Example_AJILE12_exploratory_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Example of identifying relationships between neural activity and movement behavior in AJILE12"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This script loads data directly from the DANDI remote repository using JupyterHub (https://hub.dandiarchive.org/). It can be adjusted to work on downloaded AJILE12 data files stored locally by using the commented out line in the cell after 'Load the data file' and commenting out all preceding lines in the cell."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Uncomment and run this cell if using JupyterHub\n",
    "# %pip install natsort\n",
    "# %pip install seaborn\n",
    "# %pip install nilearn\n",
    "# %pip install dandi\n",
    "# %pip install nwbwidgets==0.7.0\n",
    "# %pip install git+https://github.com/catalystneuro/brunton-lab-to-nwb.git\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import statsmodels.api as sm\n",
    "from scipy.signal import sosfiltfilt, butter, hilbert\n",
    "\n",
    "from pynwb import NWBHDF5IO\n",
    "from dandi.dandiapi import DandiAPIClient\n",
    "from ndx_events import LabeledEvents, AnnotatedEventsTable, Events\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## User-defined parameters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These parameters can be adjusted to analyze other electrodes, frequency bands, behavior types, participants, sessions, etc.\n",
    "\n",
    "We will select data from participant 1, session 3 only during times that the participant was eating (because this is an active behavior). ECoG data will be converted to spectral power in the gamma band (80-100 Hz) for electrode 7, which is located over the motor cortex. We will also look at the vertical velocity of the right wrist."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "sbj, session = 1, 3\n",
    "behavior_type = 'Eat'\n",
    "neural_freq_range = [80, 100]  # Hz\n",
    "ecog_ch_num = 7\n",
    "keypoint_of_interest = 'R_Wrist'\n",
    "pose_direction = 'vertical'  # 'vertical' or 'horizontal'\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the data file"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Data is loaded in directly from DANDI remote repository."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.9/site-packages/hdmf/spec/namespace.py:532: UserWarning: Ignoring cached namespace 'hdmf-common' version 1.4.0-alpha because version 1.5.0 is already loaded.\n",
      "  warn(\"Ignoring cached namespace '%s' version %s because version %s is already loaded.\"\n",
      "/opt/conda/lib/python3.9/site-packages/hdmf/spec/namespace.py:532: UserWarning: Ignoring cached namespace 'core' version 2.2.5 because version 2.4.0 is already loaded.\n",
      "  warn(\"Ignoring cached namespace '%s' version %s because version %s is already loaded.\"\n"
     ]
    }
   ],
   "source": [
    "with DandiAPIClient() as client:\n",
    "    asset = client.get_dandiset(\"000055\").get_asset_by_path(\n",
    "        \"sub-{0:>02d}/sub-{0:>02d}_ses-{1:.0f}_behavior+ecephys.nwb\".format(sbj, session)\n",
    "    )\n",
    "    s3_path = asset.get_content_url(follow_redirects=1, strip_query=True)\n",
    "\n",
    "io = NWBHDF5IO(s3_path, mode='r', load_namespaces=True, driver='ros3')\n",
    "# io = NWBHDF5IO(local_file_path, mode='r', load_namespaces=False)\n",
    "nwb = io.read()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Print information about selected electrode"
   ]
  },
  {
   "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>x</th>\n",
       "      <th>y</th>\n",
       "      <th>z</th>\n",
       "      <th>imp</th>\n",
       "      <th>location</th>\n",
       "      <th>filtering</th>\n",
       "      <th>group</th>\n",
       "      <th>group_name</th>\n",
       "      <th>standard_deviation</th>\n",
       "      <th>kurtosis</th>\n",
       "      <th>median_deviation</th>\n",
       "      <th>good</th>\n",
       "      <th>low_freq_R2</th>\n",
       "      <th>high_freq_R2</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></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>7</th>\n",
       "      <td>-53.896469</td>\n",
       "      <td>-29.059873</td>\n",
       "      <td>62.709102</td>\n",
       "      <td>NaN</td>\n",
       "      <td>unknown</td>\n",
       "      <td>250 Hz lowpass</td>\n",
       "      <td>GRID pynwb.ecephys.ElectrodeGroup at 0x1402148...</td>\n",
       "      <td>GRID</td>\n",
       "      <td>45.422554</td>\n",
       "      <td>2.707063</td>\n",
       "      <td>37.685382</td>\n",
       "      <td>True</td>\n",
       "      <td>0.103835</td>\n",
       "      <td>0.055118</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            x          y          z  imp location       filtering  \\\n",
       "id                                                                  \n",
       "7  -53.896469 -29.059873  62.709102  NaN  unknown  250 Hz lowpass   \n",
       "\n",
       "                                                group group_name  \\\n",
       "id                                                                 \n",
       "7   GRID pynwb.ecephys.ElectrodeGroup at 0x1402148...       GRID   \n",
       "\n",
       "    standard_deviation  kurtosis  median_deviation  good  low_freq_R2  \\\n",
       "id                                                                      \n",
       "7            45.422554  2.707063         37.685382  True     0.103835   \n",
       "\n",
       "    high_freq_R2  \n",
       "id                \n",
