eval_membrane.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "source": [
    "# computes the metrics between GT shells and predicted shells\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "import os\n",
    "from utils.utils import load_predictions, get_data, load_data, load_test_predictions, get_metrics, get_bbox_slices, get_bbox_margin, get_shells, compute_auc, get_path, train_test_auc"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:54:45.621221Z",
     "start_time": "2024-12-05T13:54:45.584452Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "execution_count": 25
  },
  {
   "cell_type": "code",
   "source": [
    "root_path = get_path()\n",
    "fig_path = os.path.join(root_path, \"figures\")\n",
    "\n",
    "data_path = os.path.join(root_path, \"RBC_labelled\")\n",
    "\n",
    "datasets = [\"valid1\", \"valid2\", \"valid3\"]"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:39:17.506194Z",
     "start_time": "2024-12-05T13:39:17.467754Z"
    }
   },
   "outputs": [],
   "execution_count": 6
  },
  {
   "cell_type": "code",
   "source": [
    "# load gt data\n",
    "data_dict = {}\n",
    "for dataset in datasets:\n",
    "    imgs, gt, file_names = load_data(data_path, dataset)\n",
    "    data_dict[dataset] = {\"imgs\": imgs,\n",
    "                          \"gt\": gt,\n",
    "                          \"file_names\": file_names}"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:39:23.270782Z",
     "start_time": "2024-12-05T13:39:18.088135Z"
    }
   },
   "outputs": [],
   "execution_count": 7
  },
  {
   "cell_type": "code",
   "source": [
    "# load prediction data\n",
    "seed = 0\n",
    "n_folds = 10\n",
    "eval_str = \"data_valid1_valid2_valid3_by_stack_mode_3D_iso_min_train_masks_0_seed_0_epoch_499_aniso_3.2\"\n",
    "for dataset in datasets:\n",
    "    test_pred, test_idx = load_test_predictions(data_path,\n",
    "                                     dataset,\n",
    "                                     data_dict[dataset][\"file_names\"],\n",
    "                                     eval_str,\n",
    "                                     seed=seed,\n",
    "                                     n_folds=n_folds)\n",
    "    reverse_idx = np.argsort(test_idx)\n",
    "    test_pred = test_pred[reverse_idx]\n",
    "    data_dict[dataset][\"pred\"] = test_pred"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:39:28.105282Z",
     "start_time": "2024-12-05T13:39:25.223733Z"
    }
   },
   "outputs": [],
   "execution_count": 8
  },
  {
   "cell_type": "code",
   "source": [
    "# test shell functionality\n",
    "img_id = 3\n",
    "pred_shell = get_shells(data_dict[\"valid1\"][\"pred\"][img_id], dilation_radius=0, erosion_radius=2)\n",
    "gt_shell = get_shells(data_dict[\"valid1\"][\"gt\"][img_id],  dilation_radius=0, erosion_radius=2)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:39:30.029297Z",
     "start_time": "2024-12-05T13:39:29.745690Z"
    }
   },
   "outputs": [],
   "execution_count": 9
  },
  {
   "cell_type": "code",
   "source": [
    "z=15\n",
    "fig, ax = plt.subplots(1,4)\n",
    "ax[0].imshow(data_dict[\"valid1\"][\"gt\"][img_id][z], interpolation=\"none\")\n",
    "ax[1].imshow(gt_shell[z], interpolation=\"none\")\n",
    "ax[2].imshow(data_dict[\"valid1\"][\"pred\"][img_id][z], interpolation=\"none\")\n",
    "ax[3].imshow(pred_shell[z], interpolation=\"none\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T13:39:31.832259Z",
     "start_time": "2024-12-05T13:39:31.426233Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f822762eaf0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 4 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 10
  },
  {
   "metadata": {},
   "cell_type": "code",
   "outputs": [],
   "execution_count": null,
   "source": "# load membrane metrics"
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-12-05T13:54:48.735463Z",
     "start_time": "2024-12-05T13:54:48.699340Z"
    }
   },
   "cell_type": "code",
   "source": [
    "cmap = matplotlib.cm.get_cmap(\"tab10\")\n",
    "thresholds = np.arange(11)/ 10\n",
    "stitch_threshold = 0.1\n",
    "n_folds = 10\n",
    "anisotropy = 3.2"
   ],
   "outputs": [],
   "execution_count": 26
