Merge branch 'main' of https://gitty.informatik.hs-mannheim.de/1826514/DSA_SS24

.gitignore

@@ -1 +1,2 @@
 /data/
+/settings.json

README.md

@@ -2,7 +2,12 @@
 
 This project was developed through the Data Science and Analytics course at the Mannheim University of Applied Sciences. A data science cycle was taught theoretically on the basis of lectures and implemented practically in the project. 
 
+<<<<<<< HEAD
 # Analysis of cardiovascular diseases using ECG data
+=======
+
+## Analysis of cardiovascular diseases using ECG data
+>>>>>>> bc4897dacb83ed2c3722b7b5d66b2fff23fdcb6a
 
 
 ## Table of Contents
@@ -15,6 +20,7 @@ This project was developed through the Data Science and Analytics course at the
 - [Acknowledgements](#acknowledgements)
 - [Contact](#contact)
 
+
 ## About
 
 Cardiovascular diseases refer to a group of diseases that affect the heart and blood vessels and represent a significant global health burden. They are a leading cause of morbidity and mortality worldwide, making effective prevention and management of these diseases critical. Physical examinations, blood tests, ECGs, stress or exercise tests, echocardiograms and CT or MRI scans are used to diagnose cardiovascular disease.
@@ -36,7 +42,7 @@ The data set used in this project was divided into four main groups: SB, AFIB, G
 The data provision provides for the following points, which can be taken from the diagram.
 
 
-![Alt-Text](readme_data/flow_diag.png)
+![Alt-Text](readme_data/Projektablauf.drawio.png)
 
 
 ## Getting Started
@@ -69,7 +75,7 @@ This project is licensed under the [MIT License](https://opensource.org/licenses
 We would like to especially thank our instructor, Ms. Jacqueline Franßen, for her enthusiastic support in helping us realize this project.
 
 ## Contact
-- Klara Tabea Bracke 
+- Klara Tabea Bracke (3015256@hs-mannheim.de)
 - Arman Ulusoy (3016148@stud.hs-mannheim.de)
-- Nils Rekus
+- Nils Rekus (1826514@stud.hs-mannheim.de)
 - Felix Jan Michael Mucha (felixjanmichael.mucha@stud.hs-mannheim.de)

filter_params.json

@@ -0,0 +1 @@
+{"names": ["butterlowpass", "lowess", "non_local_means"], "order": 1, "fs": 500.0, "cutoff": 25, "filter_strength": 50, "template_window_size": 7, "search_window_size": 21, "frac": 0.003, "it": 1}

