python/validation/psfgeneration/validation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PSF Generation Validation\n",
    "\n",
    "This validation process involves the following:\n",
    "\n",
    "- Determine a set of PSF configurations to test\n",
    "- Create a PSF for each of those configurations using both Flowdec and PSFGenerator\n",
    "- Store any more detailed visualizations/statistics in a separate notebook for each configuration\n",
    "- Aggregate top-line statistics like SSIM and PSNR to visualize across all configurations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "import papermill as pm\n",
    "import os\n",
    "import os.path as osp\n",
    "import pandas as pd\n",
    "from flowdec import psf as fd_psf\n",
    "nb_dir = 'results'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "usage: ipykernel_launcher.py [-h] [--size-x SIZE_X] [--size-y SIZE_Y]\n",
      "                             [--size-z SIZE_Z] [--na NA]\n",
      "                             [--wavelength WAVELENGTH] [--m M] [--ns NS]\n",
      "                             [--ng0 NG0] [--ng NG] [--ni0 NI0] [--ni NI]\n",
      "                             [--ti0 TI0] [--tg0 TG0] [--tg TG]\n",
      "                             [--res-lateral RES_LATERAL]\n",
      "                             [--res-axial RES_AXIAL] [--pz PZ]\n",
      "                             [--num-basis NUM_BASIS]\n",
      "                             [--num-samples NUM_SAMPLES]\n",
      "                             [--oversampling OVERSAMPLING]\n",
      "                             [--min-wavelength MIN_WAVELENGTH]\n",
      "\n",
      "optional arguments:\n",
      "  -h, --help            show this help message and exit\n",
      "\n",
      "Dimensions:\n",
      "  --size-x SIZE_X       Number of pixels in result (x dimension) (default:\n",
      "                        256)\n",
      "  --size-y SIZE_Y       Number of pixels in result (y dimension) (default:\n",
      "                        256)\n",
      "  --size-z SIZE_Z       Number of pixels in result (z dimension) (default:\n",
      "                        128)\n",
      "\n",
      "Microscope Parameters:\n",
      "  --na NA               Numerical aperture of device (default: 1.4)\n",
      "  --wavelength WAVELENGTH\n",
      "                        Wavelength of emitted light in vacuum (microns)\n",
      "                        (default: 0.61)\n",
      "  --m M                 Magnification factor (default: 100)\n",
      "  --ns NS               Specimen refractive index (RI) (default: 1.33)\n",
      "  --ng0 NG0             Coverslip RI, design value (default: 1.5)\n",
      "  --ng NG               Coverslip RI, experimental (defaults to ng0 if not\n",
      "                        given) (default: None)\n",
      "  --ni0 NI0             Immersion RI, design value (default: 1.5)\n",
      "  --ni NI               Immersion RI, experimental (defaults to ni0 if not\n",
      "                        given) (default: None)\n",
      "  --ti0 TI0             Working distance (microns) (default: 150)\n",
      "  --tg0 TG0             Coverslip thickness, design value (microns) (default:\n",
      "                        170)\n",
      "  --tg TG               Coverslip thickness, experimental (microns) (defaults\n",
      "                        to tg0 if not given) (default: None)\n",
      "  --res-lateral RES_LATERAL\n",
      "                        Lateral pizel size / resolution (microns) (default:\n",
      "                        0.1)\n",
      "  --res-axial RES_AXIAL\n",
      "                        Axial pizel size / resolution (microns) (default:\n",
      "                        0.25)\n",
      "  --pz PZ               Particle distance from coverslip (microns) (default:\n",
      "                        2)\n",
      "\n",
      "Precision Parameters:\n",
      "  --num-basis NUM_BASIS\n",
      "                        Number of rescaled Bessels that approximate the phase\n",
      "                        function (default: 100)\n",
      "  --num-samples NUM_SAMPLES\n",
      "                        Number of pupil samples along radial direction\n",
      "                        (default: 1000)\n",
      "  --oversampling OVERSAMPLING\n",
      "                        Defines the upsampling ratio on the image space grid\n",
      "                        for computations (default: 2)\n",
      "  --min-wavelength MIN_WAVELENGTH\n",
      "                        Reference wavelength used in computation of scaling\n",
      "                        factor (microns); See section titled 'B. Bessel series\n",
      "                        approximation' in [1] for more details (default:\n",
      "                        0.436)\n"
     ]
    }
   ],
   "source": [
    "# Show default arguments and use these to derive test cases\n",
