Deep-Learning-Based Accelerated and Noise-Suppressed Estimation (DANSE) of quantitative Gradient Recalled Echo (qGRE) MRI metrics associated with Human Brain Neuronal Structure and Hemodynamic Properties

Kahali, S.; Kothapalli, S. V. V. N.; Xu, X.; Kamilov, U. S.; Yablonskiy, D. A.

doi:10.1101/2021.09.10.459810

Home/Archive/bioRxiv

bioRxiv · 10.1101/2021.09.10.459810

Deep-Learning-Based Accelerated and Noise-Suppressed Estimation (DANSE) of quantitative Gradient Recalled Echo (qGRE) MRI metrics associated with Human Brain Neuronal Structure and Hemodynamic Properties

Kahali, S.·Kothapalli, S. V. V. N.·Xu, X.·Kamilov, U. S.·Yablonskiy, D. A.

View original source ↗Export metadata ↓

Abstract

PurposeTo introduce a Deep-Learning-Based Accelerated and Noise-Suppressed Estimation (DANSE) method for reconstructing quantitative maps of biological tissue cellular-specific, R2t* and hemodynamic-specific, R2' from Gradient-Recalled-Echo (GRE) MRI data with multiple gradient-recalled echoes. MethodsDANSE method adapts supervised learning paradigm to train a convolutional neural network for robust estimation of R2t* and R2' maps free from the adverse effects of macroscopic (B0) magnetic field inhomogeneities directly from the GRE magnitude images without utilizing phase images. The corresponding ground-truth maps were generated by means of a voxel-by-voxel fitting of a previously-developed biophysical quantitative GRE (qGRE) model accounting for tissue, hemodynamic and B0-inhomogeneities contributions to GRE signal with multiple gradient echoes using nonlinear least square (NLLS) algorithm. ResultsWe show that the DANSE model efficiently estimates the aforementioned brain maps and preserves all features of NLLS approach with significant improvements including noise-suppression and computation speed (from many hours to seconds). The noise-suppression feature of DANSE is especially prominent for data with SNR characteristic for typical GRE data (SNR~50), where DANSE-generated R2t* and R2' maps had three times smaller errors than that of NLLS method. ConclusionsDANSE method enables fast reconstruction of magnetic-field-inhomogeneity-free and noise-suppressed quantitative qGRE brain maps. DANSE method does not require any information about field inhomogeneities during application. It exploits spatial patterns in the qGRE MRI data and previously-gained knowledge from the biophysical model, thus producing clean brain maps even in the environments with high noise levels. These features along with fast computational speed can lead to broad qGRE clinical and research applications.

Source connections

Published article ↗

Explore related subjects

neuroscience Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Kahali, S., Kothapalli, S. V. V. N., Xu, X., Kamilov, U. S., Yablonskiy, D. A.. 2021-09-11. Deep-Learning-Based Accelerated and Noise-Suppressed Estimation (DANSE) of quantitative Gradient Recalled Echo (qGRE) MRI metrics associated with Human Brain Neuronal Structure and Hemodynamic Properties. https://doi.org/10.1101/2021.09.10.459810

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

A Mechanistic Account of Dedifferentiation in the Aging Visual System

To uncover the mechanisms of age-related dedifferentiation, we simulated neural degradation in a computational model of the ventral visual stream. Severing connections or adding noise, whether in inferotemporal cortex (IT) or upstream, reduced IT differentiation, indicating that dedifferentiation need not originate locally. Degrading peripheral V1, which preferentially drove place-selective IT neurons, selectively reduced place differentiation, a candidate mechanism for the stronger place dedifferentiation reported in aging.

neuroscience↗

bioRxiv2026

Reusing pipettes in Patch-seq: a foundation for automated multimodal profiling

Patch-seq is a technique for single-cell multimodal characterization. It combines whole-cell electrophysiology, single-cell transcriptomics, and morphological reconstruction in single neurons. Patch-seq morphologies are a critical bridge among cell types, genes, physiology, and large-scale connectivity platforms such as EM connectomes and single-cell-resolution projectomes. However, patch-seq experimental complexity and operator dependence limit throughput. Automation could enable scalable Patch-seq but requires procedural adaptations, including detergent-based pipette cleaning, robotic nucleus extraction and collection. The impact of these automations on Patch-seq quality has not been systematically evaluated. Here, we show that each of these modifications preserves or improves key Patch-seq outcomes. Automated nucleus extraction performs comparably to, or better than, expert manual operation, preserving key RNA and morphological metrics. Detergent-cleaned, reused pipettes supported a viable sealing success rate across multiple cleaning cycles and preserved electrophysiological quality and transcriptomic yield without introducing excess contamination. Sonication effectively allows collection of nuclear material without damaging RNA and supports subsequent pipette cleaning and reuse. Enhanced automated sealing and break-in procedures increased throughput regardless of brain region, cell type or species. Together, these results validate the core procedural adaptations required for automated Patch-seq and establish a foundation for high-throughput, standardized multimodal single-cell characterization.

neuroscience↗

bioRxiv2026

Development of direction selectivity in the Xenopus tadpole optic tectum and midbrain tegmentum

Detecting motion direction is a fundamental function performed by direction selective (DS) neurons. DS neurons in the optic tectum (superior colliculus in mammals) enable the tracking of moving objects. DS neurons of the accessory optic system support reflexive eye / head movements that stabilize gaze during self-motion. By developmental stage 48 (10 days postfertilization), Xenopus tadpoles display visually guided behaviors that require both types of DS neurons. Our previous work has shown that these two regions are not only functionally distinct but also follow different developmental programs. Here we characterized and compared how direction selectivity develops in both visual processing centers. Using in vivo electrophysiological recordings during moving bar visual stimulation, we measured DS synaptic inputs and spiking outputs in both regions across three key developmental stages of the retinotectal projection. Tegmental neurons were direction selective ab initio, and both input and output tuning remained stable across development. Tectal neurons were observed to receive DS inputs early, and their input selectivity also remained stable across development. However, their spiking output grew substantially more DS at the later developmental stage. This was due at least in part to a naturally occurring decrease in intrinsic excitability which preferentially dampened overall spike rates, but especially in response to the non-preferred direction of motion, thereby increasing direction selectivity. These results reveal distinct developmental trajectories in parallel visual pathways. Tegmental direction selectivity is established early and stays stable, whereas tectal output tuning sharpens during development through a decrease in intrinsic excitability.

neuroscience↗