KICA: A MATLAB package for high-throughput analysis of calcium and voltage imaging data from cardiac cells

Introduction

The beat-to-beat cycling of cytosolic calcium and membrane voltage are key processes that are critical for normal func- tion of the heart.^1–3 Measuring the dynamics of these processes is therefore a key experimental approach in understanding cardiac physiology, disease modelling and drug screening. To achieve this, cardiac cells or tissue are commonly labelled with calcium or voltage sensitive fluorophores,^4,5 or genetically encoded sensors,^6–9 and fluorescent amplitude changes, representing changes in cytosolic calcium concentration or membrane voltage, are measured over time. Recently, there has been significant growth in the use of high-throughput platforms, such as high content screening systems, or automated microscopes for this application. For example, one such advanced technique is kinetic imaging cytometry (KIC) that leverages microscopy-based imaging platforms to continuously track and quantify the fluorescent readout from populations of many individual cells over time.^10,11 When applied in plate-based assays, this approach can result in acquisition of data from tens of thousands of individual cells. As a result of emerging approaches such as this, the requirement for automated, hands-off quality control, processing and analysis of these datasets has increased as the size and complexity of the datasets has grown.

To address this, we developed KICA (Kinetic Imaging Cytometry Analyser), a MATLAB-based¹² package designed for efficient high-throughput processing and analysis of fluorescent time-series data. KICA automates key processes such as signal denoising, drift correction, signal-to-noise ratio (SNR) calculation, pulse detection and analysis. In addition to these features, KICA also offers scalability for large datasets by utilising parallel computing to significantly reduce computation time, so streamlining the pipeline for researchers and improving the accuracy and efficiency data analysis. Although KICA is platform agnostic (requiring only CSV describing fluorescent amplitude over time) it also specifically handles data inputs in export format from Vala Sciences CyteSeer software.¹³ KICA has been used in several research studies with applications to both calcium and voltage signals, in acutely isolated as well as induced pluripotent stem cell (IPSC) derived cardiomyocytes, to study both pharmacological interventions and to investigate cardiac pathologies.^14–18 These studies demonstrate the utility of KICA in extracting meaningful biological insights from complex datasets.

In this paper, we describe the development and implementation of KICA, detailing its architecture and features. We discuss the software’s quality control mechanisms and validate its performance using real-world data. Finally, we demonstrate KICA’s scalability and reuse potential for various cytometry applications supporting its use as a tool for laboratories or facilities undertaking high-throughput experimentation in cardiac physiology and drug screening.

Methods

KICA is implemented in MATLAB using an object-oriented programming approach, which provides flexibility and mod- ularity. The software is composed of several core components, each responsible for a specific aspect of the data analysis pipeline, from data ingestion to signal processing and result reporting. The overall workflow of KICA is structured into five main stages:

Each of the above stages is handled by a dedicated module, which makes up the core components of KICA.

Implementation

• • Data import module: Imports time-series data from CSV files generated by imaging software (e.g., CyteSeer). The module also checks for the correct format and extracts key metadata such as sampling rate. The csvread or readtable function in MATLAB is used to load data into the workspace. The sampling rate is inferred by analysing the timestamps or metadata in the CSV file header.

• • Signal processing module: Responsible for denoising, baseline drift correction, and SNR calculation. Figure 1 gives an overview of the signal processing performed by KICA. The initial raw signal, including a polynomial fit used for baseline correction, is depicted in Figure 1(A). Following denoising and baseline correction, the processed signal with detected individual beats is shown in Figure 1(B). Figure 1(C) provides an expanded view of two beats from the processed signal, illustrating the detection of fiducial points corresponding to depolarisation and repolarisation.

  • − Denoising: Utilises a wavelet-based denoising technique to reduce noise in the time-series data. This method decomposes the signal into wavelets, removes noise components, and reconstructs the signal with minimal distortion. The MATLAB wdenoise function is applied, which automatically selects optimal thresholds based on the signal-to-noise ratio.

  • − Baseline Drift Correction: Removes low-frequency components in the signal, which cause drift over time. This is achieved by fitting a polynomial to the baseline and subtracting it from the raw signal. A polynomial fitting algorithm (polyfit) is used to model the baseline drift, which is then subtracted from the signal.

  • − SNR Calculation: The signal-to-noise ratio is calculated to assess the quality of the signal. This helps in iden- tifying poor-quality data before pulse detection. SNR is calculated as the ratio of the mean signal amplitude to the standard deviation of the noise.

