3d-cnn-predicting-mask-to-genomic-features.ipynb

{"cells":[{"metadata":{"trusted":true},"cell_type":"code","source":"\"\"\"\nLink to kaggle notebook:\nhttps://www.kaggle.com/parthvora/3d-cnn-predicting-mask-to-genomic-features/\n\n\"\"\"","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"import os\nfrom glob import glob\nimport random\n\nfrom collections import OrderedDict\nimport numpy as np\nimport pandas as pd\nimport cv2 as cv\nfrom PIL import Image\nfrom sklearn.model_selection import train_test_split\nfrom matplotlib import pyplot as plt\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom torch.utils.data import Dataset\nfrom torch.utils.data import DataLoader\nimport torchvision.transforms as transforms\nimport torchvision.transforms.functional as TF","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"df = pd.read_csv(\"/kaggle/input/lgg-mri-segmentation/lgg-mri-segmentation/kaggle_3m/data.csv\")\nlabels = df.COCCluster.to_numpy()\nlabels -= 1\nprint(np.unique(labels))\n\npatients = []\ndata_dir = '/kaggle/input/lgg-mri-segmentation/kaggle_3m/'\nfor dirname, _, filenames in os.walk(data_dir):\n    if dirname == data_dir:\n        continue\n        \n    patients.append(dirname)\n\n#sort(patients)\n\ncoc = []\ncoc.append(len(np.where(labels == 0)[0]))\ncoc.append(len(np.where(labels == 1)[0]))\ncoc.append(len(np.where(labels == 2)[0]))\n\nplt.bar([1, 2, 3], coc)\nplt.xlabel(\"Type\")\nplt.xticks([1, 2, 3])\nplt.ylabel(\"Num Patients\")\nplt.title(\"Histogram of patients for COC subtype\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"class MRIDataset3D(Dataset):\n    \n    # Takes in a list of images\n    def __init__(self, patients, labels):\n        self.patients = patients\n        self.labels = labels\n    \n    def __len__(self):\n        return(len(self.patients))\n    \n    def __getitem__(self, i):\n\n        patient = self.patients[i]\n        label = self.labels[i]\n        \n        files = os.listdir(patient)\n        \n        # 1 channel, max of 88 frames\n        masks = np.zeros((1, 88, 256, 256))\n        \n        # Stack masks together to make a 3D box\n        for i in files:\n            if \"mask\" in i:\n                fn = patient + \"/\" + i\n                m = cv.imread(fn, cv.IMREAD_GRAYSCALE)               \n                index = int(i.split(\"_\")[-2]) - 1\n                masks[0, index, :, :] = m\n\n        return masks, label","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"X_train, X_test, y_train, y_test = train_test_split(patients, labels, test_size=0.2, shuffle=True, stratify=labels) \n\nprint(len(y_test))\n\ntrain_dataset = MRIDataset3D(X_train, y_train)\ntest_dataset = MRIDataset3D(X_test, y_test)\n\ntrain_loader = DataLoader(train_dataset, batch_size=10, shuffle=True)\ntest_loader = DataLoader(test_dataset, batch_size=1, shuffle=False)\n\nfor (i, (x, y)) in enumerate(train_loader):\n    print(x.dtype)\n    break\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"class CNN3D(nn.Module):\n    def __init__(self):\n        super(CNN3D, self).__init__()\n        \n        self.layer1 = nn.Sequential(\n                        nn.Conv3d(in_channels=1, out_channels=2, kernel_size=7, stride=2), \n                        nn.ReLU(),\n                        nn.MaxPool3d(kernel_size=3, stride=2)\n                      )\n        \n        self.layer2 = nn.Sequential(\n                          nn.Conv3d(in_channels=2, out_channels=4, kernel_size=5, stride=1), \n                          nn.ReLU(),\n                          nn.MaxPool3d(kernel_size=3, stride=2)\n                      )\n        \n        self.layer3 = nn.Sequential(\n                          nn.Conv3d(in_channels=4, out_channels=8, kernel_size=3, stride=1), \n                          nn.ReLU(),\n                          nn.MaxPool3d(kernel_size=3, stride=2)\n                      )\n        \n        self.linear1 = nn.Sequential(\n                           nn.Linear(8 * 2 * 12 * 12, 100),\n                           