classify_inhale_exhale.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "4e6bc4ef-fb68-4378-bbf6-a08eb45b0e0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4d7bff84-7e42-4931-83f9-cebd7f28e042",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "70e6c5fb22d1457b95b175874eedb8ac",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "from sklearn.utils.class_weight import compute_class_weight\n",
    "from data.aware_raw import AwareRaw, AwareSpectrogram\n",
    "\n",
    "classes = [\n",
    "    'Control / healthy / no pulmonary disease',\n",
    "    'Asthma',\n",
    "    'CF',\n",
    "    'COPD'\n",
    "]\n",
    "\n",
    "dataset_raw = AwareRaw(\"data/AWARE_DATA_LABELS_2023-12-08_1611.csv\", \"data/id_map.csv\", \"data/aware_full_1704385505.db\", pickle_file=\"data/aware_segmented.pkl\")\n",
    "dataset = AwareSpectrogram(\n",
    "    dataset_raw, \n",
    "    target_classes=classes, \n",
    "    age_balanced=False, \n",
    "    output_demogr=False, \n",
    "    output_spiro_raw=False, \n",
    "    output_spiro_pred=False, \n",
    "    output_oscil_raw=False, \n",
    "    output_oscil_zscore=False, \n",
    "    output_disease_label=False,\n",
    "    output_inhale_exhale=True,\n",
    "    relative_change=False, \n",
    "    calibration=True, \n",
    "    averaged=False, \n",
    "    num_channels=3,\n",
    "    dim_order='BTCHW',\n",
    "    modality='ir'\n",
    ")\n",
    "imflip = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "54ce16e6-9faa-45a9-8673-038f464b607b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(18, 24000)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_1405226/4226547782.py:52: UserWarning: The palette list has more values (10) than needed (1), which may not be intended.\n",
      "  sns.scatterplot(\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 300x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/eric/miniconda3/envs/aware/lib/python3.11/site-packages/umap/umap_.py:1945: UserWarning: n_jobs value 1 overridden to 1 by setting random_state. Use no seed for parallelism.\n",
      "  warn(f\"n_jobs value {self.n_jobs} overridden to 1 by setting random_state. Use no seed for parallelism.\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(18, 24000)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_1405226/4226547782.py:76: UserWarning: The palette list has more values (10) than needed (1), which may not be intended.\n",
      "  sns.scatterplot(\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 300x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(18, 24000) (18,) (18,)\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "from torchview import draw_graph\n",
    "import math\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib import animation\n",
    "from sklearn.manifold import TSNE\n",
    "from umap import UMAP\n",
    "from sklearn.metrics import pairwise_distances\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "\n",
    "RANDOM_SEED = [4399,114514,1234,1024,304,1,2,3,4,5]\n",
    "# RANDOM_SEED = [1]\n",
    "BATCH_SIZE = 1000\n",
    "LEARN_RATE = 1e-3\n",
    "MAX_NUM_EPOCH = 10\n",
    "MODEL_NAME = \"cnn\"\n",
    "VISUALIZE_MODEL = False\n",
    "\n",
    "# data_loader = torch.utils.data.DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True)\n",
    "# inputs, labels = next(iter(data_loader))\n",
    "# print('Input Size:', inputs.size())\n",
    "\n",
    "LABEL_TO_NUM = {\n",
    "    'Control / healthy / no pulmonary disease': 0,\n",
    "    'Asthma': 1,\n",
    "    'CF': 2,\n",
    "    'COPD': 3,\n",
    "    'Others': 4\n",
    "}\n",
    "\n",
    "X = dataset.data['IR']\n",
    "X = np.array(X.to_list())\n",
    "X = X.reshape(X.shape[0],-1)\n",
    "y = dataset.data['isExhale'].to_numpy()\n",
    "group = dataset.data['AWARE STUDY ID:'].to_numpy()\n",
    "\n",
    "idx = (group==300)\n",
    "X,y,group = X[idx,:],y[idx],group[idx]\n",
    "\n",
    "def tsne_plot(X, y, legend, group=None, perplexity=30):\n",
    "    print(X.shape)\n",
    "    tsne = TSNE(n_components=2, verbose=0, perplexity=perplexity, max_iter=300, random_state=RANDOM_SEED[0])\n",
    "    tsne_results = tsne.fit_transform(X)\n",
    "    \n",
    "    results = {'tsne-2d-one': tsne_results[:,0],\n",
    "               'tsne-2d-two': tsne_results[:,1],\n",
    "               'isExhale': y,\n",
    "               'id': group}\n",
    "    \n",
    "    fig, ax = plt.subplots(figsize=(3,2))\n",
    "    sns.scatterplot(\n",
    "        x=\"tsne-2d-one\", y=\"tsne-2d-two\",\n",
    "        hue=\"id\",\n",
    "        style=\"isExhale\",\n",