notebooks/statistics.ipynb

@@ -0,0 +1,151 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Hypothesis\n",
+    "This notebook is used to read the data from the pickle files and to test the hypothesis that in the age group of 60-70 the frequency of a sinus bradycardia is significantly higher than in the other age groups.\n",
+    "For that instance the chi-squared test is used."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "import seaborn as sns\n",
+    "import matplotlib.pyplot as plt\n",
+    "import pickle\n",
+    "from scipy.stats import chi2_contingency\n",
+    "from data_helper import *\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Reading SB\n",
+      "Length of SB: 50\n",
+      "Reading AFIB\n",
+      "Length of AFIB: 27\n",
+      "Reading GSVT\n",
+      "Length of GSVT: 0\n",
+      "Reading SR\n",
+      "Length of SR: 13\n",
+      "Chi-Square Statistic: 38.266574797751275\n",
+      "P-value: 0.0004730210823940083\n",
+      "Chi-Square Statistic for SB in 60-70 vs others: 1.4858035714285718\n",
+      "P-value for SB in 60-70 vs others: 0.22286870264719977\n"
+     ]
+    }
+   ],
+   "source": [
+    "#path = \"C:/Studium/dsa/data\"\n",
+    "#path = \"C:/Users/Nils/Documents/HS-Mannheim/0000_MASTER/DSA/EKG_Prog/data\"\n",
+    "path = \"C:/Users/klara/projects/DSA/data\"\n",
+    "\n",
+    "categories_dict = {\n",
+    "'SB':    [426177001],\n",
+    "'AFIB':  [164889003, 164890007],\n",
+    "'GSVT':  [426761007, 713422000, 233896004, 233897008, 713422000],\n",
+    "'SR':    [426783006, 427393009]\n",
+    "}\n",
+    "\n",
+    "data = {}\n",
+    "for cat_name in categories_dict.keys():\n",
+    "    print(f\"Reading {cat_name}\")\n",
+    "    with open(f'{path}/{cat_name}.pkl', 'rb') as f:\n",
+    "        records = pickle.load(f)\n",
+    "        data[cat_name] = records\n",
+    "        print(f\"Length of {cat_name}: {len(records)}\")\n",
+    "\n",
+    "data_demographic = {'age':[], 'diag':[], 'gender':[]}\n",
+    "for cat_name, records in data.items():\n",
+    "    for record in records:\n",
+    "        age = record.comments[0].split(' ')[1]\n",
+    "        sex = record.comments[1].split(' ')[1]\n",
+    "        if age == 'NaN' or sex == 'NaN':\n",
+    "            continue\n",
+    "        # cut Age: from alter string \n",
+    "        data_demographic['age'].append(int(age))\n",
+    "        data_demographic['diag'].append(cat_name)\n",
+    "        data_demographic['gender'].append(sex)\n",
+    "\n",
+    "df_dgc = pd.DataFrame(data_demographic)\n",
+    "\n",
+    "# Change from group to category\n",
+    "age_categories = [0, 10, 20, 30, 40, 50, 60, 70, 80, 90]\n",
+    "df_dgc['age_group'] = pd.cut(df_dgc['age'], bins=age_categories)\n",
+    "corr_matrix_age_diag= pd.crosstab(df_dgc['age_group'], df_dgc['diag'])\n",
+    "\n",
+    "# Chi-square test\n",
+    "chi2, p, _, _ = chi2_contingency(corr_matrix_age_diag)\n",
+    "\n",
+    "# Difference between observed and expected frequencies\n",
+    "print(f\"Chi-Square Statistic: {chi2}\")\n",
+    "print(f\"P-value: {p}\")\n",
+    "\n",
+    "# Check if SB (Sinusbradykardie) has a significantly higher frequency in the 60-70 age group\n",
+    "sb_60_70 = corr_matrix_age_diag.loc[pd.Interval(60, 70, closed='right'), 'SB']\n",
+    "sb_other = corr_matrix_age_diag.drop(pd.Interval(60, 70, closed='right')).sum()['SB']\n",
+    "total_60_70 = corr_matrix_age_diag.loc[pd.Interval(60, 70, closed='right')].sum()\n",
+    "total_other = corr_matrix_age_diag.drop(pd.Interval(60, 70, closed='right')).sum().sum()\n",
+    "\n",
+    "# Frequency table for the specific Chi-Square test\n",
+    "observed = [[sb_60_70, total_60_70 - sb_60_70], [sb_other, total_other - sb_other]]\n",
+    "chi2_sb, p_sb = chi2_contingency(observed)[:2]\n",
+    "\n",
+    "\n",
+    "print(f\"Chi-Square Statistic for SB in 60-70 vs others: {chi2_sb}\")\n",
+    "print(f\"P-value for SB in 60-70 vs others: {p_sb}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The results can be interpreted as followed:\n",
+    "\n",
+    "- The first value returned is the Chi-Square Statistic that shows the difference between the observed and the expected frequencies. Here, a bigger number indicates a bigger difference. The p-value shows the probability of this difference being statistically significant. If the p-value is below the significance level of 0.05, the difference is significant.\n",
+    "\n",
+    "- The Chi-Square Statistic for sinus bradycardia in the age group 60-70 compared to the other age groups, is a value that shows whether there is a significant difference in the frequency of sinus bradycardia in the age group 60-70 in comparison to the other age groups. If the p-value is smaller than the significance level of 0.05, the difference in the frequency between the age group 60-70 and the other age groups is significant."
+   ]
+  },
+  {
+   "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": "3.11.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

