Virginia Via Research Day Book 2026

Undergraduate Student Research Biomedical

08 OPTIMIZING DATA ANALYSIS OF MOUSE ELECTROCARDIOGRAM (ECG)

Austin Mason, Junting Zhou, and Dr. Jia-Qiang He Corresponding author: awmason200@vt.edu

Department of Biomedical Sciences and Pathobiology, College of Veterinary Medicine, Virginia Tech, Blacksburg, Virginia

Obtaining high-quality, reproducible lectrocardiography (ECG) data from mice is more challenging than from humans, mainly because of the significantly smaller size of the mouse heart. Due to the same reason, the amplitudes of mouse ECG are also much smaller than those of human ECG, which makes it difficult to analyze, especially when the signal is contaminated by noise. Although automatic and AI-based analysis have been used in human ECG, such an approach is not well established in analyzing mouse ECG, especially when the ratio of signal to noise is low. In this case, manual measurements of noisy ECG are needed to avoid potential identification of wrong peaks or duration. In addition to noise, manual analysis of ECG is limited by the low efficiency of human analysts. Oftentimes, researchers unintentionally introduce bias during their manual interpretation of the data, which can easily lead to unreliable, unreproducible results and conclusions. To solve this problem, previous studies have

superimposed ECG traces to obtain an averaged human ECG trace that is used during analysis. However, this method has rarely been used on mouse ECG data. To this end, we designed a standardized method to create more accurate ECG data by superimposing mouse ECG traces. If the standard deviation of the superimposed ECG data is significantly less than the standard deviation of the non-superimposed ECG data, then we can assume that our method produces more consistent, accurate results. Briefly, the original ECG data recorded from both male and female mice (C57BL/6) were exported from BIOPAC, an acquisition software, to a nx2 matrix using Python. Data corresponding to the peaks of the R waves were parsed from the matrix and saved in a new nx2 matrix. For consistency in the data, the time for every RR interval was calculated automatically, and only the traces with the mode of delta time were kept. During ECG acquisition, the amplitude (mV) was measured every millisecond. Thus,

every millisecond correlated to a set that contained the amplitude data for the desired traces. These sets were then averaged into a single data point, and the averaged points were saved in a new nx2 matrix. This matrix was graphed using MATLAB, which provided a superimposed ECG graph. For each mouse, data analysis was manually performed on the superimposed and non-superimposed ECG data. From each ECG graph, the RR interval, P wave amplitude, PQ interval, QRS interval, QT interval, R wave amplitude, J wave amplitude, ST interval, and heart rate were measured. We expect that there will be a statistical difference between the superimposed and non-superimposed standard deviations, specifically in terms of P, R, and J wave amplitudes.

Table of Contents

129 2026 Research Recognition Day