    "fd_psf.GibsonLanni().get_arg_parser().print_help()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build test case argument sets\n",
    "\n",
    "args = [\n",
    "    {\n",
    "        'name': 'particle-pos-p1um',\n",
    "        # \"num_samples\" is increased here from default of 1000\n",
    "        # because approximations are significantly further off otherwise\n",
    "        'args': {'psf_args': '{\"pz\": 1.0, \"num_samples\": 2000}'}   \n",
    "    }, {\n",
    "        'name': 'particle-pos-n1um',\n",
    "        'args': {'psf_args': '{\"pz\": -1.0, \"num_samples\": 2000}'}   \n",
    "    }, {\n",
    "        'name': 'red-light',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"wavelength\": 0.7}'}   \n",
    "    }, {\n",
    "        'name': 'blue-light',\n",
    "        # \"min_wavelength\" is lowered here from default of .436 microns\n",
    "        # because approximations are significantly further off otherwise\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"wavelength\": 0.45, \"min_wavelength\": 0.35}'}   \n",
    "    }, {\n",
    "        'name': 'green-light',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"wavelength\": 0.55, \"min_wavelength\": 0.35}'}   \n",
    "    }, {\n",
    "        'name': 'low-zres',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"res_axial\": 1.5, \"num_samples\": 2000}'}   \n",
    "    }, {\n",
    "        'name': 'hi-zres',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"res_axial\": 0.1}'}   \n",
    "    }, {\n",
    "        'name': 'large-xy',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"size_x\": 512, \"size_y\": 400, \"size_z\": 16}'}   \n",
    "    }, {\n",
    "        'name': 'large-z',\n",
    "        'args': {'psf_args': '{\"pz\": 0.0, \"size_x\": 32, \"size_y\": 64, \"size_z\": 128}'}   \n",
    "    },\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"particle-pos-p1um\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/particle-pos-p1um.ipynb\n",
      "100%|██████████| 27/27 [00:40<00:00,  1.49s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"particle-pos-n1um\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/particle-pos-n1um.ipynb\n",
      "100%|██████████| 27/27 [00:39<00:00,  1.47s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"red-light\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/red-light.ipynb\n",
      "100%|██████████| 27/27 [00:34<00:00,  1.28s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"blue-light\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/blue-light.ipynb\n",
      "100%|██████████| 27/27 [00:42<00:00,  1.56s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"green-light\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/green-light.ipynb\n",
      "100%|██████████| 27/27 [00:38<00:00,  1.44s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"low-zres\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/low-zres.ipynb\n",
      "100%|██████████| 27/27 [01:32<00:00,  3.42s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"hi-zres\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/hi-zres.ipynb\n",
      "100%|██████████| 27/27 [00:31<00:00,  1.17s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"large-xy\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/large-xy.ipynb\n",
      "100%|██████████| 27/27 [00:15<00:00,  1.73it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Executing comparisons for configuration \"large-z\"\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Input Notebook:  template.ipynb\n",
      "Output Notebook: results/large-z.ipynb\n",
      "100%|██████████| 27/27 [00:11<00:00,  2.45it/s]\n"
     ]
    }
   ],
   "source": [
    "# Run a comparison process for each configuration defined above\n",
    "os.makedirs(nb_dir, exist_ok=True)\n",
    "for arg in args:\n",
    "    print('Executing comparisons for configuration \"{}\"'.format(arg['name']))\n",
    "    nb_template = 'template.ipynb'\n",
    "    out_path = osp.join(nb_dir, arg['name'] + '.ipynb')\n",
    "    pm.execute_notebook(nb_template, out_path, parameters=arg['args'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "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>name</th>\n",
       "      <th>value</th>\n",
       "      <th>type</th>\n",
       "      <th>filename</th>\n",
       "      <th>key</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>psf_args</td>\n",
       "      <td>{\"pz\": 0.0, \"wavelength\": 0.45, \"min_wavelengt...</td>\n",
       "      <td>parameter</td>\n",
       "      <td>blue-light.ipynb</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>df_log</td>\n",
       "      <td>{'Diff': {'25%': -0.0016736984252929688, '50%'...</td>\n",
       "      <td>record</td>\n",
       "      <td>blue-light.ipynb</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>df_original</td>\n",
       "      <td>{'Diff': {'25%': -7.486232789233327e-08, '50%'...</td>\n",
