Automated detection and analysis of depolarization events in human cardiomyocytes using MaDEC

Abstract

Optical imaging-based methods for assessing the membrane electrophysiology of in vitro human cardiac cells allow for non-invasive temporal assessment of the effect of drugs and other stimuli. Automated methods for detecting and analyzing the depolarization events (DEs) in image-based data allow quantitative assessment of these different treatments. In this study, we use 2-photon microscopy of fluorescent voltage-sensitive dyes (VSDs) to capture the membrane voltage of actively beating human induced pluripotent stem cell-derived cardiomyocytes (hiPS-CMs). We built a custom and freely available Matlab software, called MaDEC, to detect, quantify, and compare DEs of hiPS-CMs treated with the β-adrenergic drugs, propranolol and isoproterenol. The efficacy of our software is quantified by comparing detection results against manual DE detection by expert analysts, and comparing DE analysis results to known drug-induced electrophysiological effects. The software accurately detected DEs with true positive rates of 98-100% and false positive rates of 1-2%, at signal-to-noise ratios (SNRs) of 5 and above. The MaDEC software was also able to distinguish control DEs from drug-treated DEs both immediately as well as 10min after drug administration.

Keywords: Classification; Matched filter; Signal detection; Voltage-sensitive dye imaging; hiPS derived cardiomyocytes.

Figure 1

Examples of training templates s from control (left), isoproterenol (middle), and propranolol (right) data. Each template is 0.48 s long.

Figure 2

Detector Performance. The figure consists of three pre-processed intensity traces from the control (Top), isoproterenol (Middle), and propranolol (Bottom) drug treatments. The circles above each data trace represent the DEs identified by analysts 1–3 (bottom to top respectively). The triangles above each data trace represent MaD detected DEs, using the optimal thresholds identified in the corresponding ROC curves to the right. The optimal threshold is identified as the one resulting in the detector performance closest to TP = 100% and FP = 0%. The error bars represent 95% confidence intervals. The optimal performance for each drug treatment case is presented under each ROC curve. The MaD detector performed with a TP rate of 98–100% and a FP rate of 1–2% for all 3 drug treatments.

Figure 3

Detector Performance At Low SNR. The figure consists of two pre-processed intensity traces from the low SNR control (Top), and the low SNR isoproterenol (Bottom) drug treatments. The circles above each data trace represent the DEs identified by analysts 1–3 (bottom to top respectively). The triangles above each data trace represent MaD detected DEs, using the optimal thresholds identified in the corresponding ROC curves to the right. The error bars represent 95% confidence intervals. The optimal performance for each drug treatment case is presented under each ROC curve. As the SNR decreased, the performance of the MaD detector also decreased. Performance was still adequate for SNR ≃ 3, but not for SNR ≤ 1.

Figure 4

DE analysis with respect to drug treatment and time after drug administration. 6 panels are shown. Each panel shows the average DEs of the two specified drug treatments, with standard deviations shaded around the average. The title of each panel identifies the two treatments being compared. The text below the title indicates whether the DE populations were different in width, upslope, and downslope, and provides p-values. The top 3 panels, going from left to right, compare Iso-1 0min, (green) to Control-1 (red), Iso-1 10min (light teal) to Control-1, and Iso-1 0min to Iso-1 10min. Both Iso-1 0min and Iso-1 10min were different from Control-1, and were different from each other, in all three parameters (width, upslope, downslope). The bottom 3 panels, going from left to right, compare Pro-1 0min (purple) to Control-1 (red), the Pro-1 10min (blue) to Control-1, and Pro-1 0min to Pro-1 10min. Both Pro-1 0min and Pro-1 10min were different from Control-1 in width, upslope, and downslope. They were also different from each other in width, and upslope.

Figure 5

DE analysis with respect to drug treatment for the lower SNR data. Iso-2 15min (orange) and Control-2 (red) were different in width, but not in upslope or downslope.