Searcharxiv⌕ Search

arXiv · 2611.11121

Phase-Constrained Self-Supervised Reconstruction and Denoising for Ultra-High-Resolution MRI

Abstract

Ultra-high-field gradient-echo (GRE) MRI reveals fine anatomy with strong susceptibility contrast, but small voxels reduce signal-to-noise ratio (SNR), while susceptibility-induced phase variation and wrapping complicate reconstruction. These challenges are especially consequential for quantitative susceptibility mapping (QSM), which requires denoising GRE phase without blurring anatomy or distorting phase. We propose a measurement-domain self-supervised learning (SSL) framework for joint reconstruction and denoising of complex-valued images from a single noisy acquisition. By splitting acquired k-space samples into input and held-out subsets, the method trains without clean or fully sampled references. An explicit phase-modulation operator, estimated from low-resolution phase and embedded in the unrolled network's encoding model, simplifies phase patterns encountered by the learned regularizer while preserving consistency with acquired complex k-space measurements. We evaluated ex vivo mouse brain MRI at 15, 18, and 72 um isotropic resolution; in vivo mouse brain MRI at 30 um isotropic resolution; and in vivo human brain MRI at 0.3x0.3x0.4mm3 resolution. Across controlled noise levels, SSL improved image-quality metrics in most conditions, preserved fine anatomy, and reduced error propagation into T2* maps and QSM. Ablation studies showed that phase regularization suppressed localized artifacts from rapid phase variation, while in vivo experiments demonstrated denoising with structural preservation. These findings support phase-constrained, measurement-domain SSL for high-resolution anatomical and quantitative MRI when clean references are unavailable.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Juan Liu, Zhuoheng Liu, Hongbo Wu, Xinyue Han, Jiaen Liu, Nian Wang. 2026-11-08. Phase-Constrained Self-Supervised Reconstruction and Denoising for Ultra-High-Resolution MRI. https://searcharxiv.com/abs/2611.11121

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Deep Learning-based Multiparametric Quantitative Ultrasound for Hepatic Steatosis Assessment

Quantitative ultrasound (QUS) has shown promise for liver fat quantification in metabolic dysfunction-associated fatty liver disease (MAFLD). Incorporating multiple parameters has demonstrated the potential to outperform single QUS parameter measures. Here, we propose a deep learning-based multiparametric QUS framework that learns effective integration of multiple features from QUS measures to estimate pixel-wise hepatic fat fraction, enabling the detecting of heterogeneous fat accumulation within the liver. We developed a U-Net based network architecture to predict local fat fraction percentages, which takes 2D QUS maps as input and is trained with magnetic resonance imaging-proton density fat fraction (MRI-PDFF) as the ground truth. We extracted 2D QUS maps (H-scan frequency, at-tenuation, Nakagami, and RF/envelope/log-compressed data) and tested various input configura-tions consisting of different combinations of the QUS parameters. We enrolled 30 patients with healthy livers or suspected/confirmed MASLD, and acquired ultrasound RF data, and validated our approach using the patient data. Among all tested input configurations, the model combining H-scan, attenuation, and Nak-agami parameters achieved the best performance, with a Spearman correlation coefficient (Rs) of 0.92 and an area under the curve (AUC) of 0.98. The predicted fat fraction maps exhibited color changes concordant with the gradual increasing in MRI-PDFF, from normal liver to 45% fat ac-cumulation. These findings demonstrate that deep learning can integrate multiparametric QUS in-formation in a fully data-driven manner, providing a localized fat fraction map rather than a single whole-liver value. We anticipate that our deep learning framework can offer potential advantages for future assessment of heterogeneous disease development in liver for more complex, progres-sive liver diseases such as steatohepatitis.

physics.med-ph↗

BEAM3R: Beam's-eye-view architecture with Mamba-3 for implicit dose reconstruction

To enable accurate and rapid photon control point and proton beamlet dose calculation in the DoseRAD2026 challenge, we present BEAM3R, a dose estimation framework operating in beam's-eye-view (BEV). Our core innovation combines a Mamba-3 state-space depth-sequence core with physics-based transport conditioning to model long-range depth transport without expensive 3D convolutions. BEAM3R shares a 2D CNN encoder-decoder architecture for photon and proton dose tasks, processing per-plane BEV slices. Proton beamlets are conditioned on water equivalent thickness and remaining range, encoding the parameters determining Bragg peak position. Photon models use a bidirectional Mamba-3 core to capture dose contributions from materials downstream of the calculation point, while the proton model uses a forward core with learned energy-prefix tokens and a Bragg-peak refinement module. To reduce interpolation artifacts and support high spatial resolution, we introduce axial grid alignment of BEV lattices with CT slices and an implicit super-resolution representation arranged as channels, evaluated by a differentiable Triton-accelerated resampler that reconstructs packed cubic B-spline coefficients directly in CT space. For MRI-based tasks, synthetic CTs (sCT) are generated using a conditional GAN with a SwinUNETR generator and a PatchGAN discriminator. On the preliminary DoseRAD2026 test set, CT-to-photon and CT-to-proton models achieved 1%/1 mm local gamma pass rates of 96.8% and 96.0%, with stratified plan-level MAEs of 0.0041 and 0.0079. Substituting sCT reduced gamma pass rates to 89.7% for photon and 75.4% proton plan level doses, with stratified plan-level MAEs of 0.0093 and 0.0336. Standardised runtimes were 23.4 s and 18.4 s for CT-to-photon and CT-to-proton prediction, increasing to 39.7 s and 42.8 s for the corresponding MRI-based pipelines.

physics.med-ph↗

Dynamic infrared thermography and computed tomography angiography fusion for DIEP flap perforator selection: a novel measurement system

Accurate perforator selection is critical for successful deep inferior epigastric artery perforator (DIEP) flap breast reconstruction. While computed tomography angiography (CTA) provides high-resolution anatomical detail, it lacks functional perfusion assessment. Dynamic infrared thermography (DIRT), in contrast, captures superficial rewarming dynamics but lacks objective anatomical validation. This study presents a standardized multimodal measurement system enabling automated fusion of DIRT with CTA-derived perforator coordinates. The framework integrates controlled thermographic acquisition, real-time hotspot detection, metric spatial calibration, and global one-to-one assignment of thermal hotspots to CTA perforators within a predefined periumbilical region of interest. The workflow was evaluated in 12 patients. Of 46 CTA perforators within the region of interest, 84.8% were successfully linked to thermographic hotspots, with a median spatial offset of 14.74 mm. Automated assignment excluded 76.1% of detected hotspots due to lack of anatomical correspondence, thereby reducing subjective interpretation. The proposed system enables quantitative multimodal validation of functional and anatomical imaging and provides objective support for perforator selection in clinical practice.

physics.med-ph↗