       "      <td>record</td>\n",
       "      <td>blue-light.ipynb</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>measures</td>\n",
       "      <td>{'psnr_log': 40.984099215285056, 'psnr_origina...</td>\n",
       "      <td>record</td>\n",
       "      <td>blue-light.ipynb</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>psf_args</td>\n",
       "      <td>{\"pz\": 0.0, \"wavelength\": 0.55, \"min_wavelengt...</td>\n",
       "      <td>parameter</td>\n",
       "      <td>green-light.ipynb</td>\n",
       "      <td>green-light</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          name                                              value       type  \\\n",
       "0     psf_args  {\"pz\": 0.0, \"wavelength\": 0.45, \"min_wavelengt...  parameter   \n",
       "1       df_log  {'Diff': {'25%': -0.0016736984252929688, '50%'...     record   \n",
       "2  df_original  {'Diff': {'25%': -7.486232789233327e-08, '50%'...     record   \n",
       "3     measures  {'psnr_log': 40.984099215285056, 'psnr_origina...     record   \n",
       "4     psf_args  {\"pz\": 0.0, \"wavelength\": 0.55, \"min_wavelengt...  parameter   \n",
       "\n",
       "            filename          key  \n",
       "0   blue-light.ipynb   blue-light  \n",
       "1   blue-light.ipynb   blue-light  \n",
       "2   blue-light.ipynb   blue-light  \n",
       "3   blue-light.ipynb   blue-light  \n",
       "4  green-light.ipynb  green-light  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Aggregate results across all configurations\n",
    "df = pm.read_notebooks(nb_dir).dataframe\n",
    "df['key'] = df['key'].apply(lambda v: v.replace('.ipynb', ''))\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "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>value</th>\n",
       "      <th>key</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>psnr_log</th>\n",
       "      <td>40.984099</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>psnr_original</th>\n",
       "      <td>116.153428</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ssim_log</th>\n",
       "      <td>0.998636</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ssim_original</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>blue-light</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>psnr_log</th>\n",
       "      <td>66.781530</td>\n",
       "      <td>green-light</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    value          key\n",
       "psnr_log        40.984099   blue-light\n",
       "psnr_original  116.153428   blue-light\n",
       "ssim_log         0.998636   blue-light\n",
       "ssim_original    1.000000   blue-light\n",
       "psnr_log        66.781530  green-light"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def extract_series(df, prop):\n",
    "    df = df[df['name'] == prop].set_index('key')\n",
    "    return pd.concat([pd.Series(r['value']).rename('value').to_frame().assign(key=k) for k, r in df.iterrows()])\n",
    "df_meas = extract_series(df, 'measures')\n",
    "df_meas.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Scores\n",
    "\n",
    "This table shows SSIM and PSNR scores (between Flowdec and PSFGenerator) on the original PSF image scale as well as a logarithmic scale:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SSIM**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "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>ssim_log_scale</th>\n",
       "      <th>ssim_original_scale</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>key</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>blue-light</th>\n",
       "      <td>0.998636</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>green-light</th>\n",
       "      <td>0.999990</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>hi-zres</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-xy</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-z</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>low-zres</th>\n",
       "      <td>0.999932</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-n1um</th>\n",
       "      <td>0.999886</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-p1um</th>\n",
       "      <td>0.999886</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>red-light</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   ssim_log_scale  ssim_original_scale\n",
       "key                                                   \n",
       "blue-light               0.998636                  1.0\n",
       "green-light              0.999990                  1.0\n",
       "hi-zres                  1.000000                  1.0\n",
       "large-xy                 1.000000                  1.0\n",
       "large-z                  1.000000                  1.0\n",
       "low-zres                 0.999932                  1.0\n",
       "particle-pos-n1um        0.999886                  1.0\n",