nn.ReLU()\n                       )\n        \n        #self.linear2 = nn.Sequential(\n        #                   nn.Linear(1000, 100),\n        #                   nn.ReLU()\n        #               )\n        \n        self.linear3 = nn.Sequential(\n                           nn.Linear(100, 3),\n                       )\n        \n    def forward(self, x):\n        out = self.layer1(x)\n        out = self.layer2(out)\n        out = self.layer3(out)\n        out = out.reshape(out.size(0), -1)\n\n        out = self.linear1(out)\n        nn.Dropout(0.5)\n        #out = self.linear2(out)\n        #nn.Dropout(0.5)\n        out = self.linear3(out)\n\n        return out","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def train(model, train_loader, test_loader, epochs, lr):\n    \n    # Can try different optimizers\n    optimizer = torch.optim.Adam(model.parameters(), lr)\n    compute_loss = nn.CrossEntropyLoss()\n    \n    train_loss, test_loss = [], []\n    \n    if torch.cuda.is_available():\n        model = model.cuda()\n    else:\n        print(\"ERROR: CUDA NOT AVAILABLE\")\n        return [], []\n    \n    for epoch in range(epochs):\n        \n        # Train\n        for step, (X, y) in enumerate(train_loader):\n            \n            optimizer.zero_grad()\n            \n            X = X.float()           \n            X = X.cuda()\n                \n            y_hat = model(X)\n            y_hat = y_hat.cpu()\n            \n            loss = compute_loss(y_hat, y)\n            \n            train_loss.append(loss.item())\n            \n            loss.backward()\n            optimizer.step()\n\n            del X, y, y_hat\n            \n        # Test\n        t = []\n        for step, (X, y) in enumerate(test_loader):\n            \n            X = X.float()\n            X = X.cuda()\n                \n            y_hat = model(X)\n            y_hat = y_hat.cpu()\n            \n            loss = compute_loss(y_hat, y)\n            t.append(loss.item())\n            \n            del X, y, y_hat\n        test_loss.append(np.mean(t))\n        \n        print(\"Epoch: {0}\".format(epoch + 1))\n        print(\"Train Loss: {0}\".format(train_loss[-1]))\n        print(\"Validation Loss: {0}\".format(test_loss[-1]))\n     \n    return train_loss, test_loss","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model = CNN3D()\n\nepochs = 10\nlr = 0.00005\ntrain_loss, test_loss = train(model, train_loader, test_loader, epochs, lr)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"plt.plot(range(len(train_loss)), train_loss)\nplt.plot([0, 10, 20, 30, 40, 50, 60, 70, 80, 90], test_loss)\nplt.xticks([0, 10, 20, 30, 40, 50, 60, 70, 80, 90], [0, 1, 2, 3, 4, 5, 6, 7, 8, 9])\nplt.xlabel(\"Epochs\")\nplt.ylabel(\"Cross Entropy Loss\")\nplt.title(\"Training and Validation Loss for 3D CNN\")\nplt.legend([\"Training\", \"Validation\"])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def test(model, test_loader):\n    \n    acc = []\n    preds = []\n    true_labels = []\n    for step, (X, y) in enumerate(test_loader):\n        \n        X = X.float()\n        X = X.cuda()\n        \n        y_hat = model(X)\n        y_hat = y_hat.cpu()\n        \n        y_hat = y_hat.detach().numpy().reshape(3)\n        pred = np.argmax(y_hat)\n        y = y.numpy()\n\n        preds.append(pred)\n        true_labels.append(y)\n        acc.append(pred == y)\n        \n    return acc, true_labels, preds\n    \n    ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"acc, true_labels, preds = test(model, test_loader)\nprint(np.mean(acc))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.metrics import confusion_matrix\nimport seaborn as sn\ncm = confusion_matrix(true_labels, preds)\nsn.heatmap(cm, annot=True, xticklabels=[1, 2, 3], yticklabels=[1, 2, 3])\nplt.xlabel(\"Predicted Labels\")\nplt.ylabel(\"True Labels\")\nplt.title(\"Confusion matrix for 3D CNN (20% Test Set)\")\n","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.7.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":4}