Noisy Imaging Data Research Articles

Diffusion tensor field estimation based on 3D U-Net and diffusion tensor imaging model constraint

To propose a diffusion tensor field estimation network based on 3D U-Net and diffusion tensor imaging (DTI) model constraint (3D DTI-Unet) to accurately estimate DTI quantification parameters from a small number of diffusion-weighted (DW) images with a low signal-to-noise ratio. The input of 3D DTI-Unet was noisy diffusion magnetic resonance imaging (dMRI) data containing one non-DW image and 6 DW images with different diffusion coding directions. The noise-reduced non-DW image and accurate diffusion tensor field were predicted through 3D U-Net. The dMRI data were reconstructed using the DTI model and compared with the true value of dMRI data to optimize the network and ensure the consistency of the dMRI data with the physical model of the diffusion tensor field. We compared 3D DTI-Unet with two DW image denoising algorithms (MP-PCA and GL-HOSVD) to verify the effect of the proposed method. The proposed method was better than MP-PCA and GL-HOSVD in terms of quantitative results and visual evaluation of DW images, diffusion tensor field and DTI quantification parameters. The proposed method can obtain accurate DTI quantification parameters from one non-DW image and 6 DW images to reduce image acquisition time and improve the reliability of quantitative diagnosis.

Identification of electrical rotational activity in noisy cardiac tissue recordings using a deep neural network

Abstract Funding Acknowledgements Type of funding sources: None. Background Deep learning is increasingly used in modern biomedical research and applications due to the substantial availability of large clinical datasets. These approaches are invaluable in tasks involving noisy imaging data, such as tumour segmentation in histological images. In cardiology, a deep learning approach could be helpful in real-time tracking of the sources of arrhythmia, i.e. electrical rotational activity in the heart. However, the existing optical or electrophysiological recordings that could be used for training such a model are recorded under highly variable conditions and are not always annotated, thereby requiring data augmentation. Purpose To use deep neural networks trained on synthetic data to obtain concise (low dimensional) representations of noisy optical mapping recordings of cardiac arrhythmias and rapidly locate spiral wave centres. Methods To overcome the lack of experimental training data, a digital twin of a neonatal rat ventricular cardiomyocyte monolayer was used to create a large synthetic training dataset of noiseless spiral wave recordings. Spiral wave centres were detected and labelled by making use of classical algorithms which are proven to work well on noiseless data. After labelling the centres, noise was added to the spiral wave recordings to simulate realistic experimental measurements. Subsequently, these data were fed into three different deep learning architectures: 1) a variational auto-encoder (VAE) to denoise optical mapping recordings of cardiac arrhythmias in an unsupervised manner, 2) a convolutional neural network (CNN) to detect the spiral centres, and 3) a combination of both to denoise the recording and detect centres simultaneously. Results After training on synthetic datasets, each architecture could accurately predict what it was designed for (noiseless wave fronts, spiral centres including chirality, or both) on both synthetic and experimental data. These spiral centre detection results were compared with 5 classical methods of denoising and spiral centre detection for accuracy and speed. Our method was as accurate as the best performing yet slow classical algorithm, which can only detect centres after observing a full rotation cycle (~300ms). At the same time, it was as fast as the fastest classical method, needing only 30ms after enabling the algorithm to detect spiral wave centres. This allows quasi-real-time tracking of arrhythmic sources. Conclusion This study reveals that modern deep learning strategies in combination with synthetic simulation datasets can be used on experimental measurements to aid in the development of new technologies, here applied to the detection of spiral wave centres in optical mapping recordings of cardiac arrhythmias. Given the combination of speed and accuracy at which these algorithms produce results, further exploration and refinement may improve the identification of targets for catheter ablation, thereby potentially improving the outcome.

Bayesian nonparametric analysis for the detection of spikes in noisy calcium imaging data.