       "particle-pos-p1um        0.999886                  1.0\n",
       "red-light                1.000000                  1.0"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([\n",
    "    df_meas.loc['ssim_log'].set_index('key')['value'].rename('ssim_log_scale'),\n",
    "    df_meas.loc['ssim_original'].set_index('key')['value'].rename('ssim_original_scale')\n",
    "], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**PSNR**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "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>psnr_log_scale</th>\n",
       "      <th>psnr_original_scale</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>key</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>blue-light</th>\n",
       "      <td>40.984099</td>\n",
       "      <td>116.153428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>green-light</th>\n",
       "      <td>66.781530</td>\n",
       "      <td>128.495587</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>hi-zres</th>\n",
       "      <td>159.330339</td>\n",
       "      <td>200.770418</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-xy</th>\n",
       "      <td>144.351244</td>\n",
       "      <td>205.861848</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-z</th>\n",
       "      <td>83.021613</td>\n",
       "      <td>137.138291</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>low-zres</th>\n",
       "      <td>65.626906</td>\n",
       "      <td>113.602541</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-n1um</th>\n",
       "      <td>44.805171</td>\n",
       "      <td>121.536585</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-p1um</th>\n",
       "      <td>44.805171</td>\n",
       "      <td>121.536585</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>red-light</th>\n",
       "      <td>108.406498</td>\n",
       "      <td>161.949216</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   psnr_log_scale  psnr_original_scale\n",
       "key                                                   \n",
       "blue-light              40.984099           116.153428\n",
       "green-light             66.781530           128.495587\n",
       "hi-zres                159.330339           200.770418\n",
       "large-xy               144.351244           205.861848\n",
       "large-z                 83.021613           137.138291\n",
       "low-zres                65.626906           113.602541\n",
       "particle-pos-n1um       44.805171           121.536585\n",
       "particle-pos-p1um       44.805171           121.536585\n",
       "red-light              108.406498           161.949216"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([\n",
    "    df_meas.loc['psnr_log'].set_index('key')['value'].rename('psnr_log_scale'),\n",
    "    df_meas.loc['psnr_original'].set_index('key')['value'].rename('psnr_original_scale')\n",
    "], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Maximum Differences\n",
    "\n",
    "This section shows the maximum difference for any single voxel value between Flowdec and PSFGenerator PSFs versus the range of the original PSF images:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_df(df, prop):\n",
    "    df = df[df['name'] == prop].set_index('key')\n",
    "    return pd.concat([pd.DataFrame(r['value']).assign(key=k) for k, r in df.iterrows()])\n",
    "\n",
    "df_stat = extract_df(df, 'df_original')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "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>Fowdec PSF Max Value</th>\n",
       "      <th>flowdec PSF Min Value</th>\n",
       "      <th>PSFGenerator PSF Max Value</th>\n",
       "      <th>PSFGenerator PSF Max Value</th>\n",
       "      <th>Absolute Max Difference</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>key</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>blue-light</th>\n",
       "      <td>1.0</td>\n",
       "      <td>9.857429e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>9.857429e-10</td>\n",
       "      <td>4.638507e-05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>green-light</th>\n",
       "      <td>1.0</td>\n",
       "      <td>6.726014e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>6.726014e-10</td>\n",
       "      <td>2.790959e-05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>hi-zres</th>\n",
       "      <td>1.0</td>\n",
       "      <td>3.078059e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.078059e-10</td>\n",
       "      <td>2.980232e-08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-xy</th>\n",
       "      <td>1.0</td>\n",
       "      <td>1.237140e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.237140e-10</td>\n",
       "      <td>1.490116e-08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>large-z</th>\n",
       "      <td>1.0</td>\n",
