{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[],"dockerImageVersionId":28755,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"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\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../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        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session\n\n# Use the kagglehub client library to attach Kaggle resources like competitions, datasets, and models to your session\n# Learn more about kagglehub: https://github.com/Kaggle/kagglehub/blob/main/README.md\n\nimport kagglehub\n# kagglehub.dataset_download('<owner>/<dataset-slug>')","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# الخلية الأولى: تهيئة الجلسة وقراءة بيانات المسابقة\n\n!pip install -q pydicom\n\nimport os\nimport glob\nimport random\nfrom pathlib import Path\n\nimport numpy as np\nimport pandas as pd\nfrom PIL import Image\nimport pydicom\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom torch.utils.data import Dataset, DataLoader\nfrom torchvision import models, transforms\n\nfrom sklearn.model_selection import train_test_split\n\n# الإعدادات الأساسية\nSEED = 42\nIMAGE_SIZE = 224\nBATCH_SIZE = 4\nEPOCHS = 1\n\nrandom.seed(SEED)\nnp.random.seed(SEED)\ntorch.manual_seed(SEED)\n\nDEVICE = torch.device(\n    \"cuda\" if torch.cuda.is_available() else \"cpu\"\n)\n\n# مسار بيانات المسابقة الذي ظهر عندك\nDATA_DIR = Path(\n    \"/kaggle/input/competitions/rsna-knee-abnormality-detection\"\n)\n\nif not (DATA_DIR / \"train.csv\").exists():\n    raise FileNotFoundError(\n        \"لم أجد train.csv في المسار المحدد. تأكد من إضافة بيانات المسابقة.\"\n    )\n\n# قراءة الملفات\ntrain_df = pd.read_csv(DATA_DIR / \"train.csv\")\nseries_df = pd.read_csv(DATA_DIR / \"train_series.csv\")\ntest_df = pd.read_csv(DATA_DIR / \"test.csv\")\ntest_series_df = pd.read_csv(DATA_DIR / \"test_series.csv\")\nsample_submission = pd.read_csv(DATA_DIR / \"sample_submission.csv\")\n\nID_COL = \"StudyInstanceUID\"\nTARGETS = [\n    column for column in sample_submission.columns\n    if column != ID_COL\n]\n\nprint(\"تم تشغيل الخلية الأولى بنجاح\")\nprint(\"مسار البيانات:\", DATA_DIR)\nprint(\"بيانات التدريب:\", train_df.shape)\nprint(\"بيانات الاختبار:\", test_df.shape)\nprint(\"عدد التصنيفات:\", len(TARGETS))\nprint(\"الجهاز المستخدم:\", DEVICE)\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# الخلية الثانية: تجهيز صور DICOM وبيانات التدريب\n\n\ndef make_series_map(meta):\n    result = {}\n\n    for uid, group in meta.groupby(ID_COL):\n        group = group.copy()\n\n        if \"Fluid_Sensitive\" in group.columns:\n            group = group.sort_values(\n                \"Fluid_Sensitive\",\n                ascending=False\n            )\n\n        result[str(uid)] = str(\n            group.iloc[0][\"SeriesInstanceUID\"]\n        )\n\n    return result\n\n\ntrain_series_map = make_series_map(series_df)\ntest_series_map = make_series_map(test_series_df)\n\n\ndef read_middle_dicom(series_folder):\n    files = sorted(\n        glob.glob(str(series_folder / \"*.dcm\"))\n    )\n\n    if len(files) == 0:\n        return np.zeros(\n            (IMAGE_SIZE, IMAGE_SIZE),\n            dtype=np.float32\n        )\n\n    # اختيار الصورة الوسطى من سلسلة MRI\n    file_path = files[len(files) // 2]\n\n    try:\n        ds = pydicom.dcmread(\n            file_path,\n            force=True\n        )\n        image = ds.pixel_array.astype(np.float32)\n\n        if getattr(\n            ds,\n            \"PhotometricInterpretation\",\n            \"\"\n        ) == \"MONOCHROME1\":\n            image = image.max() - image\n\n        image = np.nan_to_num(image)\n        low, high = np.percentile(image, [1, 99])\n        image = np.clip(\n            (image - low) / (high - low + 1e-6),\n            0,\n            1\n        )\n\n        image = Image.fromarray(\n            (image * 255).astype(np.uint8)\n        )\n        image = image.resize(\n            (IMAGE_SIZE, IMAGE_SIZE)\n        )\n\n        return np.asarray(\n            image,\n            dtype=np.float32\n        ) / 255.0\n\n    except Exception:\n        return np.zeros(\n            (IMAGE_SIZE, IMAGE_SIZE),\n            dtype=np.float32\n        )\n\n\ndef get_study_image(uid, base_dir, series_map):\n    uid = str(uid)\n    series_uid = series_map.get(uid)\n\n    if series_uid is None:\n        image = np.zeros(\n            (IMAGE_SIZE, IMAGE_SIZE),\n            dtype=np.float32\n        )\n    else:\n        series_folder = (\n            base_dir / uid / series_uid\n        )\n        image = read_middle_dicom(series_folder)\n\n    # تحويل الصورة الرمادية إلى 3 قنوات\n    image = np.stack(\n        [image, image, image],\n        axis=0\n    )\n\n    return torch.tensor(\n        image,\n        dtype=torch.float32\n    )\n\n\n# الاحتفاظ بالدراسات التي لديها تسمية واحدة على الأقل\nlabeled_df = train_df[\n    