• • Pulse detection module: Detects pulses and performs pulse-related analyses such as peak and upstroke detection, APD calculation, and pulse-stimulus validation.

  • − Peak Detection: Identifies peaks in the signal that correspond to action potentials or calcium transients. The findpeaks function in MATLAB detects local maxima based on specified thresholds for peak prominence and width.

  • − Upstroke Detection: Determines the rate of the signal’s upstroke, which corresponds to the rapid depolari- sation phase in action potentials. The upstroke is detected by calculating the slope of the signal between consecutive data points, and the maximum slope is identified as the upstroke.

  • − Action Potential/Calcium Transient Duration (APD/CTD): Calculates the duration between the upstroke and the return of the signal to baseline levels. APD/CTD is calculated as the time difference between the upstroke time and user specified percentages (e.g., 50, 75, 90%) of the return to baseline.

  • − Rise and Fall times: Calculates the time from 10% to 90% of peak in the rising phase of the calcium transient (Rise time) and from 90% to 10% of peak in the falling phase of the calcium transient (Fall time).

• • Quality control module: Ensures data integrity by validating signals based on SNR and pulse detection.

  • − Ensures the quality of the data by validating each signal based on the SNR and the presence of valid pulses. Signals that do not meet the criteria are flagged and excluded from further analysis. Threshold values for SNR and pulse detection are predefined by the user. If the SNR is below the threshold or no valid pulses are detected, the signal is flagged as failing quality control.

• • Reporting module: Generates Excel reports with pulse characteristics and saves images of the analysed signals for each cell.

  • − Generates detailed Excel reports that summarise the results of the analysis. The report includes key metrics for each cell, including pulse duration, peak amplitude, and APD, as well as visualisations of the analysed signals. The xlswrite function in MATLAB is used to create structured reports with embedded figures for each analysed cell. The visualisation is produced using MATLAB’s plotting functions and saved as images in the report. In addition to the Excel report, the software also allows for saving MATLAB figures (.fig) for each individual cell for more in depth interrogation and preparation of publication quality figures.

• • Parallelisation module: Utilises MATLAB’s Parallel Computing Toolbox to distribute tasks across multiple CPU cores for efficient high-throughput analysis.

  • − KICA takes advantage of MATLAB’s Parallel Computing Toolbox to distribute the computational load across multiple CPU cores. This is especially important for high-throughput datasets, where the number of cells analysed can exceed thousands. The parfor function is used to parallelise the processing loop, allowing the analysis of each cell to run concurrently across available CPU cores. The time-series data from each cell is processed independently, making it an ideal candidate for parallelisation.

Figure 1. Overview of signal processing using KICA.

A) Raw signal showing fluorescent reports of ten paced cardiac action potential waveforms. A polynomial fit to the baseline (used for baseline correction) is shown. B) Baseline corrected and denoised signal. Individual beats/pulses are detected via threshold crossing. C) Expanded view of first two action potentials from (B), showing fiducial points related to action potential depolarisation and repolarisation.

Use cases

To help users quickly understand if the software is functioning correctly and how to apply it to their data, KICA includes example input datasets and a detailed user guide. The example datasets can be processed using the default parameters, and the expected output is provided in the form of Excel reports and graphical visualisations. The user guide also provides instructions on how to set user parameters such as stimulation start time, stimulation period, number of stimuli, fluorescent signal type (voltage or calcium), pulse detection threshold percentage, pulse action potential durations, pulse start on stimulus detection delta duration, pulse start point type (upstroke start, upstroke end or activation point) and quality control SNR threshold.

By providing these sample inputs and outputs, KICA ensures that users can quickly validate the software’s functionality upon installation and understand how to interpret the results. Detailed instructions for running these tests and explaining parameters are included in the software’s user guide, along with URL and contact details to report issues.

Discussion

While KICA was developed for use with fluorescent time-series data related to calcium cycling and membrane voltage, its flexible design makes it applicable to analysis of other periodic physiological datasets such as contraction data, or direct electrophysiological recordings of action potentials. Researchers working in fields such as immunology, neuroscience, and pharmacology can adapt KICA to their needs by modifying its parameters and incorporating additional signal processing steps. The object-oriented architecture facilitates these extensions.

KICA’s ability to handle large datasets and perform quality control checks makes it particularly useful for researchers conducting high-throughput screenings, where manual data analysis is impractical.

Ethics and consent

Ethical approval and consent were not required.

Software availability