{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nfrom skimage import io,feature\nimport matplotlib.pyplot as plt\nfrom skimage.io import MultiImage\nfrom skimage.color import rgb2gray\nfrom skimage.filters import threshold_otsu\nfrom skimage.transform import rescale, resize, downscale_local_mean\nfrom PIL import Image\nimport os\nimport shutil\nfrom tqdm import tqdm\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        os.path.join(dirname, filename)\n        #print(os.path.join(dirname, filename))\n\n# Any results you write to the current directory are saved as output.\npath_2img = \"../input/prostate-cancer-grade-assessment/train_images/\"\npath_2mask = \"../input/prostate-cancer-grade-assessment/train_label_masks/\"\npath_2train = \"../input/prostate-cancer-grade-assessment/train.csv\"\nremove_index = 2227\n#os.mkdir('/kaggle/working/train_imgs/')\n#os.mkdir('/kaggle/working/train_imgs_canny/')","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"#Analysing the files\ndata_train = pd.read_csv(path_2train)\ncols = data_train.columns\nprint(\"Shape ({0} , {1})\".format(len(data_train),len(cols)))\ndata_train.head()\ndata_train = data_train.drop(data_train.index[remove_index])\ndata_train = data_train.reset_index()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#Helper functions\ndef read_img(df, idx, level=2,show = False):\n    img_id = df['image_id'][idx]\n    provider = df['data_provider'][idx]\n    grade = df['isup_grade'][idx]\n    score = df['gleason_score'][idx]\n    image_path = path_2img + img_id + '.tiff'\n\n    img_im = MultiImage(image_path)\n    img_im = img_im[level]\n    if(show):\n        image_msk = path_2mask + img_id + '_mask.tiff'\n        img_msk = MultiImage(image_msk)\n        img_msk = img_msk[level]\n        img_msk = img_msk[:,:,0]\n        im_plot(img_im,\"Level 2\",level)\n        im_plot(img_msk,\"Level 0\",level)\n    return img_id,provider,grade,score,img_im\ndef im_plot(img,nm=''):\n    plt.figure(figsize=(10, 10))\n    plt.imshow(img)\n    plt.title(nm, fontsize=20)\n    return\ndef get_slice_index(arr, assign_val = 10000,debug = False):\n    ls1 = []\n    ls2 = []\n    for i in range(arr.shape[0]):\n        temp = np.where(arr[i,:] == False)\n        if(temp[0].shape[0]==0):\n            ls1.append(assign_val)\n            ls2.append(assign_val)\n        else:\n            ls1.append(temp[0][0])\n            ls2.append(temp[0][-1])\n    y1 = 0\n    y2 = arr.shape[0]\n    for i in range(len(ls1)):\n        if(ls1[i] != assign_val):\n            y1 = i\n            break\n    for i in range(len(ls2)-1,-1,-1):\n        if(ls2[i] != assign_val):\n            y2 = i\n            break\n    x1=np.min(ls1)\n    test_list = [assign_val]\n    #print(ls2)\n    res = [i for i in ls2 if i not in test_list] \n    \n    x2=np.max(res)\n    #print(ls1)\n    if(debug):\n        print(\"Boundries [x1:{0},y1:{1}],[x2:{2},y2:{3}]\".format(x1,y1,x2,y2))\n    return [x1,y1,x2,y2]\ndef crop_image(data):\n    list_index = get_slice_index(data,debug = False)\n    return data[list_index[1]:list_index[3], list_index[0]:list_index[2]]\ndef convert_2_gray(data):\n    data = rgb2gray(data)\n    thresh = threshold_otsu(data)\n    binary = data > thresh\n    #print(data)\n    #print(\"Size in Mb : \",binary.nbytes / 1000000)\n    return binary  \ndef delete_tree(pfad=\"/kaggle/working/train_imgs/\"):\n    shutil.rmtree(pfad)\n    return \ndef create_img_cropped(df,l=256,w=256):\n    for i in tqdm(range(len(df))):\n        id_img,_,_,_,data = read_img(df, i)\n        data = convert_2_gray(data)\n        crop_img = crop_image(data)\n        #crop_img = resize(crop_img, (l, w))\n        if(i%1000==0):\n            print(crop_img.shape)\n        Image.fromarray(crop_img).save(\"/kaggle/working/train_imgs/\"+id_img+\".tiff\")\n    return\ndef get_canny(df,sig=0.4):\n    return feature.canny(df, sigma=sig)\ndef get_shannon(df):\n    s_ent = []\n    for i in tqdm(range(len(df))):\n        id_img,_,_,_,data = read_img(df, i)\n        data = convert_2_gray(data)\n        crop_img = crop_image(data)\n        s_ent.append(shannon_entropy(crop_img))\n        \n    return s_ent\ndef feature_engineering(df):\n    df['shannon'] = get_shannon(df)\n    