    "        data=results,\n",
    "        palette=sns.color_palette(),\n",
    "        legend=False,\n",
    "    )\n",
    "#     plt.xlim([-15,15])\n",
    "#     plt.ylim([-15,15])\n",
    "    plt.title(f\"t-SNE Scatter Plot\\nPerplexity={perplexity}\")\n",
    "    plt.show()\n",
    "\n",
    "def umap_plot(X, y, legend, group=None, n_neighbors=15):\n",
    "    print(X.shape)\n",
    "    umap = UMAP(n_components=2, n_neighbors=n_neighbors, min_dist=0.1, random_state=RANDOM_SEED[0])\n",
    "    umap_results = umap.fit_transform(X)\n",
    "    \n",
    "    results = {'umap-2d-one': umap_results[:,0],\n",
    "               'umap-2d-two': umap_results[:,1],\n",
    "               'isExhale': y,\n",
    "               'id': group}\n",
    "    \n",
    "    fig, ax = plt.subplots(figsize=(3,2))\n",
    "    sns.scatterplot(\n",
    "        x=\"umap-2d-one\", y=\"umap-2d-two\",\n",
    "        hue=\"id\",\n",
    "        style=\"isExhale\",\n",
    "        data=results,\n",
    "        palette=sns.color_palette(),\n",
    "        legend=False,\n",
    "    )\n",
    "#     plt.xlim([-15,15])\n",
    "#     plt.ylim([-15,15])\n",
    "    plt.title(f\"UMAP Scatter Plot\\nN_neighbors={n_neighbors}\")\n",
    "    plt.show()\n",
    "\n",
    "# tsne_plot(inputs.view(inputs.size(0),-1).numpy(), labels.numpy(), ['In','Ex'])\n",
    "tsne_plot(X,y,['In','Ex'],group,perplexity=8)\n",
    "# umap_plot(inputs.view(inputs.size(0),-1).numpy(), labels.numpy(), ['In','Ex'])\n",
    "umap_plot(X,y,['In','Ex'],group,n_neighbors=8)\n",
    "\n",
    "print(X.shape, group.shape, y.shape)\n",
    "# df = pd.DataFrame(X)\n",
    "# display(df)\n",
    "# df.to_csv(\"feature_raw.csv\", header=False, index=False)\n",
    "\n",
    "# df = pd.DataFrame(group)\n",
    "# display(df)\n",
    "# df.to_csv(\"id.csv\", header=False, index=False)\n",
    "\n",
    "# df = pd.DataFrame(y.astype(int))\n",
    "# display(df)\n",
    "# df.to_csv(\"label.csv\", header=False, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "fdd4fe63-ad7a-411d-b5e0-ae8a3f7f84fb",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e0a1b7cbef294a808830ff27bafce80b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a90f276f88cd4795a2c375133b97f765",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "df562864144f44a7821fd50d3290ae06",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9eebe705fa78432e9284ec3befe22ecd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "127a899d26e74a68904cedc73258082f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Linear SVM</th>\n",
       "      <th>Logi Regr (L2, default)</th>\n",
       "      <th>Logi Regr (L1)</th>\n",
       "      <th>MLP</th>\n",
       "      <th>Naive Bayes</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.759898</td>\n",
       "      <td>0.757982</td>\n",
       "      <td>0.825032</td>\n",
       "      <td>0.724777</td>\n",
       "      <td>0.594508</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.725415</td>\n",
       "      <td>0.738186</td>\n",
       "      <td>0.842912</td>\n",
       "      <td>0.734994</td>\n",
       "      <td>0.607280</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.717949</td>\n",
       "      <td>0.732692</td>\n",
       "      <td>0.794872</td>\n",
       "      <td>0.724359</td>\n",
       "      <td>0.628846</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.726923</td>\n",
       "      <td>0.733974</td>\n",
       "      <td>0.856410</td>\n",
       "      <td>0.732051</td>\n",
       "      <td>0.623077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.762821</td>\n",
       "      <td>0.759615</td>\n",
       "      <td>0.856410</td>\n",
       "      <td>0.740385</td>\n",
       "      <td>0.590385</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Avg</th>\n",
       "      <td>0.738601</td>\n",
       "      <td>0.744490</td>\n",
       "      <td>0.835127</td>\n",
       "      <td>0.731313</td>\n",
       "      <td>0.608819</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Linear SVM  Logi Regr (L2, default)  Logi Regr (L1)       MLP  \\\n",
       "0      0.759898                 0.757982        0.825032  0.724777   \n",
       "1      0.725415                 0.738186        0.842912  0.734994   \n",
       "2      0.717949                 0.732692        0.794872  0.724359   \n",
       "3      0.726923                 0.733974        0.856410  0.732051   \n",
       "4      0.762821                 0.759615        0.856410  0.740385   \n",
       "Avg    0.738601                 0.744490        0.835127  0.731313   \n",
