image","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.522596Z","iopub.execute_input":"2023-06-07T13:38:01.523036Z","iopub.status.idle":"2023-06-07T13:38:01.560695Z","shell.execute_reply.started":"2023-06-07T13:38:01.522994Z","shell.execute_reply":"2023-06-07T13:38:01.559675Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def process(file_path, size=(TARGET_WIDTH, TARGET_HEIGHT), crop_image=CROP_IMAGE, apply_clahe=APPLY_CLAHE, apply_eq_hist=APPLY_EQ_HIST, debug=False, save=True):\n    # Read Dicom File\n    dicom = dicomsdl.open(file_path)\n    image = dicom.pixelData()\n    \n    # Save original image for debug purposes\n    if debug:\n        fig, axes = plt.subplots(1, 5, figsize=(20,10))\n        image0 = np.copy(image)\n        axes[0].imshow(image0)\n        axes[0].set_title('Original Image')\n        axes[0].axis('off')\n    else:\n        axes = False\n    \n    # voi_lut\n    try:\n        image = voi_lut(image, dicom)\n    except:\n        pass\n    \n    # Some images have 0 values as highest intensity and need to be inverted\n    if dicom.getPixelDataInfo()['PhotometricInterpretation'] == 'MONOCHROME1':\n        image = np.max(image) - image\n\n    # Normalize [0,1] range\n    image = (image - image.min()) / (image.max() - image.min())\n\n    # Convert to uint8 image in range [0, 255]\n    image = (image * 255).astype(np.uint8)\n    \n    # Flip T0 Left/Right Orientation\n    h0, w0 = image.shape\n    if image[:,int(-w0 * 0.10):].sum() > image[:,:int(w0 * 0.10)].sum():\n        image = np.flip(image, axis=1)\n    \n    # Crop Image\n    if crop_image:\n        image = crop(image, debug=axes)\n        \n    # Resize\n    if size is not None:\n        # Pad black pixels to make square image\n        h, w = image.shape\n        if (h / w) > TARGET_HEIGHT_WIDTH_RATIO:\n            pad = int(h / TARGET_HEIGHT_WIDTH_RATIO - w)\n            image = np.pad(image, [[0,0], [0, pad]])\n            h, w = image.shape\n        else:\n            pad = int(0.50 * (w * TARGET_HEIGHT_WIDTH_RATIO - h))\n            image = np.pad(image, [[pad, pad], [0,0]])\n            h, w = image.shape\n        # Resize\n        image = cv2.resize(image, size, interpolation=cv2.INTER_AREA)\n        \n    # Apply CLAHE contrast enhancement\n    if apply_clahe:\n        image = CLAHE.apply(image)\n        \n     # Apply Histogram Equalization\n    if apply_eq_hist:\n        image = cv2.equalizeHist(image)\n        \n    # Show Processed Image    \n    if debug:\n        axes[4].imshow(image)\n        axes[4].set_title('Processed Image')\n        axes[4].axis('off')\n        plt.show()\n        \n    # Save Only\n    if save:\n        image_id = file_path.split('/')[-1].split('.')[0]\n        if IMAGE_FORMAT == 'png':\n            cv2.imwrite(f'{image_id}.png', image)\n        else:\n            cv2.imwrite(f'{image_id}.jpg', image, [cv2.IMWRITE_JPEG_QUALITY, 95])","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.562044Z","iopub.execute_input":"2023-06-07T13:38:01.562442Z","iopub.status.idle":"2023-06-07T13:38:01.58415Z","shell.execute_reply.started":"2023-06-07T13:38:01.562382Z","shell.execute_reply":"2023-06-07T13:38:01.582801Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def process(file_path, size=(TARGET_WIDTH, TARGET_HEIGHT), crop_image=CROP_IMAGE, apply_clahe=APPLY_CLAHE, apply_eq_hist=APPLY_EQ_HIST, debug=False, save=True):\n    # Read Dicom File\n    dicom = dicomsdl.open(file_path)\n    image = dicom.pixelData()\n    \n    # Save original image for debug purposes\n    if debug:\n        fig, axes = plt.subplots(1, 5, figsize=(20,10))\n        image0 = np.copy(image)\n        axes[0].imshow(image0)\n        axes[0].set_title('Original Image')\n        axes[0].axis('off')\n    else:\n        axes = False\n    \n    # voi_lut\n    try:\n        image = voi_lut(image, dicom)\n    except:\n        pass\n    \n    # Some images have 0 values as highest intensity and need to be inverted\n    if dicom.getPixelDataInfo()['PhotometricInterpretation'] == 'MONOCHROME1':\n        image = np.max(image) - image\n\n    # Normalize [0,1] range\n    image = (image - image.min()) / (image.max() - image.min())\n\n    # Convert to uint8 image in range [0, 255]\n    image = (image * 255).astype(np.uint8)\n    \n    # Flip T0 Left/Right Orientation\n    h0, w0 = image.shape\n    if image[:,int(-w0 * 0.10):].sum() > image[:,:int(w0 * 0.10)].sum():\n        image = np.flip(image, axis=1)\n    \n    # Crop Image\n    if crop_image:\n        image = crop(image, debug=axes)\n        \n    # Resize\n    if size is not None:\n        # Pad black pixels to make square image\n        h, w = image.shape\n        if (h / w) > TARGET_HEIGHT_WIDTH_RATIO:\n            pad = int(h / TARGET_HEIGHT_WIDTH_RATIO - w)\n            image = np.pad(image, [[0,0], [0, pad]])\n            h, w = image.shape\n        else:\n            pad = int(0.50 * (w * TARGET_HEIGHT_WIDTH_RATIO - h))\n            image = np.pad(image, [[pad, pad], [0,0]])\n            h, w = image.shape\n        # Resize\n        image = cv2.resize(image, size, interpolation=cv2.INTER_AREA)\n        \n    # Apply CLAHE contrast enhancement\n    if apply_clahe:\n        image = CLAHE.apply(image)\n        \n     # Apply Histogram Equalization\n    if apply_eq_hist:\n        image = cv2.equalizeHist(image)\n        \n    # Show Processed Image    \n    if debug:\n        axes[4].imshow(image)\n        axes[4].set_title('Processed Image')\n        axes[4].axis('off')\n        plt.show()\n        \n    # Save Only\n    if save:\n        image_id = file_path.split('/')[-1].split('.')[0]\n        if IMAGE_FORMAT == 'png':\n            cv2.imwrite(f'{image_id}.png', image)\n        else:\n            cv2.imwrite(f'{image_id}.jpg', image, [cv2.IMWRITE_JPEG_QUALITY, 95])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train = pd.read_csv('/kaggle/input/rsna-breast-cancer-detection/train.csv')\n    \ndef get_file_path(args):\n    patient_id, image_id = args\n    return f'/kaggle/input/rsna-breast-cancer-detection/train_images/{patient_id}/{image_id}.dcm'\n    \ntrain['file_path'] = train[['patient_id', 'image_id']].apply(get_file_path, axis=1)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:01.58614Z","iopub.execute_input":"2023-06-07T13:38:01.586718Z","iopub.status.idle":"2023-06-07T13:38:02.307832Z","shell.execute_reply.started":"2023-06-07T13:38:01.58668Z","shell.execute_reply":"2023-06-07T13:38:02.306678Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"N = 1\n\nfor fp in tqdm(train['file_path'].head(N)):\n    process(fp, crop_image=True, size=(TARGET_WIDTH, TARGET_HEIGHT), debug=True, save=False)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:02.312329Z","iopub.execute_input":"2023-06-07T13:38:02.313445Z","iopub.status.idle":"2023-06-07T13:38:07.873222Z","shell.execute_reply.started":"2023-06-07T13:38:02.313406Z","shell.execute_reply":"2023-06-07T13:38:07.872138Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def normalize(image):\n    # Repeat channels to create 3 channel images required by pretrained ConvNextV2 models\n    image = tf.repeat(image, repeats=3, axis=3)\n    # Cast to float 32\n    image = tf.cast(image, tf.float32)\n    # Normalize with respect to ImageNet mean/std\n    image = tf.keras.applications.imagenet_utils.preprocess_input(image, mode='torch')\n\n    return image","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:07.875125Z","iopub.execute_input":"2023-06-07T13:38:07.875511Z","iopub.status.idle":"2023-06-07T13:38:07.8825Z","shell.execute_reply.started":"2023-06-07T13:38:07.87548Z","shell.execute_reply":"2023-06-07T13:38:07.881348Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def get_model():\n    # Inputs, note the names are equal to the dictionary keys in the dataset\n    image = tf.keras.layers.Input(INPUT_SHAPE, name='image', dtype=tf.uint8)\n\n    # Normalize Input\n    image_norm = normalize(image)\n\n    # CNN Feature Maps\n    x = efficientnet.EfficientNetV2S(\n        input_shape=(TARGET_HEIGHT, TARGET_WIDTH, 3),\n        pretrained=None,\n        num_classes=0,\n    )(image_norm)\n\n    # Average Pooling BxHxWxC -> BxC\n    x = tf.keras.layers.GlobalAveragePooling2D()(x)\n    # Dropout to prevent Overfitting\n    x = tf.keras.layers.Dropout(0.30)(x)\n    # Output value between [0, 1] using Sigmoid function\n    outputs = tf.keras.layers.Dense(1, activation='sigmoid')(x)\n\n    # Define model with inputs and outputs\n    model = tf.keras.models.Model(inputs=image, outputs=outputs)\n\n    # Load pretrained Model Weights\n    model.load_weights('/kaggle/input/model13/model_13.h5')\n\n    # Set model non-trainable\n    model.trainable = False\n\n    # Compile model\n    model.compile()\n\n    return model","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:07.88419Z","iopub.execute_input":"2023-06-07T13:38:07.884547Z","iopub.status.idle":"2023-06-07T13:38:07.894698Z","shell.execute_reply.started":"2023-06-07T13:38:07.884518Z","shell.execute_reply":"2023-06-07T13:38:07.89359Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Pretrained