scripts/data_helper.py

@@ -0,0 +1,234 @@
+import pickle
+import json
+import copy
+from matplotlib import pyplot as plt
+import numpy as np
+import wfdb.processing
+import scipy.signal
+from scipy.signal import butter, lfilter
+from statsmodels.nonparametric.smoothers_lowess import lowess
+import cv2 as cv
+
+""""
+TODO create overall description
+"""
+
+def load_data(only_demographic:bool=False, path_settings:str="../settings.json"):
+    """
+    Loads data from pickle files based on the specified settings.
+
+    Args:
+        only_demographic (bool, optional): If True, only loads demographic data (age, diagnosis, gender). Defaults to False.
+        path_settings (str, optional): Path to the settings file. Defaults to "./settings.json".
+
+    Returns:
+        dict: A dictionary containing the loaded data.
+    """
+    settings = json.load(open(path_settings))
+    path_data = settings["data_path"] 
+    labels = settings["labels"]
+
+    data = {}
+    if only_demographic:
+        data = {'age': [], 'diag': [], 'gender': []}
+
+    for cat_name in labels.keys():
+        print(f"Reading {cat_name}")
+        with open(f'{path_data}/{cat_name}.pkl', 'rb') as f:
+            records = pickle.load(f)
+            if only_demographic:
+                for record in records:
+                    age = record.comments[0].split(' ')[1]
+                    gender = record.comments[1].split(' ')[1]
+                    if age == 'NaN' or gender == 'NaN':
+                        continue       
+                    data['age'].append(int(age))
+                    data['diag'].append(cat_name)
+                    data['gender'].append(gender)
+            else:
+                data[cat_name] = records
+    return data
+
+
+def format_data_input(data):
+    """
+    Formats the input data into a standardized format.
+
+    Parameters:
+        data (np.ndarray or wfdb.Record or list or dict): The input data to be formatted.
+
+    Returns:
+        dict: The formatted data.
+
+    """
+    if isinstance(data, np.ndarray):
+        data = wfdb.Record(p_signal=data.copy())
+    if isinstance(data, wfdb.Record):
+        data = [data]
+    if isinstance(data, list):
+        temp_dict = {}
+        temp_dict['temp_key'] = data
+        data = temp_dict.copy()
+    return data
+
+
+def format_data_output(data):
+    """
+    Formats the output data into a less redundant format.
+
+    Args:
+        data (dict, list, wfdb.Record, or ndarray): The input data to be formatted.
+
+    Returns:
+        The formatted data.
+
+    """
+    if len(data.keys()) == 1 and 'temp_key' in data.keys():
+        data = data['temp_key']
+    if isinstance(data, list) and len(data) == 1:
+        data = data[0]
+    if isinstance(data, wfdb.Record) and len(data.p_signal.shape) == 1:
+        data = data.p_signal[0]
+    return data
+
+def butterlowpass_filter(data, cutoff:int, fs:int, order:int=5):
+    """
+    Apply a Butterworth lowpass filter to the input data.
+
+    Parameters:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        The input data to be filtered.
+    - cutoff: float
+        The cutoff frequency of the filter.
+    - fs: float
+        The sampling frequency of the input data.
+    - order: int, optional
+        The order of the filter (default is 5).
+
+    Returns:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        The filtered output data.
+
+    """
+    data = copy.deepcopy(data)
+    data = format_data_input(data)
+    for label, wfdb_objs in data.items():
+        for wfdb_obj in wfdb_objs:
+            for idx in range(wfdb_obj.p_signal.shape[1]):
+                signal = wfdb_obj.p_signal[:, idx]
+                nyq = 0.5 * fs
+                normal_cutoff = cutoff / nyq
+                b, a = butter(order, normal_cutoff, btype='low', analog=False)
+                wfdb_obj.p_signal[:, idx] = lfilter(b, a, signal)
+
+    return format_data_output(data)
+
+
+def lowess_filter(data, frac:float=0.03, it:int=1):
+    """
+    Applies the lowess filter to the given data.
+
+    Parameters:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        A dictionary containing the data to be filtered.
+    - frac (float): 
+        The fraction of the data used to compute each fitted value. Default is 0.03.
+    - it (int): 
+        The number of iterations for the smoothing process. Default is 1.
+
+    Returns:
+    (dict, list, wfdb.Record, or ndarray): The filtered data.
+
+    """
+    data = copy.deepcopy(data)