Recent advancements in miniaturized fluorescence microscopy have made it possible to investigate neuronal responses to external stimuli in awake behaving animals through the analysis of intracellular calcium signals. An ongoing challenge is deconvolving the temporal signals to extract the spike trains from the noisy calcium signals' time series. In this article, we propose a nested Bayesian finite mixture specification that allows the estimation of spiking activity and, simultaneously, reconstructing the distributions of the calcium transient spikes' amplitudes under different experimental conditions. The proposed model leverages two nested layers of random discrete mixture priors to borrow information between experiments and discover similarities in the distributional patterns of neuronal responses to different stimuli. Furthermore, the spikes' intensity values are also clustered within and between experimental conditions to determine the existence of common (recurring) response amplitudes. Simulation studies and the analysis of a dataset from the Allen Brain Observatory show the effectiveness of the method in clustering and detecting neuronal activities.

Automated weak signal extraction of hyperspectral Raman imaging data by adaptive low‐rank matrix approximation

AbstractHyperspectral Raman imaging has emerged as a promising spectroscopic tool that can provide spatial and molecular information of the sample in a label‐free and noninvasive manner, which is very suitable to the biological and biomedical research. However, the intrinsically weak Raman scattering effect results in the low signal quality of the measured Raman spectra, which has largely limited the application of Raman imaging. In this paper, we develop an adaptive low‐rank matrix approximation method to automatically extract the signal from the noisy hyperspectral Raman imaging data. After spike removal, the hyperspectral Raman imaging data are decomposed into a linear combination of submatrices by singular value decomposition. Next, the submatrices are classified into positive and negative groups according to the SNR contribution. The negative group, reflecting the instrumental noise, is discarded, and the positive group is used to reconstruct the denoised signal. We prove on the simulated data that this algorithm can significantly decrease the normalized mean noise error from 34.35% down to 1.9%. Such a strong denoising performance enables it to efficiently extract the signal from the noisy hyperspectral Raman imaging data, especially under the low SNR condition. We finally apply this algorithm to the fast speed Raman imaging of HeLa cell and surprisingly find that the slight difference of the spectra can be differentiated after signal extraction. This algorithm offers a promising tool in the Raman imaging and also can be extended to other spectroscopic imaging platforms such as fluorescent and IR spectroscopy.

Generating patient-specific virtual tumor populations with reaction-diffusion models and molecular imaging data.

The use of mathematical tumor growth models coupled to noisy imaging data has been suggested as a possible component in the push towards precision medicine. We discuss the generation of population and patient-specific virtual populations in this context, providing in silico experiments to demonstrate how intra- and inter-patient heterogeneity can be estimated by applying rigorous statistical procedures to noisy molecular imaging data, and how the noise properties of such data can be analyzed to estimate uncertainties in predicted patient outcomes.

Representational structure or task structure? Bias in neural representational similarity analysis and a Bayesian method for reducing bias.

The activity of neural populations in the brains of humans and animals can exhibit vastly different spatial patterns when faced with different tasks or environmental stimuli. The degrees of similarity between these neural activity patterns in response to different events are used to characterize the representational structure of cognitive states in a neural population. The dominant methods of investigating this similarity structure first estimate neural activity patterns from noisy neural imaging data using linear regression, and then examine the similarity between the estimated patterns. Here, we show that this approach introduces spurious bias structure in the resulting similarity matrix, in particular when applied to fMRI data. This problem is especially severe when the signal-to-noise ratio is low and in cases where experimental conditions cannot be fully randomized in a task. We propose Bayesian Representational Similarity Analysis (BRSA), an alternative method for computing representational similarity, in which we treat the covariance structure of neural activity patterns as a hyper-parameter in a generative model of the neural data. By marginalizing over the unknown activity patterns, we can directly estimate this covariance structure from imaging data. This method offers significant reductions in bias and allows estimation of neural representational similarity with previously unattained levels of precision at low signal-to-noise ratio, without losing the possibility of deriving an interpretable distance measure from the estimated similarity. The method is closely related to Pattern Component Model (PCM), but instead of modeling the estimated neural patterns as in PCM, BRSA models the imaging data directly and is suited for analyzing data in which the order of task conditions is not fully counterbalanced. The probabilistic framework allows for jointly analyzing data from a group of participants. The method can also simultaneously estimate a signal-to-noise ratio map that shows where the learned representational structure is supported more strongly. Both this map and the learned covariance matrix can be used as a structured prior for maximum a posteriori estimation of neural activity patterns, which can be further used for fMRI decoding. Our method therefore paves the way towards a more unified and principled analysis of neural representations underlying fMRI signals. We make our tool freely available in Brain Imaging Analysis Kit (BrainIAK).

