{"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":"import numpy as np\nimport os\nimport pandas as pd\nimport matplotlib.pyplot as plt\nfrom IPython.core.display import HTML\nimport glob\nimport pydicom\nfrom pydicom.pixel_data_handlers.util import apply_voi_lut\nimport math\nimport ast\nimport cv2\n\ndef css_styling():\n styles = open(\"../input/titlestyle/style2.css\", \"r\").read()\n return HTML(styles)\ncss_styling()","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-08-10T19:02:02.263683Z","iopub.execute_input":"2021-08-10T19:02:02.264086Z","iopub.status.idle":"2021-08-10T19:02:02.788835Z","shell.execute_reply.started":"2021-08-10T19:02:02.263995Z","shell.execute_reply":"2021-08-10T19:02:02.787930Z"},"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Intro

\n
\n
\n\n📔 Public notebooks - 546+
\n\n🥇 Gold medals - 26+
\n\n🥈 Silver medals - 40+
\n\n🥉 Bronze medals - 94+
\n \n1️ place solution - https://www.kaggle.com/c/siim-covid19-detection/discussion/263658
\n \n2️ place solution - https://www.kaggle.com/c/siim-covid19-detection/discussion/263674
\n \n3️ place solution - https://www.kaggle.com/c/siim-covid19-detection/discussion/263654
\n\n\nWordcloud made of notebook titles:\n\"img1\"/\n
","metadata":{}},{"cell_type":"markdown","source":"
\nSumma summarum: EfficientNet, YOLOv5 and dicom files \n
","metadata":{}},{"cell_type":"markdown","source":"
\n

Goal

\n
\n
\n

😷 Categorize chest radiographs as negative for pneumonia, typical, indeterminate, or atypical for COVID-19. If some abnormalities are found, provide the bounding boxes.

\n

😷 Image can have positive value for only one label but multiple bounding boxes. This competition is both object detection and classification

\n
","metadata":{}},{"cell_type":"code","source":"train_study = pd.read_csv(\"../input/siim-covid19-detection/train_study_level.csv\")\ntrain_image = pd.read_csv(\"../input/siim-covid19-detection/train_image_level.csv\")\ntrain_study.head()","metadata":{"execution":{"iopub.status.busy":"2021-08-10T14:40:58.484828Z","iopub.execute_input":"2021-08-10T14:40:58.48559Z","iopub.status.idle":"2021-08-10T14:40:58.584477Z","shell.execute_reply.started":"2021-08-10T14:40:58.485531Z","shell.execute_reply":"2021-08-10T14:40:58.583175Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_image.head()","metadata":{"execution":{"iopub.status.busy":"2021-08-10T14:41:01.969937Z","iopub.execute_input":"2021-08-10T14:41:01.970378Z","iopub.status.idle":"2021-08-10T14:41:01.987441Z","shell.execute_reply.started":"2021-08-10T14:41:01.97034Z","shell.execute_reply":"2021-08-10T14:41:01.985449Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def fix_inverted_radiograms(data, img):\n '''Fixes inverted radiograms - with PhotometricInterpretation == \"MONOCHROME1\"\n data: the .dcm dataset\n img: the .dcm pixel_array'''\n \n if data.PhotometricInterpretation == \"MONOCHROME1\":\n img = np.amax(img) - img\n \n img = img - np.min(img)\n img = img / np.max(img)\n img = (img * 255).astype(np.uint8)\n \n return img\n\ndef get_image_metadata(study_id, df):\n '''Returns the label and bounding boxes (if any)\n for a speciffic study id.'''\n \n data = df[df[\"study_id\"] == study_id]\n \n if data[\"Negative for Pneumonia\"].values == 1:\n label = \"negative_for_pneumonia\"\n elif data[\"Typical Appearance\"].values == 1:\n label = \"typical\"\n elif data[\"Indeterminate Appearance\"].values == 1:\n label = \"indeterminate\"\n else:\n label = \"atypical\"\n \n bbox = list(data[\"boxes\"].values)\n \n return label, bbox\n\ndef return_coords(box):\n '''Returns coordinates from a bbox'''\n # Get the list of dictionaries\n box = ast.literal_eval(box)[0]\n # Get the exact x and y coordinates\n x1, y1, x2, y2 = box[\"x\"], box[\"y\"], box[\"x\"] + box[\"width\"], box[\"y\"] + box[\"height\"]\n # Save coordinates\n return (int(x1), int(y1), int(x2), int(y2))\n\ndef show_dcm_info(study_ids, df):\n '''Show .dcm images along with description.'''\n \n fig, axes = plt.subplots(nrows=2, ncols=3, figsize=(21,10))\n\n # Get .dcm paths\n dcm_paths = [glob.glob(f\"../input/siim-covid19-detection/train/{study_id}/*/*\")[0]\n for study_id in study_ids]\n datasets = [pydicom.dcmread(path) for path in dcm_paths]\n images = [apply_voi_lut(dataset.pixel_array, dataset) for dataset in datasets]\n\n # Loop through the information\n for study_id, data, img, i in zip(study_ids, datasets, images, range(2*3)):\n # Fix inverted images\n img = fix_inverted_radiograms(data, img)\n\n # Below function available in functions section ;)\n label, bbox = get_image_metadata(study_id, df)\n \n # Check for bounding box and add if it's the case\n try: \n # For no bbox, the list is [nan]\n no_box = math.isnan(bbox[0])\n pass\n except TypeError:\n # Retrieve the bounding box\n all_coords = []\n for box in bbox:\n all_coords.append(return_coords(box))\n\n for (x1, y1, x2, y2) in all_coords:\n cv2.rectangle(img, (x1, y1), (x2, y2), (0, 80, 255), 15)\n cv2.putText(img, label, (x1, y1-14), \n cv2.FONT_HERSHEY_SIMPLEX, 3, (0, 0, 0), 4)\n \n # Plot the image\n x = i // 3\n y = i % 3\n \n axes[x, y].imshow(img, cmap=\"binary\")\n axes[x, y].set_title(f\"Label: {label} \\n Sex: {data.PatientSex} | Body Part: {data.BodyPartExamined}\", \n fontsize=14, weight='bold')\n axes[x, y].axis('off')\n \ntrain_study[\"study_id\"] = train_study[\"id\"].apply(lambda x: x.split(\"_\")[0])\ntrain = pd.merge(train_image, train_study, \n left_on=\"StudyInstanceUID\", right_on=\"study_id\")\n\ntrain.drop([\"id_x\", \"StudyInstanceUID\", \"id_y\"], axis=1, inplace=True)\n\nshow_dcm_info(study_ids=[\"72044bb44d41\", \"f807cd855d31\", \"b949689a9ef1\",\n \"612ea5194007\", \"082cafb03942\", \"d3e83031ebea\"], \n df=train)","metadata":{"execution":{"iopub.status.busy":"2021-08-10T13:26:53.664465Z","iopub.execute_input":"2021-08-10T13:26:53.66494Z","iopub.status.idle":"2021-08-10T13:27:05.070797Z","shell.execute_reply.started":"2021-08-10T13:26:53.664907Z","shell.execute_reply":"2021-08-10T13:27:05.069704Z"},"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# dicom file example, it contains additional meta data\npydicom.dcmread('../input/siim-covid19-detection/train/72044bb44d41/c39be19d56ba/7452ddd5b44b.dcm')","metadata":{"execution":{"iopub.status.busy":"2021-08-10T15:03:39.754393Z","iopub.execute_input":"2021-08-10T15:03:39.754905Z","iopub.status.idle":"2021-08-10T15:03:39.804654Z","shell.execute_reply.started":"2021-08-10T15:03:39.754857Z","shell.execute_reply":"2021-08-10T15:03:39.803445Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Data preparation