       "      <td>1.484220e-06</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.484220e-06</td>\n",
       "      <td>1.096936e-05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>low-zres</th>\n",
       "      <td>1.0</td>\n",
       "      <td>4.496030e-08</td>\n",
       "      <td>1.0</td>\n",
       "      <td>4.496030e-08</td>\n",
       "      <td>1.301039e-04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-n1um</th>\n",
       "      <td>1.0</td>\n",
       "      <td>7.155783e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>7.155783e-10</td>\n",
       "      <td>1.077533e-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>particle-pos-p1um</th>\n",
       "      <td>1.0</td>\n",
       "      <td>7.155785e-10</td>\n",
       "      <td>1.0</td>\n",
       "      <td>7.155785e-10</td>\n",
       "      <td>1.077533e-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>red-light</th>\n",
       "      <td>1.0</td>\n",
       "      <td>2.110213e-09</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.110213e-09</td>\n",
       "      <td>1.150649e-06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   Fowdec PSF Max Value  flowdec PSF Min Value  \\\n",
       "key                                                              \n",
       "blue-light                          1.0           9.857429e-10   \n",
       "green-light                         1.0           6.726014e-10   \n",
       "hi-zres                             1.0           3.078059e-10   \n",
       "large-xy                            1.0           1.237140e-10   \n",
       "large-z                             1.0           1.484220e-06   \n",
       "low-zres                            1.0           4.496030e-08   \n",
       "particle-pos-n1um                   1.0           7.155783e-10   \n",
       "particle-pos-p1um                   1.0           7.155785e-10   \n",
       "red-light                           1.0           2.110213e-09   \n",
       "\n",
       "                   PSFGenerator PSF Max Value  PSFGenerator PSF Max Value  \\\n",
       "key                                                                         \n",
       "blue-light                                1.0                9.857429e-10   \n",
       "green-light                               1.0                6.726014e-10   \n",
       "hi-zres                                   1.0                3.078059e-10   \n",
       "large-xy                                  1.0                1.237140e-10   \n",
       "large-z                                   1.0                1.484220e-06   \n",
       "low-zres                                  1.0                4.496030e-08   \n",
       "particle-pos-n1um                         1.0                7.155783e-10   \n",
       "particle-pos-p1um                         1.0                7.155785e-10   \n",
       "red-light                                 1.0                2.110213e-09   \n",
       "\n",
       "                   Absolute Max Difference  \n",
       "key                                         \n",
       "blue-light                    4.638507e-05  \n",
       "green-light                   2.790959e-05  \n",
       "hi-zres                       2.980232e-08  \n",
       "large-xy                      1.490116e-08  \n",
       "large-z                       1.096936e-05  \n",
       "low-zres                      1.301039e-04  \n",
       "particle-pos-n1um             1.077533e-03  \n",
       "particle-pos-p1um             1.077533e-03  \n",
       "red-light                     1.150649e-06  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_extrema = pd.concat([\n",
    "    df_stat.loc['max'].set_index('key')['Diff'].rename('max_diff'),\n",
    "    df_stat.loc['min'].set_index('key')['Diff'].rename('min_diff'),\n",
    "    df_stat.loc['max'].set_index('key')['Flowdec'].rename('Fowdec PSF Max Value'),\n",
    "    df_stat.loc['min'].set_index('key')['Flowdec'].rename('flowdec PSF Min Value'),\n",
    "    df_stat.loc['max'].set_index('key')['PSFGenerator'].rename('PSFGenerator PSF Max Value'),\n",
    "    df_stat.loc['min'].set_index('key')['PSFGenerator'].rename('PSFGenerator PSF Max Value'),\n",
    "], axis=1)\n",
    "df_extrema['Absolute Max Difference'] = df_extrema[['min_diff', 'max_diff']].abs().max(axis=1)\n",
    "df_extrema = df_extrema.drop(['min_diff', 'max_diff'], axis=1)\n",
    "df_extrema"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Dynamic Range of PSFs vs Largest Difference by Test Case')"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = df_extrema.sort_values('Absolute Max Difference').plot(figsize=(18, 6), kind='bar', legend='side')\n",
    "ax.set_yscale('log')\n",
    "ax.set_xlabel('Test Case')\n",
    "ax.set_ylabel('Voxel Value')\n",
    "ax.set_title('Dynamic Range of PSFs vs Largest Difference by Test Case')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}