TraceSpecks: A Software for Automated Idealization of Noisy Patch-Clamp and Imaging Data

Experimental records of single molecules or ion channels from fluorescence microscopy and patch-clamp electrophysiology often include high-frequency noise and baseline fluctuations that are not generated by the system under investigation and have to be removed. Moreover, multiple channels or conductance levels can be present at a time in the data that need to be quantified to accurately understand the behavior of the system. Manual procedures for removing these fluctuations and extracting conducting states or multiple channels are laborious, prone to subjective bias, and likely to hinder the processing of often very large data sets. We introduce a maximal likelihood formalism for separating signal from a noisy and drifting background such as fluorescence traces from imaging of elementary Ca2+ release events called puffs arising from clusters of channels, and patch-clamp recordings of ion channels. Parameters such as the number of open channels or conducting states, noise level, and background signal can all be optimized using the expectation-maximization algorithm. We implement our algorithm following the Baum-Welch approach to expectation-maximization in the portable Java language with a user-friendly graphical interface and test the algorithm on both synthetic and experimental data from the patch-clamp electrophysiology of Ca2+ channels and fluorescence microscopy of a cluster of Ca2+ channels and Ca2+ channels with multiple conductance levels. The resulting software is accurate, fast, and provides detailed information usually not available through manual analysis. Options for visual inspection of the raw and processed data with key parameters are provided, in addition to a range of statistics such as the mean open probabilities, mean open times, mean close times, dwell-time distributions for different number of channels open or conductance levels, amplitude distribution of all opening events, and number of transitions between different number of open channels or conducting levels in asci format with a single click.

MIN1PIPE: A Miniscope 1-Photon-Based Calcium Imaging Signal Extraction Pipeline

Invivo calcium imaging using a 1-photon-based miniscope and a microendoscopic lens enables studies of neural activities in freely behaving animals. However, the high and fluctuating background, the inevitable movements and distortions of imaging field, and the extensive spatial overlaps of fluorescent signals emitted from imaged neurons inherent in this 1-photon imaging method present major challenges for extracting neuronal signals reliably and automatically from the raw imaging data. Here, we develop a unifying algorithm called the miniscope 1-photon imaging pipeline (MIN1PIPE), which contains several stand-alone modules and can handle a wide range of imaging conditions and qualities with minimal parameter tuning and automatically and accurately isolate spatially localized neural signals. We have quantitatively compared MIN1PIPE with other existing partial methods using both synthetic and real datasets obtained from different animal models and show that MIN1PIPE has superior efficiency and precision in analyzing noisy miniscope calcium imaging data.

Inference of neuronal network spike dynamics and topology from calcium imaging data

Two-photon calcium imaging enables functional analysis of neuronal circuits by inferring action potential (AP) occurrence (“spike trains”) from cellular fluorescence signals. It remains unclear how experimental parameters such as signal-to-noise ratio (SNR) and acquisition rate affect spike inference and whether additional information about network structure can be extracted. Here we present a simulation framework for quantitatively assessing how well spike dynamics and network topology can be inferred from noisy calcium imaging data. For simulated AP-evoked calcium transients in neocortical pyramidal cells, we analyzed the quality of spike inference as a function of SNR and data acquisition rate using a recently introduced peeling algorithm. Given experimentally attainable values of SNR and acquisition rate, neural spike trains could be reconstructed accurately and with up to millisecond precision. We then applied statistical neuronal network models to explore how remaining uncertainties in spike inference affect estimates of network connectivity and topological features of network organization. We define the experimental conditions suitable for inferring whether the network has a scale-free structure and determine how well hub neurons can be identified. Our findings provide a benchmark for future calcium imaging studies that aim to reliably infer neuronal network properties.

