Epicardially attached cardiac accelerometer data from canines and porcines

Abstract

The database contains acceleration signals from accelerometers attached directly on the epicardium in canines and porcines during various interventions that altered cardiac function and myocardial motion. The tri-axial accelerometer was attached in the anterior, apical region of the left ventricle (LV). The database contains a total of 289 recordings from 19 canines and 27 porcines with an average duration of 15 seconds, ranging from 5 to 30 seconds. Acceleration data was recorded simultaneously with ECG and LV pressure (LVP).

Background

In the past decade, accelerometers have been miniaturized enough so they can be incorporated in pacing leads attached to the heart which allows the possibility to measure cardiac motion in addition to the pacing capability. Currently, two different companies offer this commercially in cardiac resynchronization therapy devices [1,2]. Furthermore, our group has proposed and patented the use of accelerometers incorporated in temporary pacing leads which are routinely attached to the epicardium during open chest cardiac surgery [3]. The heart is stopped during the surgery, and when the heart is restarted, the patient may require pacing to get a regular or faster heart rate to increase cardiac output during the recovery phase up to a week following surgery when the patient is still under post-operative care in the hospital. The pacing leads are pulled out through the skin before the patient leaves the hospital. Thus, the incorporation of a miniaturized accelerometer on the pacing lead attached to the heart, facilitates direct measurements of the heart motion. As the motion is directly linked to cardiac function, this method provides a means for continuous monitoring of cardiac function in the critical post-operative phase.

As part of the research of this method, our group has performed several animal studies with different protocols to investigate various aspects of the technology. In a new retrospective study using artificial intelligence to analyze the accelerometers signal, we pooled the data from some of our previous studies. The data extracted from the previous studies were simultaneously recorded tri-axial accelerometer signal, LVP measured with micromanometer, and ECG from precordial leads. A deep neural network was trained to automatically detect the aortic valve opening (AVO) and closing (AVC) time points using only the accelerometer signal. LVP measurements were used as reference labels for AVO and AVC as the time points of maximum and minimum LV dP/dt, respectively. ECG R-peaks were also detected in the labeling-process for detection of heart-beats. The pooled data in this database is the data used for this AVO and AVC detection study and consists of a total of 289 recordings from 19 canines and 27 porcines with an average duration of 15 seconds and ranging from 5 to 30 seconds.

Methods

The data was recorded during different experimental protocols performed at Oslo University Hospital, Oslo, Norway [4,5,6,7]. All protocols were approved by the Norwegian Food Safety Authority (FOTS), and all experiments were carried out in accordance with Norwegian regulations concerning the use of animals in experiments. The experiments followed different instrumentation protocols, but all of them had at least LVP and acceleration signals recorded simultaneously during different interventions. LVP was recorded using a calibrated micromanometer-tipped catheter (MPC-500, Millar Instruments Inc, Houston, TX).

A tri-axial accelerometer sensor (MPU9250, InvenSense Inc, San Jose, CA, USA) sutured to the epicardium in the LV apical, anterior region was used to measure the acceleration. The accelerometer's x-, y-, and z-axis were aligned with the longitudinal (positive x-direction towards the apex), circumferential (positive y-axis towards the LV lateral wall), and radial directions (positive z-axis pointing outward), respectively. ECG was recorded from 3 precordial leads. Depending on the protocol, data were recorded at either 650 Hz or 1000 Hz in canines and either 250 Hz or 500 Hz in porcines. However, for this dataset all data has been resampled to 500 Hz for consistency.

The static gravity component of the acceleration signal was also removed with a moving average filter (Tukey window of length 3 s with a cosine fraction of 0.5). Lastly, the accelerometer sensor was calibrated to a unit of g and normalized. Most recordings were performed under open chest conditions with three exceptions that were performed with a closed chest in some porcines and are explicitly labelled “with closed chest” in the data description.

Data Description

Not all interventions were performed in all animals due to differences in protocols. Some animals have multiple recordings for some interventions, for example during fluid loading recordings could be obtained at different levels of loading.

The lists below describe the interventions with the total number of animals (A) and the total number of recordings (S) for each intervention.

In canine experiments, data were obtained from the following interventions:

1. Baseline (A=19, S=34)
2. Right ventricular pacing (A=10, S=11)
3. Infusion of dobutamine (A=4, S=8)
4. Induction of ischemia (ischemia was induced by temporary occlusion of the proximal left anterior descending coronary artery (LAD)) (A=3, S=4)
5. Induction of left bundle branch block (LBBB) (LBBB was induced by radio-frequency ablation of the left bundle branch) (A=17, S=25)
6. Bi-ventricular pacing, i.e. cardiac resynchronization therapy (CRT) (A=11, S=38)
7. Combining LBBB with infusion of dobutamine (lbbbdob) (A=2, S=2)
8. Combining LBBB with induction of ischemia (lbbbisc) (A=3, S=3)
9. Combining LBBB with fluid loading (lbbbloading) (A=2, S=5)

In the porcine experiments, data were obtained from the following interventions:

