What problem does this paper attempt to address?

The problem that this paper attempts to solve is to remove noise in transient fluorescence signals in functional imaging, especially for glutamate imaging. Traditional denoising methods such as Gaussian filtering or median filtering will significantly reduce the temporal and spatial resolutions while smoothing the entire recording, and are not suitable for removing noise in glutamate recordings because the kinetics of these sensors are very fast and the signals are located in the synaptic region, presenting unique challenges. Therefore, a specialized tool is required to address the challenges posed by glutamate recordings, namely faster kinetics and synaptic localization. ### Background of the Paper - **Research Significance**: Accurate measurement of neuronal activity is crucial for understanding synaptic plasticity and learning mechanisms. However, noise in the imaging system can severely distort the data. - **Limitations of Existing Methods**: Traditional denoising methods (such as Gaussian filtering or median filtering), although effective to a certain extent, will smooth the entire image indiscriminately, leading to a significant decrease in temporal and spatial resolutions. Moreover, manual selection of filtering parameters is labor - intensive and subjective, often resulting in sub - optimal denoising effects. - **Deficiencies of Existing Frameworks**: Existing denoising frameworks are not suitable for glutamate recordings because glutamate sensors have very fast kinetics and require higher spatio - temporal resolutions. ### Goals of the Paper - **Develop a New Denoising Tool**: A new model based on the U - Net architecture, Neuroimage Denoiser, has been developed to effectively remove noise from transient local fluorescence signals in microscope recordings. - **Be Independent of Recording Frequency and Sensor Kinetics**: The design of Neuroimage Denoiser makes the denoising process independent of the recording frequency and the kinetics of the sensors used, thereby increasing the flexibility and universality of the model. - **Maintain Response Amplitude**: While removing noise, maintain the integrity of the response amplitude to ensure the accuracy and reliability of the data. ### Solutions - **Model Architecture**: Adopt the U - Net architecture, which includes encoder and decoder parts, and retains untransformed data through skip connections. The model uses 3x3 convolutional layers, ReLU activation functions, and batch normalization. - **Training Data Preparation**: Automatically screen out frames and regions showing active synaptic signals without manual intervention. Efficiently filter the background by calculating the activity map of each frame. - **Denoising Performance Evaluation**: Evaluate the denoising performance of the model by calculating the mean absolute percentage error (MAPE) and the standard deviation of the non - responsive part. The selection of the best model is based on the minimum combined score. ### Results - **Denoising Effect**: Neuroimage Denoiser can effectively remove noise while maintaining a high degree of consistency in the response amplitude (correlation coefficient R = 0.9966, p < 0.0001). - **Processing Speed**: Tested on images of different sizes, Neuroimage Denoiser shows a fast inference speed, and it only takes about 152 seconds to process a 1,000 - frame recording of 800x800 pixels. Through these improvements, Neuroimage Denoiser significantly improves the quality of functional microscope recordings, providing support for more accurate and reliable analysis of neural activity.

Neuroimage Denoiser for removing noise from transient fluorescent signals in functional imaging.

Frequency-Relevant Residual Learning for Multi-Modal Image Denoising.

Fast, efficient, and accurate neuro-imaging denoising via supervised deep learning

Convolutional Neural Network Denoising in Fluorescence Lifetime Imaging Microscopy (FLIM)

Denoising Across Data Acquisition Modalities for Mesoscopic Scale Optical Neuroimaging

DENOISING: Dynamic enhancement and noise overcoming in multimodal neural observations via high-density CMOS-based biosensors

Framework for denoising Monte Carlo photon transport simulations using deep learning

Real-time Image Denoising of Mixed Poisson-Gaussian Noise in Fluorescence Microscopy Images using ImageJ

A multi-artifact EEG denoising by frequency-based deep learning

Weak signal extraction enabled by deep neural network denoising of diffraction data

FM2S: Self-Supervised Fluorescence Microscopy Denoising With Single Noisy Image

A robust deep neural network for denoising task-based fMRI data: An application to working memory and episodic memory

Deep Learning-Based Denoising of CEST MR Data: A Feasibility Study on Applying Synthetic Phantoms in Medical Imaging

Shot-Noise Limited Nonlinear Optical Imaging Excited With GHz Femtosecond Pulses and Denoised by Deep-Learning

Real-time denoising enables high-sensitivity fluorescence time-lapse imaging beyond the shot-noise limit

Neural network denoising of x-ray images from high-energy-density experiments

Developing and Evaluating Deep Neural Network-Based Denoising for Nanoparticle TEM Images with Ultra-Low Signal-to-Noise

Deep Joint Denoising and Detection for Enhanced Intracellular Particle Analysis

Robust self-supervised denoising of voltage imaging data using CellMincer

Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging

Fluorescence Microscopy Images Denoising Via Deep Convolutional Sparse Coding