GBBANet, electrocardiographic records database
GBBANet | DICARDIA

DICARDIA Database

If you wish to use this database in your research, you must cite these articles:

  1. Ledezma, C., Severeyn, E., Perpinan, G., Altuve, M., & Wong, S. (2014, August). A new on-line electrocardiographic records database and computer routines for data analysis. In Engineering in Medicine and Biology Society (EMBC), 2014 36th Annual International Conference of the IEEE (pp. 2738-2741). IEEE.
  2. Quintero, L. (2007). Desarrollo de una base de datos para la evaluación de la neuropatía autonómica cardíaca de origen diabético (Doctoral dissertation, Tesis de Maestría, Universidad Simón Bolívar, Coord. de Ing. Electrónica).

Diabetes mellitus is an illness characterized by the accumulation of glucose in the blood, nowadays it affects around 135 million people in the world. Diabetes is also one of the main causes of death worldwide.
According to numerous studies the number of diabetic people will rise 40% in developped countries and 170% in developping countries. These increments are due to the fact that there is no cure for diabetes, nevertheless it can be controled since the apperance of insuline in 1921.
Many research projects show the importance of some physiological parameters, signs and symptoms in the diagnostic of diabetes mellitus. Electrocardiographic changes have been reported in diabetic subjects, these changes could turn out to be indicators in the diagnostic of this illness.
It has been proved that diabetic complications start happening sooner in life than it was beleived, therefore the early diagnosticit is of vital importance if we want to treat and prevent serious complications. It is important to note that these complications can result in a serious cost to the patient and to the public healthcare system.
One of the most frequent, and at the same time more dangerous, alterations produced by diabetes is the cardiac autonomic neuropathy (CAN), which is a result of the deterioration of the autonomic nervous system's fibres which innervate the heart and the blood vessels' receptors. The direct consequence of the CAN is a disorder in the cardiac frequency control and in the vascular dynamic.
Certain works have revealed that patients suffering of CAN don't have a heart's physiological response to certain physical stress stimuli such as exercise. It has been proved that CAN may be diagnosed in the early stages of diabetes because there is a relation between the CAN and the cardiac frequency variability.
Even though the relation between CAN and the cardiovascular mortality is well known, we still don't fully understand this illness. It is therefore of great importance to dispose of a database of diabetic patients which present CAN, in order to study the different complications of this sickness.
It is so that the Biologic and Medical Computation Group (GCMB) and the Applied Biophysics and Bioengineering Group (GBBA) of the Simon Bolivar University (USB) along with collaborators from the Caracas University Hospital (HUC) formed , in 2007, the Group for Diabetic Cardiac Neuropathy Diagnostic and Modeling (DICARDIA). The objective of this group is to develop methods that help increment the diagnostic value of clinical tests in CAN. This project was divided in 3 stages. The first one was the design of a clinical protocol to collect and develop a database, the second stage is the extraction and processing of the collected data, the third stage is the presentation of the results. The database here presented is the result of the accomplishment first stage of this project.
The data source for this project were rutinary procedures that took place in the HUC. The main idea was to stablish a method to value the autonomous nervous system, the cardiovascular system and the glucose regulation system in the diabetic patient and hence contribute with the solution of a public health problem.
The collected data was introduced in a database called ecgML, which helps visualizing the demographic, clinical an signal characteristics with notes for each patient.

The clinical protocol was designed and applied in the cardiology service of the HUC, the study was composed by 65 subjects divided in three groups:

  • Diabetic with cardiac comlpications: 51 subjects of age 57 +/- 10 years old and weight 73 +/- 15 Kg
  • Diabetic without cardiac complications: 3 subjects of age 49 +/- 12 years old and weight 79 +/- 8 Kg
  • Control group: 11 subjects of age 50 +/- 6 years old and weight 81 +/- 20 Kg
These subjects (save the control group) are patients presenting diabetes mellitus type II. The control group was composed by asymptomatic patients with no known pathology and repose ECG with no alterations.
Demographic data was collected and saved for all patients.
The subjects underwent an effort test following the Bruce protocol to which was added a three minute warm-up stage. The test began with a 1,7 mph speed and 0° slope and finished with a 6mph speed and a 22° slope. In total there were 7 stages, 3 minute long each. For a non-trained subject the average test duration is 20 minutes, this means most subjects will finish the test between stages 3 and 5.
For this protocol it was taken into account the HbA1c, glycemia, cholesterol and triglycerides levels.
All the demographic data may be downloaded and consulted by using the ecgMLbrowser (see usefull programs section).
The signals used for this project were collected using the Ergocid-At-Plus effort test system. This is an equipment that allows the collection of 12 ECG derivations (500 samples per second) and the O2, CO2 and respiratory volume analysis, among other functions. This system was used to collect 8 ECG derivations during the effort test.
The data collection is serial, so the data stream produced by the equipment is as follows:

STREAM = S11 - S21 - S31 - S41 - S51 - S61 - S71 - S81 - S12 - S22 - S32 - ......... - S7N - S8N

Where SMN is the Nth sample of channel M.
It is so that the raw signal is given in the signal download section. In this database we provide certain programs that may be used to comprehend the structure and way of treatment of the signals.

ecgML

This is a method to describe a document by inserting tags in it specificaly designed for ECG data collection and analisis. The specifications are coded in XML language.
The ecgML structure allows to represent the data from the ECG following a hierarchical structure, taking into account the events, patient's data, diagnostic and other clinical data.
There are two main components for each file: i) The demographic data, containing information such as age, gender, etc; ii) the ECG registries with notations.
All the XML files downloaded from this database may be opened by using ecgML and following these simple instructions:

  1. Download the ecgMLbrowser application by clicking here
  2. Open the ecgMLbrowser and press the button file->open
  3. Select the registry you wish to open

    ecgML


  4. On the left side of the window you will find the structure panel, containing the clinical and demographical information, the clinical protocol and the notations on each P, QRS and T waves. On the right side you will find the waveform panel, showing 10 seconds of signal.



  5. If you wish to see the annotations on each heartbeat you must search on the structure panel the "wavenotations" . When double-clicking, the annotations will appear on the waveform panel.

ReadECG

This MATLAB function reads an ECG record from the DICARDIA database:

[signal t] = ReadECG(filename)



It opens the desired file and stores the 8 leads from the ECG in the independent variable 'signal'. The function also returns a time vector that may be used to plot the signals.
The input 'filename' is a sring containing the full path to the ECG record that will be read. The output 'signal' is an array of doubles containing 8 rows that correspond to the ECG derivations contained in the file. The rows contain the DI, DII, V1, V2, V3, V4, V5 and V6 derivations respectively. These signals have a 500Hz sampling frequency. The output 't' is a time axis that is coherent with the signals. It may be used to plot the signals having a time reference.

Make sure you have read the prior sections before using these signals. Their correct usage and results interpretation may depend on it.
To download the DICARDIA database signals click here.



Website designed, programmed and maintained by Carlos A. Ledezma

Creative Commons License
GBBANet by Grupo de Bioingeniería y Biofísica Aplicada - USB is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.