+    data = format_data_input(data)
+    for label, wfdb_objs in data.items():
+        for wfdb_obj in wfdb_objs:
+            for idx in range(wfdb_obj.p_signal.shape[1]):
+                signal = wfdb_obj.p_signal[:, idx]
+                d_range = np.arange(len(signal))
+                # [:, 1] needed to get only the smoothed values
+                wfdb_obj.p_signal[:, idx] = lowess(signal, d_range, is_sorted=True, frac=frac, it=it)[:, 1]
+    return format_data_output(data)
+
+
+def non_local_means_filter(data, filter_strength:int = 50, template_window_size:int = 7, search_window_size:int = 21):
+    """
+    Applies the Non-Local Means filter to the given data.
+
+    Parameters:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        A dictionary containing the data to be filtered.
+    - filter_strength (int): 
+        Parameter controlling the strength of the filtering process. Default is 50.
+    - template_window_size (int): 
+        Size in pixels of the template patch that is used to compute weights. Default is 7.
+    - search_window_size (int): 
+        Size in pixels of the window that is used to compute weighted average for given pixel. Default is 21.
+
+    Returns:
+    (dict, list, wfdb.Record, or ndarray): The filtered data.
+
+    """
+    data = copy.deepcopy(data)
+    data = format_data_input(data)
+    for label, wfdb_objs in data.items():
+        for wfdb_obj in wfdb_objs:
+            for idx in range(wfdb_obj.p_signal.shape[1]):
+                signal = wfdb_obj.p_signal[:, idx]
+                # reshape data to 2d for image like processing
+                d_2d = np.reshape(signal, (-1, 1))
+                # max min scaling
+                d_2d_scaled = np.uint8((d_2d - np.min(d_2d)) / (np.max(d_2d) - np.min(d_2d)) * 255)
+                # apply non local means filter
+                d_2d_filtered = cv.fastNlMeansDenoising(d_2d_scaled, None, filter_strength, template_window_size, search_window_size)
+                # Rescale the denoised signal back to the original range
+                d_filtered = np.reshape(d_2d_filtered, -1) * (np.max(signal) - np.min(signal)) / 255 + np.min(signal)
+                wfdb_obj.p_signal[:, idx] = d_filtered
+    return format_data_output(data)
+
+def filter_data(data, filter_params:dict):
+    """
+    Apply a filter to the input data.
+
+    Parameters:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        The input data to be filtered.
+    - filter_params: dict
+        The parameters of the filter to be applied.
+
+    Returns:
+    - data: (dict, list, wfdb.Record, or ndarray)
+        The filtered output data.
+
+    """
+    data = copy.deepcopy(data)
+    #data = format_data_input(data)
+    if 'butterlowpass' in filter_params['names']:
+        data = butterlowpass_filter(data, filter_params['cutoff'], filter_params['fs'], filter_params['order'])
+    if 'loess' in filter_params['names']:
+        data = lowess_filter(data, filter_params['frac'], filter_params['it'])
+    if 'non_local_means' in filter_params['names']:
+        data = non_local_means_filter(data, filter_params['filter_strength'], filter_params['template_window_size'], filter_params['search_window_size'])
+    if not any(name in filter_params['names'] for name in ['butterlowpass', 'loess', 'non_local_means']):
+        print("Warning: No valid filter names found in filter_params['names']. Data will be returned as is.")
+    return data #format_data_output(data)
+
+
+if __name__ == "__main__":
+    data = load_data(only_demographic=False, path_settings="./settings.json")
+    # print shape of data for each category
+    for cat_name in data.keys():
+        print(f"{cat_name}: {len(data[cat_name])}")
+
+
+    order = 1
+    fs = 500.0
+    cutoff = 25#25
+
+    # Apply filter to the signal
+    data_test = butterlowpass_filter(data, cutoff, fs, order)
+    data_test = butterlowpass_filter(data['SB'], cutoff, fs, order)
+    data_test = butterlowpass_filter(data['SB'][0], cutoff, fs, order)
+
+

settings.json

@@ -0,0 +1,11 @@
+{
+    "data_path_comment": "Path to the data folder. This is the folder where the data is stored.",
+    "data_path": "C:/Studium/dsa/data",
+    "labels_comment": "Labels for the different classes. The labels are the SNOMED CT codes.",
+    "labels": {
+        "SB":    [426177001],
+        "AFIB":  [164889003, 164890007],
+        "GSVT":  [426761007, 713422000, 233896004, 233897008, 713422000],
+        "SR":    [426783006, 427393009]
+        }
+}