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-12-05T13:54:49.399422Z",
     "start_time": "2024-12-05T13:54:49.394873Z"
    }
   },
   "cell_type": "code",
   "source": "",
   "outputs": [],
   "execution_count": null
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-12-05T14:08:00.686584Z",
     "start_time": "2024-12-05T14:07:56.928869Z"
    }
   },
   "cell_type": "code",
   "source": [
    "dataset = [\"valid1\", \"valid2\", \"valid3\"]\n",
    "\n",
    "cv_model_names =[\"cellpose_data_valid1_valid2_valid3_by_stack_mode_3D_iso_min_train_masks_0_seed_0_epoch_499\"\n",
    "                 ]\n",
    "ax, aps_train, aps_test = train_test_auc(cv_model_names,\n",
    "                                         data_path,\n",
    "                                         dataset,\n",
    "                                         thresholds=thresholds,\n",
    "                                         n_folds=n_folds,\n",
    "                                         stitch_threshold=stitch_threshold,\n",
    "                                         anisotropy=anisotropy,\n",
    "                                         return_metrics=True,\n",
    "                                         shells=True)\n",
    "\n",
    "plt.savefig(os.path.join(fig_path,\n",
    "                         \"metric_RBC_membrane_before_pp.pdf\"),\n",
    "            bbox_inches='tight',\n",
    "            pad_inches=0,\n",
    "            dpi=200)\n",
    "\n",
    "np.save(os.path.join(fig_path, \"metric_RBC_membrane_before_pp_split_fold_threshold.npy\"),\n",
    "        np.array(np.concatenate([aps_train, aps_test])))"
   ],
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best method: cellpose_data_valid1_valid2_valid3_by_stack_mode_3D_iso_min_train_masks_0_seed_0_epoch_499\n",
      "Best AUC: 0.625214392029327\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 33
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "# check if other shell widths and positions improve the metric (which they won't)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "source": [
    "# compute dilation and erosion radii fo a given shell with and position\n",
    "def get_d_e_radii(width, excess_dilation):\n",
    "    d_radius = np.ceil(width / 2) + excess_dilation\n",
    "    e_radius = np.floor(width / 2) - excess_dilation\n",
    "\n",
    "    d_radius = int(d_radius)\n",
    "    e_radius = int(e_radius)\n",
    "\n",
    "    assert d_radius + e_radius == width\n",
    "\n",
    "    return d_radius, e_radius\n",
    "\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:09:32.544732Z",
     "start_time": "2024-12-05T14:09:32.502696Z"
    }
   },
   "outputs": [],
   "execution_count": 34
  },
  {
   "cell_type": "code",
   "source": [
    "get_d_e_radii(1, -2)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:09:33.282141Z",
     "start_time": "2024-12-05T14:09:33.244258Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(-1, 2)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "execution_count": 35
  },
  {
   "cell_type": "code",
   "source": [
    "%%time\n",
    "# varies position and width of the gt and pred shell jointly, will always favour wide shells\n",
    "shell_widths = [1, 2, 3, 4]\n",
    "excess_dilations = [-1, 0, 1]\n",
    "\n",
    "\n",
    "shell_aucs = np.zeros(shape=(len(shell_widths), len(excess_dilations)))\n",
    "\n",
    "\n",
    "for i, width in enumerate(shell_widths):\n",
    "    for j, excess_dilation in enumerate(excess_dilations):\n",
    "        d_radius, e_radius = get_d_e_radii(width, excess_dilation)\n",
    "        gt_d_radius, gt_e_radius = get_d_e_radii(width, 0)\n",
    "        for dataset in data_dict.keys():\n",
    "            shell_segs = []\n",
    "            for seg in data_dict[dataset][\"pred\"]:\n",
    "                shell_segs.append(get_shells(seg, erosion_radius=e_radius, dilation_radius=d_radius))\n",
    "            data_dict[dataset][\"pred_shell\"] = np.stack(shell_segs)\n",
    "\n",
    "            gt_shell_segs = []\n",
    "            for seg in data_dict[dataset][\"gt\"]:\n",
    "                gt_shell_segs.append(get_shells(seg, erosion_radius=gt_e_radius, dilation_radius=gt_d_radius))\n",
    "            data_dict[dataset][\"gt_shell\"] = np.stack(gt_shell_segs)\n",
