Merge pull request #54 from sensein/alistair/refactor_constants

tie together the various constants used for parsing the redcap CSV

docs/tutorial.ipynb

@@ -67,14 +67,16 @@
     "# every user has a sessionschema which we can get info for the users from\n",
     "qs = dataset.load_questionnaires('sessionschema')\n",
     "q_dfs = []\n",
-    "for subject_id, questionnaire in qs.items():\n",
+    "for i, questionnaire in enumerate(qs):\n",
     "    # get the dataframe for this questionnaire\n",
     "    df = dataset.questionnaire_to_dataframe(questionnaire)\n",
+    "    df['dataframe_number'] = i\n",
     "    q_dfs.append(df)\n",
+    "    i += 1\n",
     "\n",
     "# concatenate all the dataframes\n",
     "sessionschema_df = pd.concat(q_dfs)\n",
-    "sessionschema_df = pd.pivot(sessionschema_df, index='record_id', columns='linkId', values='valueString')\n",
+    "sessionschema_df = pd.pivot(sessionschema_df, index='dataframe_number', columns='linkId', values='valueString')\n",
     "sessionschema_df"
    ]
   },
@@ -108,7 +110,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "demographics_df = dataset.load_and_pivot_questionnaire('demographics')\n",
+    "demographics_df = dataset.load_and_pivot_questionnaire('qgenericdemographicsschema')\n",
     "demographics_df.head()"
    ]
   },
@@ -231,14 +233,18 @@
    "outputs": [],
    "source": [
     "session_schema = dataset.load_questionnaires('sessionschema')\n",
-    "# show the first item\n",
-    "record_id = list(session_schema.keys())[0]\n",
     "\n",
     "# Each element is a QuestionnaireResponse, a pydantic object\n",
     "# you can serialize it to a python dictionary with .dict()\n",
     "# and to a json with .json()\n",
     "# otherwise attributes are accessible like any other python object\n",
-    "print(session_schema[record_id].json(indent=2))"
+    "questionnaire = session_schema[0]\n",
+    "\n",
+    "print(f'FHIR ID: {questionnaire.id}')\n",
+    "print(f'First item response: {questionnaire.item[0]}')\n",
+    "print('\\nAbridged questionnaire as JSON:')\n",
+    "# only print the first 600 characters of the JSON for brevity\n",
+    "print(questionnaire.json(indent=2)[:660], end='\\n...\\n')"
    ]
   },
   {
@@ -247,7 +253,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# helper function which loads in the above as a dataframe\n",
+    "# helper function which loads in questionnaires ending with a specific schema name as a dataframe\n",
     "session_df = dataset.load_and_pivot_questionnaire('sessionschema')\n",
     "session_df.head()"
    ]
@@ -265,15 +271,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "session_confounders = dataset.load_questionnaires('confounders')\n",
-    "# show the first item\n",
-    "record_id = list(session_confounders.keys())[0]\n",
+    "session_confounders = dataset.load_questionnaires('qgenericconfoundersschema')\n",
+    "questionnaire = session_confounders[0]\n",
     "\n",
     "# Each element is a QuestionnaireResponse, a pydantic object\n",
     "# you can serialize it to a python dictionary with .dict()\n",
     "# and to a json with .json()\n",
     "# otherwise attributes are accessible like any other python object\n",
-    "print(session_confounders[record_id].json(indent=2))"
+    "print(questionnaire.json(indent=2))"
    ]
   },
   {
@@ -310,10 +315,7 @@
    "outputs": [],
    "source": [
     "acoustic_tasks = dataset.load_questionnaires('acoustictaskschema')\n",
-    "acoustic_tasks\n",
-    "# show the first item\n",
-    "record_id = list(acoustic_tasks.keys())[0]\n",
-    "print(acoustic_tasks[record_id].json(indent=2))"
+    "print(acoustic_tasks[0].json(indent=2))"
    ]
   },
   {
@@ -348,6 +350,92 @@
    "source": [
     "acoustic_tasks_df['acoustic_task_name'].value_counts()"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Audio"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pathlib import Path\n",
+    "import numpy as np\n",
+    "from b2aiprep.process import Audio, specgram\n",
+    "import IPython.display as Ipd\n",
+    "\n",
+    "# configuration options\n",
+    "win_length = 512\n",
+    "hop_length = 256\n",
+    "\n",
+    "base_path = Path('/Users/alistairewj/git/b2aiprep/output/')\n",
+    "audio_file = base_path.joinpath(\n",
+    "    'sub-1f9475bb-f13b-4f68-969b-28f20455b3e7',\n",
+    "    'ses-CB8A74EE-0C8C-4B15-B322-D93A79ADB40A',\n",
+    "    'audio',\n",
+    "    'sub-1f9475bb-f13b-4f68-969b-28f20455b3e7_ses-CB8A74EE-0C8C-4B15-B322-D93A79ADB40A_Audio-Check_rec-Audio-Check-1.wav'\n",
+    ")\n",
+    "audio = Audio.from_file(str(audio_file))\n",
+    "audio = audio.to_16khz()\n",
+    "\n",
+    "# convert to uint32 - probably should use the bits_per_sample from the original metadata!\n",
+    "signal = audio.signal.squeeze()\n",
+    "signal = (np.iinfo(np.uint32).max * (signal - signal.min())) / (signal.max() - signal.min())\n",
+    "# display a widget to play the audio file\n",
+    "Ipd.display(Ipd.Audio(data=signal, rate=audio.sample_rate))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We can also calculate a spectrogram of the data to visualize the frequency components over time. A spectrogram is essentially a frequency spectrum repeated `N` times spaced out throughout the original audio."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "win_length = 30\n",
+    "hop_length = 10\n",
+    "nfft = 512\n",
+    "features_specgram = specgram(audio, win_length=win_length, hop_length=hop_length, n_fft=nfft)\n",
+    "features_specgram = features_specgram.numpy()\n",
+    "# convert to db\n",
+    "log_spec = 10.0 * np.log10(np.maximum(features_specgram, 1e-10)).T\n",
+    "fig, ax = plt.subplots(1, 1)\n",
+    "ax.set_ylabel('Frequency (Hz)')\n",
+    "ax.matshow(log_spec, origin=\"lower\", aspect=\"auto\")\n",
+    "\n",
+    "xlim = ax.get_xlim()\n",
+    "xticks = ax.get_xticks()\n",
+    "xticklabels = [f\"{int(t * hop_length / 1000)}\" for t in xticks]\n",
+    "ax.set_xticks(xticks)\n",
+    "ax.set_xticklabels(xticklabels)\n",
+    "# reset the xlim, which may have been modified by setting the xticks\n",
+    "ax.set_xlim(xlim)\n",
+    "ax.set_xlabel('Time (s)')\n",
+    "\n",
+    "# y-axis is frequency\n",
+    "ylim = ax.get_ylim()\n",
+    "yticks = ax.get_yticks()\n",
+    "# convert yticks into frequencies\n",
+    "frequencies = yticks / nfft * audio.sample_rate\n",
+    "frequencies = [f\"{int(f)}\" for f in frequencies]\n",
+    "ax.set_yticks(yticks)\n",
+    "ax.set_yticklabels(frequencies)\n",
+    "ax.set_ylim(ylim)\n",
+    "\n",
+    "# Display the image\n",
+    "plt.show()\n"
+   ]
   }
  ],
  "metadata": {