File Path: '/kaggle/input/sartorius-training-dataset/model.h5'\ntf.keras.backend.clear_session()\n# enable XLA optmizations\ntf.config.optimizer.set_jit(True)\n\nmodel = get_model()","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:07.896041Z","iopub.execute_input":"2023-06-07T13:38:07.896404Z","iopub.status.idle":"2023-06-07T13:38:16.774342Z","shell.execute_reply.started":"2023-06-07T13:38:07.896374Z","shell.execute_reply":"2023-06-07T13:38:16.773204Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Plot model summary\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:16.776Z","iopub.execute_input":"2023-06-07T13:38:16.776493Z","iopub.status.idle":"2023-06-07T13:38:16.849999Z","shell.execute_reply.started":"2023-06-07T13:38:16.776446Z","shell.execute_reply":"2023-06-07T13:38:16.848871Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test = pd.read_csv('/kaggle/input/rsna-breast-cancer-detection/test.csv')\n\ndef get_file_path(args):\n    patient_id, image_id = args\n    return f'/kaggle/input/rsna-breast-cancer-detection/test_images/{patient_id}/{image_id}.dcm'\n    \ntest['file_path'] = test[['patient_id', 'image_id']].apply(get_file_path, axis=1)\n\ndisplay(test.info())\ndisplay(test.head())","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:16.851701Z","iopub.execute_input":"2023-06-07T13:38:16.852029Z","iopub.status.idle":"2023-06-07T13:38:16.911211Z","shell.execute_reply.started":"2023-06-07T13:38:16.852002Z","shell.execute_reply":"2023-06-07T13:38:16.909971Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Preprocess a single image and saves it\ndef preprocess_and_save_image(args):\n    (patient_id, laterality), g = args\n    cancer = 0.0\n    for row_idx, row in g.iterrows():\n        process(row['file_path'])","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:16.912658Z","iopub.execute_input":"2023-06-07T13:38:16.913008Z","iopub.status.idle":"2023-06-07T13:38:16.919089Z","shell.execute_reply.started":"2023-06-07T13:38:16.912979Z","shell.execute_reply":"2023-06-07T13:38:16.917947Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Preprocess all images in parallel using Joblib\njobs = [joblib.delayed(preprocess_and_save_image)(args) for args in test.groupby(['patient_id', 'laterality'])]\nSUBMISSION_ROWS = joblib.Parallel(\n    n_jobs=cpu_count(),\n    verbose=IS_INTERACTIVE,\n    backend='multiprocessing',\n    prefer='threads',\n)(jobs)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:16.920601Z","iopub.execute_input":"2023-06-07T13:38:16.920983Z","iopub.status.idle":"2023-06-07T13:38:19.414615Z","shell.execute_reply.started":"2023-06-07T13:38:16.920944Z","shell.execute_reply":"2023-06-07T13:38:19.413404Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"SUBMISSION_ROWS = []\n# Iterate over all patient_id/laterality combinations groups\nfor idx, ((patient_id, laterality), g) in enumerate(tqdm(test.groupby(['patient_id', 'laterality']))):\n    # Cancer target is mean of predicted cancer values\n    cancer = 0\n    # Iterate over all scans in group\n    for row_idx, row in g.iterrows():\n        # Load Image\n        image_id = row['image_id']\n        image = cv2.imread(f'{image_id}.{IMAGE_FORMAT}', -1)\n        # Show First Few Images\n        if idx < 16:\n            plt.figure(figsize=(5,8))\n            plt.imshow(image)\n            plt.show()\n        \n        # Expand to Batch HxW -> 1xHxWx1\n        image = np.expand_dims(image, [0, 3])\n        # Make Prediction\n        cancer += model.predict_on_batch(image).squeeze() / len(g)\n        # Remove Image\n        os.remove(f'{image_id}.{IMAGE_FORMAT}')\n        \n    # Add Submission Row\n    SUBMISSION_ROWS.append({\n        'prediction_id': f'{patient_id}_{laterality}',\n        'cancer': np.int8(cancer > THRESHOLD_BEST),\n    })\n    \n    if np.random.rand() > 0.99:\n        gc.collect()","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:19.416614Z","iopub.execute_input":"2023-06-07T13:38:19.417018Z","iopub.status.idle":"2023-06-07T13:38:32.543586Z","shell.execute_reply.started":"2023-06-07T13:38:19.416977Z","shell.execute_reply":"2023-06-07T13:38:32.542531Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"SUBMISSION_ROWS","metadata":{"execution":{"iopub.status.busy":"2023-06-07T13:38:32.545382Z","iopub.execute_input":"2023-06-07T13:38:32.546067Z","iopub.status.idle":"2023-06-07T13:38:32.552934Z","shell.execute_reply.started":"2023-06-07T13:38:32.546031Z","shell.execute_reply":"2023-06-07T13:38:32.552031Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}