       "\n",
       "     Naive Bayes  \n",
       "0       0.594508  \n",
       "1       0.607280  \n",
       "2       0.628846  \n",
       "3       0.623077  \n",
       "4       0.590385  \n",
       "Avg     0.608819  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 2700x900 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from matplotlib.colors import ListedColormap\n",
    "\n",
    "from sklearn.datasets import make_circles, make_classification, make_moons\n",
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis\n",
    "from sklearn.ensemble import AdaBoostClassifier, RandomForestClassifier\n",
    "from sklearn.gaussian_process import GaussianProcessClassifier\n",
    "from sklearn.gaussian_process.kernels import RBF\n",
    "from sklearn.inspection import DecisionBoundaryDisplay\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "import pandas as pd\n",
    "from sklearn.manifold import TSNE\n",
    "import sklearn.metrics as M\n",
    "import seaborn as sns\n",
    "from data.aware_raw import KFoldSplitter\n",
    "from tqdm.notebook import tqdm\n",
    "\n",
    "RANDOM_SEED = 4399\n",
    "BATCH_SIZE = 2000\n",
    "\n",
    "splitter = KFoldSplitter(dataset, BATCH_SIZE, random_seed=RANDOM_SEED)\n",
    "\n",
    "df = pd.DataFrame()\n",
    "output = {\n",
    "#     \"KNN\": pd.DataFrame(),\n",
    "    \"Linear SVM\": pd.DataFrame(),\n",
    "#     \"RBF SVM\": pd.DataFrame(),\n",
    "    \"Logi Regr (L2, default)\": pd.DataFrame(),\n",
    "    \"Logi Regr (L1)\": pd.DataFrame(),\n",
    "#     \"Decision Tree\": pd.DataFrame(),\n",
    "#     \"Random Forest\": pd.DataFrame(),\n",
    "    \"MLP\": pd.DataFrame(),\n",
    "#     \"AdaBoost\": pd.DataFrame(),\n",
    "    \"Naive Bayes\": pd.DataFrame()\n",
    "}\n",
    "# cols = [\n",
    "#     'Age', 'Sex', 'Height', 'Weight',\n",
    "#     'FEV1', 'FVC', 'FEV1/FVC', 'FEF2575',\n",
    "#     'Diagnosis', 'Prediction'\n",
    "# ]\n",
    "\n",
    "names = [\n",
    "#     \"KNN\",\n",
    "    \"Linear SVM\",\n",
    "#     \"RBF SVM\",\n",
    "    \"Logi Regr (L2, default)\",\n",
    "    \"Logi Regr (L1)\",\n",
    "#     \"Decision Tree\",\n",
    "#     \"Random Forest\",\n",
    "    \"MLP\",\n",
    "#     \"AdaBoost\",\n",
    "    \"Naive Bayes\"\n",
    "]\n",
    "\n",
    "classifiers = [\n",
    "#     KNeighborsClassifier(5),\n",
    "    SVC(kernel=\"linear\", C=0.025, random_state=42),\n",
    "#     SVC(gamma=2, C=1, random_state=42),\n",
    "    LogisticRegression(max_iter=200),\n",
    "    LogisticRegression(penalty='l1', solver='liblinear', max_iter=200),\n",
    "#     DecisionTreeClassifier(max_depth=5, random_state=42),\n",
    "#     RandomForestClassifier(\n",
    "#         max_depth=5, n_estimators=10, max_features=1, random_state=42\n",
    "#     ),\n",
    "    MLPClassifier(alpha=1, max_iter=1000, random_state=42),\n",
    "#     AdaBoostClassifier(algorithm=\"SAMME\", random_state=42),\n",
    "    GaussianNB()\n",
    "]\n",
    "\n",
    "figure = plt.figure(figsize=(27, 9))\n",
    "i = 1\n",
    "# iterate over datasets\n",
    "for split_idx, (train_loader, test_loader) in enumerate(splitter):\n",
    "    for batch_idx, (inputs, labels) in enumerate(train_loader):\n",
    "        # X_train = np.concatenate((demogr, spiro), axis=1)\n",
    "        X_train = inputs.view(inputs.size(0),-1)\n",
    "        y_train = labels\n",
    "    for batch_idx, (inputs, labels) in enumerate(test_loader):\n",
    "        # X_test = np.concatenate((demogr, spiro), axis=1)\n",
    "        X_test = inputs.view(inputs.size(0),-1)\n",
    "        y_test = labels\n",
    "    \n",
    "    s = {}\n",
    "    for name, clf in tqdm(zip(names, classifiers)):\n",
    "        clf = make_pipeline(StandardScaler(), clf)\n",
    "        clf.fit(X_train, y_train)\n",
    "        y_pred = clf.predict(X_test)\n",
    "        bal_acc = M.balanced_accuracy_score(y_test, y_pred)\n",
    "        s[name] = bal_acc\n",
    "\n",
    "    s = pd.Series(s)\n",
    "    df = pd.concat([df, s.to_frame().T])\n",
    "df = df.reset_index(drop=True)\n",
    "df_avg = df.mean(axis=0).to_frame().T\n",
    "df_avg.index = ['Avg']\n",
    "display(pd.concat((df, df_avg)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9a179144-0325-4009-954f-8c915b9049e7",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8b23e7c6-ba9e-434d-8a75-855c989fca78",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (AWARE)",
   "language": "python",
   "name": "aware"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}