Quantitative Analysis of Clinical Dynamic Contrast-enhanced MR Imaging for Evaluating Treatment Response in Human Breast Cancer

To develop a method that combines a fixed-T1, fuzzy c-means (FCM) technique with a reference region (RR) model (T1-FCM method) to estimate pharmacokinetic parameters without measuring the arterial input function or baseline T1, or T1(0), and to demonstrate its feasibility in the assessment of treatment response to neoadjuvant chemotherapy (NAC) in patients with breast cancer by using data from dynamic contrast material-enhanced magnetic resonance (MR) imaging. This study was approved by the human investigation committees of the two participating institutions. All patients gave written informed consent. A conventional dual-flip-angle gradient-echo method was used to evaluate the effects of noise and the T1 in the tissue itself on the accuracy of T1 estimation. Both conventional RR and fixed-T1 methods were used to evaluate the effects of noise and preselected T1(0) on the estimation of pharmacokinetic parameters by means of a simulation study. Thirty-three women (age range, 32-66 years; mean age, 45 years) with pathologically proved breast tumors were examined to evaluate the feasibility of using the T1-FCM method as a means of assessing treatment response to NAC. A nonparametric Mann-Whitney U test was used to assess the difference in each of the MR imaging parameters between patients with a major histologic response to treatment and those with a nonmajor histologic response. With use of the dual-flip-angle method, the accuracy and distribution of T1 estimation are dependent on the T1 in the tissue itself. The T1-FCM method is more accurate than other methods and is relatively insensitive to the effects of noise and incorrect T1(0) selection. Preliminary clinical data revealed a significant difference (P < .01) in the change of the volume transfer constant after two cycles of NAC between the major and nonmajor histologic response groups. Results of the simulation study demonstrate that the T1-FCM method appears to be relatively insensitive to noisy dynamic contrast-enhanced MR imaging data. This method could prove useful in the evaluation of breast cancer therapy.

New statistical methods enhance imaging of cameleon fluorescence resonance energy transfer in cultured zebrafish spinal neurons

Cameleons are genetically encoded fluorescence resonance energy transfer (FRET)-based Ca(2+) indicators. Attempts to use cameleons to detect neural activity in vertebrate systems have been largely frustrated by the small FRET signal, in contradistinction to the higher signals seen in Drosophila and Caenorhabditis elegans. We have developed a statistical optimization method capable of detecting small ratiometric signals in noisy imaging data, called statistical optimization for the analysis of ratiometric signals. Using this method, we can detect and estimate anticorrelated ratiometric signals with subcellular resolution in cultured, dissociated zebrafish spinal neurons expressing cameleon or loaded with fluo-4 and fura-red. This method may make it possible to use yellow cameleons for measuring neural activity at high resolution in transgenic animals.

Robust anisotropic diffusion to produce enhanced statistical parametric map from noisy fMRI

This paper presents a new, simple, and elegant technique to obtain enhanced statistical parametric maps (SPMs) from noisy functional magnetic resonance imaging (fMRI) data. This technique is based on the robust anisotropic diffusion (RAD), a technique normally used as an edge-preserving filter. A direct application of the RAD to the fMRI data does not work, because in this case RAD would perform an edge-preserving filtering of the fMRI structural information, instead of enhancing its functional information. The RAD can be applied directly to SPM but, in this case, only a small improvement of the SPM quality can be achieved, because the originating fMRI is not taken into account. To overcome these difficulties, we propose to estimate the SPM from the noisy fMRI, compute the diffusion coefficients in the SPM space, and then perform the diffusion in the structural information-removed fMRI data using the coefficients previously computed. These steps are iterated until convergence. We have tested the new technique in both simulated and real fMRI images, yielding surprisingly sharp and noiseless SPMs with increased statistical significance. We also describe how to automatically estimate an appropriate scale parameter.