1. Baseline (A=25, S=25)
2. Infusion of adrenaline (epinephrine 10 mg) (A=14, S=14)
3. Infusion of beta-blocker (esmolol, 100 mg) (A=13, S=13)
4. Infusion of vasodilator (niprid, 0.1 mg) (A=13, S=13)
5. Ischemia induced as described above (A=3, S=3)
6. Fluid loading (A=12, S=12)
7. Fluid loading with closed chest (A=9, S=25)
8. Phlebotomy (i.e. unloading) with closed chest (A=9, S=18)
9. Baseline with closed chest (A=7, S=8)
10. Infusion of dobutamine (A=3, S=3)
11. Infusion of dobutamine during ischemia (ischemiadob) (A=3, S=3)

Data is divided into 5 different folders namely AP, MP, MKCMS, MK and MV, corresponding to 5 different sets of experimental protocols that were carried out at our institute. AP, MP and MKCMS are porcine experiment series, while MK and MV are canine experiment series. All of these series' folders then have an individual folder for each individual animal (e.g. AP106, AP107 etc.).

Each individual animal folder has individual folders for the interventions performed for that particular animal. Each individual intervention folder contains individual files for each recording pertaining to that particular intervention for that particular animal. Each data file is a json file with left ventricular pressure (LVP), longitudinal acceleration (acc_x), circumferential acceleration (acc_y), radial acceleration (acc_z), ECG (if available), experiment name, animal species, sample rate, type of intervention, and identifier.

Furthermore, processed data used for our publication [8], with marked labels of LV dP/dt (minimum) and LV dP/dt (maximum) representing the ground truth for AVO and AVC, respectively, is present on github [9], along with the python scripts to reproduce those marked labels.

Usage Notes

Miniaturized accelerometers incorporated in pacing leads attached to the myocardium are used to monitor cardiac function. For this purpose functional indices must be extracted from the acceleration signal. A method that automatically detects the time of aortic valve opening (AVO) and aortic valve closure (AVC) will be helpful for such extraction.

In a study using this data, we therefore developed a deep learning model to automatically detect the aortic valve opening (AVO) and closing (AVC) time points using only the accelerometer signal [8]. The Python code used for this study is available on GitHub [9].

One limitation to the data, is that ECG was not available for all the recordings, therefore some files will not have the ECG readings. Furthermore, the data has been resampled to 500 Hz and is no longer in its raw format.

Ethics

All protocols were approved by the Norwegian Food Safety Authority (FOTS), and all experiments were carried out in accordance with Norwegian regulations concerning the use of animals in experiments. FOTS ID for the three protocols are:

• FOTS ID: 17644 (Canines)
• FOTS ID: 8628 (Porcines)
• FOTS ID: 9303 (Porcines)

Conflicts of Interest

Ali Wajdan, Tollef Struksnes Jahren, Manuel Villegas-Martinez, Faraz H Khan, Andreas Espinoza and Hans Henrik Odland report no conflict of interest. Ole Jakob Elle and Per Steinar Halvorsen are patent holders of the accelerometer technology for assessment of cardiac function, and together with Espen W. Remme and Magnus R. Krogh are shareholders in Cardiaccs A/S, which is commercially exploiting cardiac accelerometers.

References

1. Chinitz, Larry, et al. "Accelerometer-based atrioventricular synchronous pacing with a ventricular leadless pacemaker: results from the Micra atrioventricular feasibility studies." Heart rhythm 15.9 (2018): 1363-1371.
2. Senoner, Thomas, et al. "Long-term performance of an atrial lead capable of accelerometer based detection of cardiac contractility in patients receiving cardiac resynchronisation therapy." Plos one 14.9 (2019): e0222269.
3. Cardiaccs website. https://www.cardiaccs.com/ [Accessed 22 Dec 2022].
4. Krogh, Magnus Reinsfelt, et al. "Continuous estimation of acute changes in preload using epicardially attached accelerometers." IEEE Transactions on Biomedical Engineering 68.7 (2020): 2067-2075.
5. Aalen, John M., et al. "Mechanism of abnormal septal motion in left bundle branch block: role of left ventricular wall interactions and myocardial scar." JACC: Cardiovascular Imaging 12.12 (2019): 2402-2413.
6. Wajdan, Ali, et al. "Automatic detection of valve events by epicardial accelerometer allows estimation of the left ventricular pressure trace and pressure–displacement loop area." Scientific Reports 10.1 (2020): 1-11.
7. Villegas-Martinez, Manuel, et al. "Tracking Early Systolic Motion for Assessing Acute Response to Cardiac Resynchronization Therapy in Real Time." Frontiers in Physiology (2022): 1075.
8. Wajdan, Ali, et al. "Automatic Detection of Aortic Valve Events Using Deep Neural Networks on Cardiac Signals From Epicardially Placed Accelerometer." IEEE Journal of Biomedical and Health Informatics (2022).
9. Source code for recreating 'Automatic Detection of Aortic Valve Events Using Deep Neural Networks on Cardiac Signals From Epicardially Placed Accelerometer'. https://github.com/TheInterventionCentre/aortic-valve-event-detection [Accessed 22 Dec 2022].