       "7       0.055118  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nwb.electrodes[ecog_ch_num]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data pre-processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Identify the start and stop times when the behavioral label of interest occurs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "min_len = 100  # (sec) only keep times when the given label appears for longer than this amount of time at once\n",
    "\n",
    "coarse_labels = nwb.intervals['epochs'].to_dataframe()\n",
    "coarse_labels = coarse_labels[coarse_labels['labels'].str.contains(behavior_type)]\n",
    "coarse_labels['diff'] = coarse_labels['stop_time'] - coarse_labels['start_time']\n",
    "coarse_labels = coarse_labels[coarse_labels['diff'] > min_len]\n",
    "coarse_labels.reset_index(inplace=True, drop=True)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the corresponding ECoG data for each behavioral label chunk and convert to spectral power via the Hilbert transform."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "filter_order = 4  # order of butterworth filter used to bandpass filter the ECoG data\n",
    "\n",
    "neural_data = nwb.acquisition['ElectricalSeries'].data\n",
    "sampling_rate = nwb.acquisition['ElectricalSeries'].rate  # (Hz) ECoG sampling rate\n",
    "neural_power = []\n",
    "for i in range(coarse_labels.shape[0]):\n",
    "    # Identify the start/end indices for each continuous chunk of the given behavioral label\n",
    "    start_t = int(coarse_labels.loc[i, 'start_time']*sampling_rate)\n",
    "    end_t = int(coarse_labels.loc[i, 'stop_time']*sampling_rate)\n",
    "    \n",
    "    # Load data snippet\n",
    "    neur_data_curr = neural_data[start_t:end_t, ecog_ch_num]\n",
    "    \n",
    "    # Bandpass filter\n",
    "    sos = butter(filter_order, neural_freq_range, btype='bandpass', output='sos', fs=sampling_rate)\n",
    "    neur_data_filtered = sosfiltfilt(sos, neur_data_curr)\n",
    "    \n",
    "    # Apply Hilbert transform and convert to decibels\n",
    "    neur_pow = np.abs(hilbert(neur_data_filtered))\n",
    "    neur_pow = 10*np.log(neur_pow)\n",
    "    \n",
    "    # Take the difference between neighboring timepoints\n",
    "    neural_power.append(np.diff(neur_pow))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the corresponding pose data for each behavioral label chunk and convert to vertical velocity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "keypoints = list(nwb.processing['behavior'].data_interfaces['Position'].spatial_series.keys())\n",
    "assert keypoint_of_interest in keypoints\n",
    "assert pose_direction in ['vertical', 'horizontal']\n",
    "keypoint_series = nwb.processing['behavior'].data_interfaces['Position'].spatial_series[keypoint_of_interest]\n",
    "sampling_rate_keypoint = keypoint_series.rate  # Hz\n",
    "keypoint_velocity = []\n",
    "for i in range(coarse_labels.shape[0]):\n",
    "    start_t = int(coarse_labels.loc[i, 'start_time']*sampling_rate_keypoint)\n",
    "    end_t = int(coarse_labels.loc[i, 'stop_time']*sampling_rate_keypoint)\n",
    "    \n",
    "    # Load pose data snippet\n",
    "    pose_data_curr = keypoint_series.data[start_t:end_t, :]\n",
    "    pose_mag_curr = pose_data_curr[:, 1 if pose_direction == 'vertical' else 0]\n",
    "    \n",
    "    # Convert to velocity (delta X / delta t)\n",
    "    velocity_curr = np.diff(pose_mag_curr)/(1/sampling_rate_keypoint)\n",
    "    keypoint_velocity.append(velocity_curr)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Align and combine neural and pose data into a pandas dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "assert len(neural_power) == len(keypoint_velocity)\n",
    "measures_all = []\n",
    "for i in range(len(neural_power)):\n",
    "    # Neural power for the given chunk\n",
    "    neur_curr = neural_power[i]\n",
    "    l_neur = len(neur_curr)\n",
    "    \n",
    "    # Pose velocity for the given chunk\n",
    "    accel_curr = keypoint_velocity[i]\n",
    "    l_accel = len(accel_curr)\n",
    "    \n",
    "    # Downsample neural data to match pose data\n",
    "    inds_split = np.array_split(np.arange(l_neur), l_accel)\n",
    "    for j, inds in enumerate(inds_split):\n",
    "        measures_all.append([neur_curr[inds].mean(), accel_curr[j]])\n",
    "\n",
    "# Combine neural/pose data into a dataframe\n",
    "df_measures_all = pd.DataFrame(np.asarray(measures_all), columns=['Neural power (dB)', 'Keypoint velocity (pixels/sec)'])\n",
    "\n",
    "# Remove any NaN's\n",
    "df_measures_all.dropna(inplace=True)\n",
    "\n",
    "# Remove instances with velocity close to 0\n",
    "df_measures_all = df_measures_all[(df_measures_all['Keypoint velocity (pixels/sec)'] > 100) |\\\n",
    "                                  (df_measures_all['Keypoint velocity (pixels/sec)'] < -100)]\n",
    "        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Print correlation between neural power and keypoint velocity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We find a small, positive correlation between neural power in the gamma band and right wrist vertical velocity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>Neural power (dB)</th>\n",
       "      <th>Keypoint velocity (pixels/sec)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Neural power (dB)</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.025379</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Keypoint velocity (pixels/sec)</th>\n",
       "      <td>0.025379</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                Neural power (dB)  \\\n",
       "Neural power (dB)                        1.000000   \n",
       "Keypoint velocity (pixels/sec)           0.025379   \n",
       "\n",
       "                                Keypoint velocity (pixels/sec)  \n",