train_df[TARGETS].notna().any(axis=1)\n].reset_index(drop=True)\n\ntrain_ids, valid_ids = train_test_split(\n    np.arange(len(labeled_df)),\n    test_size=0.20,\n    random_state=SEED\n)\n\ntrain_part = labeled_df.iloc[train_ids].reset_index(drop=True)\nvalid_part = labeled_df.iloc[valid_ids].reset_index(drop=True)\n\nnormalization = transforms.Normalize(\n    mean=[0.5, 0.5, 0.5],\n    std=[0.25, 0.25, 0.25]\n)\n\n\nclass KneeDataset(Dataset):\n    def __init__(\n        self,\n        frame,\n        folder_name,\n        series_map,\n        labeled=True\n    ):\n        self.frame = frame.reset_index(drop=True)\n        self.folder_name = folder_name\n        self.series_map = series_map\n        self.labeled = labeled\n\n    def __len__(self):\n        return len(self.frame)\n\n    def __getitem__(self, index):\n        row = self.frame.iloc[index]\n        uid = str(row[ID_COL])\n\n        x = get_study_image(\n            uid,\n            DATA_DIR / self.folder_name,\n            self.series_map\n        )\n        x = normalization(x)\n\n        if not self.labeled:\n            return x, uid\n\n        y = np.asarray(\n            row[TARGETS],\n            dtype=float\n        )\n        mask = ~np.isnan(y)\n        y = np.nan_to_num(y, nan=0.0)\n\n        return (\n            x,\n            torch.tensor(y, dtype=torch.float32),\n            torch.tensor(mask, dtype=torch.float32)\n        )\n\n\ntrain_dataset = KneeDataset(\n    train_part,\n    \"train_series\",\n    train_series_map,\n    labeled=True\n)\n\nvalid_dataset = KneeDataset(\n    valid_part,\n    \"train_series\",\n    train_series_map,\n    labeled=True\n)\n\ntest_dataset = KneeDataset(\n    test_df,\n    \"test_series\",\n    test_series_map,\n    labeled=False\n)\n\n# num_workers=0 أكثر أمانًا على جهاز CPU\ntrain_loader = DataLoader(\n    train_dataset,\n    batch_size=BATCH_SIZE,\n    shuffle=True,\n    num_workers=0\n)\n\nvalid_loader = DataLoader(\n    valid_dataset,\n    batch_size=BATCH_SIZE,\n    shuffle=False,\n    num_workers=0\n)\n\ntest_loader = DataLoader(\n    test_dataset,\n    batch_size=1,\n    shuffle=False,\n    num_workers=0\n)\n\nprint(\"تم تجهيز البيانات بنجاح\")\nprint(\"بيانات التدريب:\", len(train_dataset))\nprint(\"بيانات التحقق:\", len(valid_dataset))\nprint(\"بيانات الاختبار:\", len(test_dataset))\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# الخلية الثالثة: بناء النموذج وتدريبه\n\n# التأكد من وجود الجهاز وعدد دورات التدريب\nDEVICE = torch.device(\n    \"cuda\" if torch.cuda.is_available() else \"cpu\"\n)\nEPOCHS = 1\n\nprint(\"الجهاز المستخدم:\", DEVICE)\n\n# إنشاء نموذج ResNet18\nmodel = models.resnet18(weights=None)\nmodel.fc = nn.Linear(\n    model.fc.in_features,\n    len(TARGETS)\n)\nmodel = model.to(DEVICE)\n\n# إعداد طريقة التدريب\noptimizer = torch.optim.AdamW(\n    model.parameters(),\n    lr=0.0001,\n    weight_decay=0.0001\n)\n\n\ndef masked_loss(logits, labels, mask):\n    loss = F.binary_cross_entropy_with_logits(\n        logits,\n        labels,\n        reduction=\"none\"\n    )\n    return (loss * mask).sum() / mask.sum().clamp_min(1.0)\n\n\n# بدء التدريب\nfor epoch in range(EPOCHS):\n    model.train()\n    total_loss = 0.0\n\n    for images, labels, mask in train_loader:\n        images = images.to(DEVICE)\n        labels = labels.to(DEVICE)\n        mask = mask.to(DEVICE)\n\n        optimizer.zero_grad()\n\n        logits = model(images)\n        loss = masked_loss(\n            logits,\n            labels,\n            mask\n        )\n\n        loss.backward()\n        optimizer.step()\n\n        total_loss += loss.item() * images.size(0)\n\n    average_loss = total_loss / len(train_loader.dataset)\n\n    print(\n        f\"Epoch {epoch + 1}/{EPOCHS} - \"\n        f\"Loss: {average_loss:.4f}\"\n    )\n\n# حفظ النموذج\nmodel_path = \"/kaggle/working/knee_model.pt\"\ntorch.save(model.state_dict(), model_path)\n\nprint(\"تم التدريب بنجاح\")\nprint(\"تم حفظ النموذج في:\", model_path)\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# الخلية الرابعة: إنشاء submission.csv\n\nmodel.eval()\n\npredictions = []\nidentifiers = []\n\nwith torch.no_grad():\n    for images, uids in test_loader:\n        images = images.to(DEVICE)\n\n        logits = model(images)\n        probabilities = torch.sigmoid(logits)\n        probabilities = probabilities.cpu().numpy()[0]\n\n        identifiers.append(uids[0])\n        predictions.append(probabilities)\n\n# إنشاء جدول التوقعات\nsubmission = pd.DataFrame(\n    predictions,\n    columns=TARGETS\n)\n\nsubmission.insert(\n    0,\n    ID_COL,\n    identifiers\n)\n\n# ترتيب الأعمدة مثل ملف sample_submission.csv\nsubmission = submission[sample_submission.columns]\n\n# حفظ الملف\nsubmission_path = \"/kaggle/working/submission.csv\"\nsubmission.to_csv(\n    submission_path,\n    index=False\n)\n\nprint(submission)\nprint(\"تم إنشاء الملف بنجاح:\", submission_path)\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}