    "\n",
    "        shell_aucs[i,j] = compute_auc(data_dict, stage=\"pred_shell\", gt_stage=\"gt_shell\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:15:01.084571Z",
     "start_time": "2024-12-05T14:09:33.817027Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 5min 25s, sys: 675 ms, total: 5min 26s\n",
      "Wall time: 5min 27s\n"
     ]
    }
   ],
   "execution_count": 36
  },
  {
   "cell_type": "code",
   "source": [
    "plt.imshow(shell_aucs.T)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:15:05.057772Z",
     "start_time": "2024-12-05T14:15:01.191586Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f82266494c0>"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 37
  },
  {
   "cell_type": "markdown",
   "source": [
    "--> increasing the width of both the pred and gt shell always improves the results. Not shifting the shell inwards or outwards gives the best results"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "source": [
    "%%time\n",
    "\n",
    "# fix GT shell narrowly at 2 pix\n",
    "shell_widths = [1, 2, 3, 4]\n",
    "excess_dilations = [-1, 0, 1]\n",
    "\n",
    "\n",
    "shell_aucs = np.zeros(shape=(len(shell_widths), len(excess_dilations)))\n",
    "\n",
    "\n",
    "for i, width in enumerate(shell_widths):\n",
    "    for j, excess_dilation in enumerate(excess_dilations):\n",
    "        d_radius, e_radius = get_d_e_radii(width, excess_dilation)\n",
    "        gt_d_radius, gt_e_radius = get_d_e_radii(2, 0)\n",
    "        for dataset in data_dict.keys():\n",
    "            shell_segs = []\n",
    "            for seg in data_dict[dataset][\"pred\"]:\n",
    "                shell_segs.append(get_shells(seg, erosion_radius=e_radius, dilation_radius=d_radius))\n",
    "            data_dict[dataset][\"pred_shell\"] = np.stack(shell_segs)\n",
    "\n",
    "            gt_shell_segs = []\n",
    "            for seg in data_dict[dataset][\"gt\"]:\n",
    "                gt_shell_segs.append(get_shells(seg, erosion_radius=gt_e_radius, dilation_radius=gt_d_radius))\n",
    "            data_dict[dataset][\"gt_shell\"] = np.stack(gt_shell_segs)\n",
    "\n",
    "        shell_aucs[i,j] = compute_auc(data_dict, stage=\"pred_shell\", gt_stage=\"gt_shell\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:19:06.584505Z",
     "start_time": "2024-12-05T14:15:05.111378Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 4min, sys: 476 ms, total: 4min 1s\n",
      "Wall time: 4min 1s\n"
     ]
    }
   ],
   "execution_count": 38
  },
  {
   "cell_type": "code",
   "source": [
    "plt.imshow(shell_aucs.T)\n",
    "plt.xlabel(\"width\")\n",
    "plt.xticks(shell_widths)\n",
    "\n",
    "plt.ylabel(\"excess dilation\")\n",
    "plt.yticks([0, 1, 2], excess_dilations)\n",
    "plt.xticks([0,1,2,3], shell_widths )"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:19:06.764366Z",
     "start_time": "2024-12-05T14:19:06.676966Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([<matplotlib.axis.XTick at 0x7f82265954f0>,\n",
       "  <matplotlib.axis.XTick at 0x7f82265954c0>,\n",
       "  <matplotlib.axis.XTick at 0x7f82265f7a30>,\n",
       "  <matplotlib.axis.XTick at 0x7f82265b8070>],\n",
       " [Text(0, 0, '1'), Text(1, 0, '2'), Text(2, 0, '3'), Text(3, 0, '4')])"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 39
  },
  {
   "cell_type": "markdown",
   "source": [
    "--> dilation_radius = 1 = erosion_radius seems best"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "source": [
    "print(shell_aucs)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-12-05T14:19:06.839527Z",
     "start_time": "2024-12-05T14:19:06.802038Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.30038016 0.35892961 0.09819714]\n",
      " [0.29621506 0.62611367 0.29725047]\n",
      " [0.53310468 0.47255529 0.21815868]\n",
      " [0.45445698 0.43235116 0.34661683]]\n"
     ]
    }
   ],
   "execution_count": 40
  },
  {
   "metadata": {},
   "cell_type": "code",
   "outputs": [],
   "execution_count": null,
   "source": ""
  }
 ],
 "metadata": {
  "kernelspec": {
   "name": "conda-env-kahrp-py",
   "language": "python",
   "display_name": "Python [conda env:kahrp]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}