       "Neural power (dB)                                     0.025379  \n",
       "Keypoint velocity (pixels/sec)                        1.000000  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_measures_all.corr(method='pearson')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Perform robust linear regression to quantify any linear relationships"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Regression identifies a small, but significant (p<0.05) positive relationship between neural power in the gamma band and right wrist vertical velocity. This result makes sense because moving one's wrist upward takes more effort (fighting against gravity) than moving one's arm downward and thus may require slightly more cortical control."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Robust linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>  <td>Neural power (dB)</td> <th>  No. Observations:  </th> <td>  6447</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>RLM</td>        <th>  Df Residuals:      </th> <td>  6445</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>               <td>IRLS</td>        <th>  Df Model:          </th> <td>     1</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Norm:</th>                <td>HuberT</td>       <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Scale Est.:</th>            <td>mad</td>        <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Cov Type:</th>              <td>H1</td>         <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>           <td>Thu, 03 Feb 2022</td>  <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>               <td>08:26:34</td>      <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>        <td>16</td>         <th>                     </th>    <td> </td>  \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "                 <td></td>                   <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>const</th>                          <td>   -0.0075</td> <td>    0.006</td> <td>   -1.251</td> <td> 0.211</td> <td>   -0.019</td> <td>    0.004</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Keypoint velocity (pixels/sec)</th> <td> 1.635e-05</td> <td> 6.62e-06</td> <td>    2.470</td> <td> 0.014</td> <td> 3.37e-06</td> <td> 2.93e-05</td>\n",
       "</tr>\n",
       "</table><br/><br/>If the model instance has been used for another fit with different fit parameters, then the fit options might not be the correct ones anymore ."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                    Robust linear Model Regression Results                    \n",
       "==============================================================================\n",
       "Dep. Variable:      Neural power (dB)   No. Observations:                 6447\n",
       "Model:                            RLM   Df Residuals:                     6445\n",
       "Method:                          IRLS   Df Model:                            1\n",
       "Norm:                          HuberT                                         \n",
       "Scale Est.:                       mad                                         \n",
       "Cov Type:                          H1                                         \n",
       "Date:                Thu, 03 Feb 2022                                         \n",
       "Time:                        08:26:34                                         \n",
       "No. Iterations:                    16                                         \n",
       "==================================================================================================\n",
       "                                     coef    std err          z      P>|z|      [0.025      0.975]\n",
       "--------------------------------------------------------------------------------------------------\n",
       "const                             -0.0075      0.006     -1.251      0.211      -0.019       0.004\n",
       "Keypoint velocity (pixels/sec)  1.635e-05   6.62e-06      2.470      0.014    3.37e-06    2.93e-05\n",
       "==================================================================================================\n",
       "\n",
       "If the model instance has been used for another fit with different fit parameters, then the fit options might not be the correct ones anymore .\n",
       "\"\"\""
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = df_measures_all['Keypoint velocity (pixels/sec)']\n",
    "Y = df_measures_all['Neural power (dB)']\n",
    "\n",
    "X = sm.add_constant(X)\n",
    "rlm_model = sm.RLM(Y, X, M=sm.robust.norms.HuberT())\n",
    "rlm_results = rlm_model.fit()\n",
    "rlm_results.summary()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot data with linear fit"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Most of the data appears clustered near 0 velocity. The positive relationship between neural power and right wrist vertical velocity is only barely visible. Additional steps that may help better understand the relationship between neural spectral power and wrist velocity include: removing pose data with abnormally high standard deviation due to noisy tracking, subtracting spectral power in nearby periods with minimal movement from ECoG spectral power, and manually reviewing pose trajectories to remove noisy tracking periods."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='Keypoint velocity (pixels/sec)', ylabel='Neural power (dB)'>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.regplot(data=df_measures_all, x='Keypoint velocity (pixels/sec)', y='Neural power (dB)', robust=True, ci=None)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}