Abstract
Dynamic Ca(2+) signaling is crucial for cell survival and death, and Ca(2+) imaging approaches are commonly used to study and measure cellular Ca(2+) patterns within cells. However, the presence of image noise from instrumentation and experimentation protocols can impede the accurate extraction of Ca(2+) signals. Removing noise from Ca(2+) Spatio-Temporal Maps (STMaps) is essential for precisely analyzing Ca(2+) datasets. Current methods for denoising STMaps can be time-consuming and subjective and rely mainly on image processing protocols. To address this, we developed CalDenoise, an automated software that employs robust image processing and deep learning models to remove noise and enhance Ca(2+) signals in STMaps effectively. CalDenoise integrates four pipelines capable of efficiently removing salt-and-pepper, impulsive, and periodic noise and detecting and removing background noise. Comprising both an image-processing-based pipeline and three generative-adversarial-network-based (GAN) deep learning models, CalDenoise proficiently removes complex noise patterns. The software features adjustable parameters to enhance accuracy and is integrated into a user-friendly graphical interface for easy access and streamlined usage. CalDenoise can serve as a robust platform for denoising complex dynamic fluorescence signal images across diverse cell types, including Ca(2+), voltage, ions, and pH signals.