Search bioRxiv⌕ Search

bioRxiv · 10.64898/2026.02.27.708398

VascFlexMap: Microvascular Ultrasound Imaging at Low Frame Rates Using Sparse Data and a Transformer-Decoder Network

Abstract

ObjectiveSuper-resolution ultrasound (SR-US) reveals microvascular structures with exquisite resolution, but clinical translation remains limited by the need for ultrafast frame rates, massive data volumes, and long reconstruction times. This work proposes a deep learning framework that reconstructs microvascular maps from low-frame-rate enhanced ultrasound sequences, bypassing explicit microbubble localization and tracking. MethodsA transformer-decoder network with learned linear projections was designed to model spatiotemporal dependencies across sparse contrast-enhanced ultrasound sequences and reconstruct vessel probability maps, refined via a post-processing enhancement stage. Single-head self-attention captures temporal correlations under challenging conditions including overlapping microbubbles and low signal-to-noise ratios. Binary cross-entropy loss guided training to preserve vascular topology across synthetic and in vivo datasets. In vivo rat brain bolus data from the PALA challenge was used to evaluate this approach under up to 500 - fold data reduction (341 frames at 2 FPS vs. 170400 frames at 1000 FPS in standard ULM). ResultsDespite aggressive undersampling, the proposed pipeline recovered coherent microvascular architecture where conventional ULM pipelines applied to the same sparse data failed to produce continuous vascular networks. Major branches and higher-order microvessels remained visible with apparent vessel widths broadened by approximately three-fold relative to reference SR-US. End-to-end reconstruction completed in 28-133 seconds on an NVIDIA H100 GPU depending on the number of frames employed. ConclusionThe reported approach preserved vascular topology with fast reconstruction and low data overhead, albeit at lower resolution. The substantial reduction in frames and computation time highlights the translational potential of this SR-US-inspired microvascular imaging approach.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Dhawan, R., Agarwal, M., Jain, S., Shekhar, H.. 2026-03-02. VascFlexMap: Microvascular Ultrasound Imaging at Low Frame Rates Using Sparse Data and a Transformer-Decoder Network. https://doi.org/10.64898/2026.02.27.708398

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

KEEP EXPLORING

Related preprints

Circular-Arc Modal Migration for Ultrasound Imaging--Part II: Simulation and Experimental Evaluation in Compact Semicircular and Convex Arrays

Circular-arc ultrasound imaging requires reconstruction that accounts for curvature and offset transmit sources across a wide field. This study evaluates how Circular-Arc Modal Migration (CAMM) changes point-response width, image-energy distribution, and low-echo target separation relative to delay-and-sum (DAS). CAMM combines angular modal analysis, local spectral remapping, and radial synthesis, with the offset phase evaluated through phase screens or approximated by scalar pullback. Guided by Part I's error analysis and point-target comparisons, paired reconstructions use scalar pullback for compact semicircular cyst simulations and phase screens for acquisition-matched C5-2v simulations, six phantom records, and two carotid views from one participant. Each DAS/CAMM pair uses identical RF. The principal observation is that lower energy fractions outside a fixed neighborhood can coexist with broader tangential responses. In C5-2v point simulations, the full-field fraction is lower at 23 of 24 locations, whereas tangential width is greater at all 24. Semicircular cyst contrast improves in most configurations, with reversed ranking at the smallest target. All eight C5-2v cyst configurations at 10--15-mm radial depth favor CAMM in the three contrast measures; deeper comparisons depend on location and metric. Measured reflectors show the same width--energy-fraction pattern, and the evaluated phantom cyst and carotid lumen regions show improved contrast. At 512 x 513 pixels and 24 transmissions, median host latency is 29.79 ms for DAS and 51.87 ms for CAMM on an RTX 4090. The tested CAMM implementation therefore provides regional gains in low-echo target separation alongside broader tangential responses and higher computational cost.

bioengineering↗

QuickXL: An Open-source Ultraviolet Irradiation Device with Programmable Exposure and Active Cooling for Nucleic Acid Crosslinking

Accessible and reproducible ultraviolet (UV) exposure is important for many chemical and biochemical applications. We describe an open-source laboratory irradiation device designed for psoralen-mediated RNA photocrosslinking, a valuable tool for nucleic acid structure determination. The device is based on 24 nominally 365-370 nm light-emitting diode modules arranged as two selectable 12-emitter series arrays with separate drivers. A Nano-compatible microcontroller, four pushbuttons and an OLED display provide user-programmable light-on and light-off intervals and an elapsed-time limit. The arrays share a fan-cooled aluminum heatsink and sample chamber, and the device can operate inside a 4 {degrees}C refrigerator to limit sample warming. A 3D-printed enclosure incorporates a sliding cover and interchangeable tube and microplate configurations with an approximately 10 mm emitter-to-sample spacing. The design supports inexpensive, configurable irradiation for nucleic-acid sample preparation, and can be adapted for other light-triggered activation of biological samples.

bioengineering↗

Programmable Imaging of Locus-Specific DNA Repair Dynamics Reveals Chromatin-Controlled Therapy Resistance in Cancer

DNA double-strand breaks (DSBs) pose a significant threat to genomic integrity, and their efficient repair is critical for cell survival. However, the spatial and temporal aspects of the repair process at specific genomic loci remain a mystery. To address this gap, we developed a novel live-cell imaging platform called Genomic Locus Imaging and DSB Repair Monitoring by Elastin-like Proteins (GLIDRE). This system leverages a CRISPR-guided elastin-like protein (ELP) droplet approach to specifically label non-repetitive genomic loci and enables real-time analysis of repair-associated dynamics at the same site. Using GLIDRE, we demonstrate that DSB repair-associated dynamics are significantly faster in open chromatin regions compared to compacted regions, irrespective of local transcriptional activity, supporting a model in which chromatin accessibility defines the baseline response whereas transcriptional activity acts in a context-dependent manner. In a model of diffuse intrinsic pontine glioma (DIPG), the H3.3K27M mutation reshapes chromatin accessibility and repair dynamics in a locus-specific manner. At the tumor suppressor locus CDKN2A, H3.3K27M mutation caused a more repressive chromatin state and delayed repair-associated dynamics, which was associated with resistance to CAR T-cell therapy. Pretreatment with the histone deacetylase inhibitor suberoylanilide hydroxamic acid (SAHA) restores repair dynamics at CDKN2A and sensitizes H3.3K27M tumor spheroids to CAR T-cell therapy. Together, our findings reveal chromatin accessibility as a key determinant of DSB repair dynamics and support a mechanistic link and actionable connection between chromatin state, DNA repair, and immunotherapy response.

bioengineering↗