df['shannon'] = (df['shannon']-df['shannon'].min())/(df['shannon'].max()-df['shannon'].min())\n    return df\n\ndef plotLearningCurve(history,epochs):\n    epoch_range = range(1,epoch+1)\n    plt.plot(epoch_range,history.history['accuracy'])\n    plt.plot(epoch_range,history.history['val_accuracy'])\n    plt.title('Model Accuracy')\n    plt.ylabel('Accuracy')\n    plt.xlabel('Epochs')\n    plt.legend(['train','val'],loc='upper left')\n    plt.show()\n    epoch_range = range(1,epoch+1)\n    plt.plot(epoch_range,history.history['loss'])\n    plt.plot(epoch_range,history.history['val_loss'])\n    plt.title('Model loss')\n    plt.ylabel('loss')\n    plt.xlabel('Epochs')\n    plt.legend(['train','val'],loc='upper left')\n    plt.show()\n    return\n#data_train.drop(remove_index)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"create_img_cropped(data_train)\n#delete_tree()\n\n#print(data_train.shape)\n#_,_,_,_,test = read_img(data_train, 2227)\n#im_plot(test,nm='rgb')\n#test = convert_2_gray(test)\n#im_plot(test,nm='binary')\n#_,crop_img = crop_image(test)\n#im_plot(crop_img,nm='crop')\n\n#os.mkdir('/kaggle/working/test/')\n# remove the file\n\n#Image.fromarray(crop_img).save(\"test/test.png\")\n#data_train.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#building the tensorflow model\nimport tensorflow as tf\nfrom tensorflow.keras import Sequential\nfrom tensorflow.keras.layers import Flatten,MaxPool2D,Dense,Conv2D,Dropout,BatchNormalization\nfrom tensorflow.keras.optimizers import Adam\nfrom tensorflow.keras.preprocessing import image\nfrom sklearn.model_selection import train_test_split\nimport gc\n\nprint(tf.__version__)\ngc.collect()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"X1=[]\nfor i in tqdm(range(len(data_train))):\n        id_img,_,_,_,data = read_img(data_train, i)\n        data = convert_2_gray(data)\n        crop_img = crop_image(data)\n        X1.append(crop_img)\nX1 = np.reshape(X1.shape[0],X1.reshape[1],1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"id_img,_,_,_,data = read_img(data_train, i)\ndata = convert_2_gray(data)\ncrop_img = crop_image(data)\ncrop_img = resize(crop_img,(300,200))\ntype(crop_img)\nprint(crop_img)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"collapsed":true},"cell_type":"code","source":"data_train = pd.get_dummies(data=data_train, columns=['isup_grade'])\ndata_train = data_train.drop(['index','data_provider','gleason_score'],axis = 1)\ndata_train.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"X = []\npath_ = \"/kaggle/working/train_imgs/\"\next = \".tiff\"\nfor i in tqdm(range(len(data_train))):\n    temp = path_ + data_train['image_id'][i]+ ext\n    img = image.load_img(temp,target_size=(128,128))\n    #img = rgb2gray(img)\n    X.append(image.img_to_array(img)/255.0)\nX = np.array(X)\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"Y = data_train.drop(['image_id'],axis=1)\nY = Y.to_numpy()\nx_train,x_test,y_train,y_test=train_test_split(X,Y,random_state=0,test_size=0.2)\nX.shape","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#Build CNN\nmodel = Sequential()\nmodel.add(Conv2D(16,(3,3),activation='relu',input_shape=x_train[0].shape))\nmodel.add(BatchNormalization())\nmodel.add(MaxPool2D(2,2))\nmodel.add(Dropout(0.2))\n\nmodel.add(Conv2D(32,(3,3),activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(MaxPool2D(2,2))\nmodel.add(Dropout(0.2))\n\nmodel.add(Conv2D(64,(3,3),activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(MaxPool2D(2,2))\nmodel.add(Dropout(0.2))\n\nmodel.add(Conv2D(128,(3,3),activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(MaxPool2D(2,2))\nmodel.add(Dropout(0.2))\n\nmodel.add(Conv2D(256,(3,3),activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(MaxPool2D(2,2))\nmodel.add(Dropout(0.2))\n\n\nmodel.add(Flatten())\nmodel.add(Dense(128,activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.3))\n\nmodel.add(Dense(128,activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.3))\n\nmodel.add(Dense(6,activation='sigmoid'))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])\nmodel.fit(x_train,y_train,epochs=15,validation_data=(x_test,y_test))","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}