ECG to SQL is a collection of python scripts to store electrocardiography (ECG) data into an SQLite database. The data used to develop this repo are provided by the PhysioNet/CinC challenge 2021 on the detection of atrial fibrillation.
The database structure is mainly based on the online analytical processing (OLAP) approach with the following entity
relationship diagram.
To ensure data integrity during entries insertion and minimize data redundancy, the design of this dataset satisfies the first three normal form rules. The ecg time series were stored into a single table per patient.
Aggregate all the metadata in csv files, one for each dataset
from data_access.prepare import prepare_summary_csv
prepare_summary_csv()Create and populate the dataset
from create_db import CreateDb
from pathlib import Path
from parameters import DefaultArguments
data_dir = Path('./path')
builder = CreateDb(data_dir, db_file_name='af_detection.db')
builder.populate_schema(dataset_name=DefaultArguments.ChapmanShaoxing)
builder.populate_data_tables(dataset_name=DefaultArguments.ChapmanShaoxing)
builder.db.close()Load the dataset and retrieve one patient data
from load_db import LoadDb
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
data_dir = Path('./path')
loader = LoadDb(data_dir=data_dir,
db_file_name='af.db')
original_id, diagnosis = loader.get_single_patient_data(patient_id=42)
ecg = loader.get_ecg(patient_id=42, leads=[1])
time = np.array(list(range(len(ecg)))) / loader.SAMPLING_FREQUENCY
plt.title(f'Patient id: {original_id} | diagnosis: {diagnosis}')
plt.xlabel('Time (s)')
plt.ylabel('mV')
plt.plot(time, ecg)
Distributed under the MIT License. See LICENSE.txt for more information.
Gaetano Scebba - @GScebba