\n
\n
\n \nDICOM stands for Digital Imaging and Communications in Medicine. It is a standard, internationally accepted format to view, store, retrieve and share medical images. This file consists of a header and image data sets packed into a single file. Notebook for preparing dicom files with extracting of meta data can be found here https://www.kaggle.com/andradaolteanu/siim-covid-19-box-detect-dcm-metadata

\n \nData augmentation is a strategy that enables practitioners to significantly increase the diversity of data available for training models, without actually collecting new data. Some methods such as random contrast, random crop, RGB shift and others are presented here https://www.kaggle.com/ruchi798/siim-covid-19-detection-eda-data-augmentation

\n \nImage preprocessing with ploting boxes https://www.kaggle.com/yujiariyasu/catch-up-on-positive-samples-plot-submission-csv

\n \n \nVery informative notebook with several image transformation techniques including histogram equalization, CLAHE, noise reduction https://www.kaggle.com/yujiariyasu/catch-up-on-positive-samples-plot-submission-csv

\n\nGreat notebook with lung segmentation and a list of other processing notebooks https://www.kaggle.com/davidbroberts/lung-segmentation-without-cnn

\n \nData augumentation using albumentations package https://www.kaggle.com/boltcoder/siim-covid19-simple-data-augmentation-techniques

\n \n\n
","metadata":{}},{"cell_type":"markdown","source":"
\n

Models

\n
\n
\n \nYOLO an acronym for 'You only look once', is an object detection algorithm that divides images into a grid system. Each cell in the grid is responsible for detecting objects within itself. This model enables real-time object detection with convolutional neural networks. How to prepare data and run YOLOv5 is explained here https://www.kaggle.com/ayuraj/train-covid-19-detection-using-yolov5

\n \nConvert dicom to png and use Efficientnet with Keras for classification and YOLOv5 for detection https://www.kaggle.com/h053473666/siim-cov19-efnb7-yolov5-infer

\n \nMMDetection is an open-source toolbox based on PyTorch for Object Detection and Segmentation tasks. The toolbox supports over 50+ baselines. One of the models, CascadeRCNN is presented here https://www.kaggle.com/sreevishnudamodaran/siim-mmdetection-cascadercnn-weight-bias

\n \nFasterRCNN using PyTorch https://www.kaggle.com/piantic/train-siim-covid-19-detection-fasterrcnn

\n \nTraining several EfficientNet models using TensorFlow https://www.kaggle.com/sreevishnudamodaran/siim-effnetv2-keras-study-train-tpu-cv0-805

\n \nEfficientNet using PyTorch and timm https://www.kaggle.com/heyytanay/siim-pytorch-classification-only-training-effnets

\n \nYOLOv5 object detection training with easy to follow explanation https://www.kaggle.com/ammarnassanalhajali/covid-19-detection-yolov5-3classes-training

\n \n \nVision Transformer with Grad-CAM algorithm used to interpret prediction of the model https://www.kaggle.com/basu369victor/covid19-detection-with-vit-and-heatmap and one more notebook with Grad-CAM https://www.kaggle.com/sinamhd9/where-s-your-model-looking-at-grad-cam\n
","metadata":{}},{"cell_type":"markdown","source":"","metadata":{}},{"cell_type":"markdown","source":"
\n

Other useful notebooks

\n
\n
\nLoad 3D CT scans using dicom files and work with them https://www.kaggle.com/allunia/pulmonary-dicom-preprocessing\n \nConvert dicom to jpg https://www.kaggle.com/xhlulu/siim-covid-19-convert-to-jpg-256px\n
","metadata":{}}]}