Luca Giannoni, Uzair Hakim, Fréderic Lange, Musa Talati, Darshana Gopal, Angelos Artemiou, Niccole Ranaei-Zamani, Subhabrata Mitra, Ilias Tachtsidis
Optical imaging and spectroscopy solutions, such as near-infrared spectroscopy (NIRS) and diffuse optical tomography (DOT), have the potential to provide compact, bedside monitoring of the placenta in the clinic, thanks to recent advancements in miniaturisation and wireless wearability. This would provide neonatologist with continuous assessment of the pregnancy status in real-time, as well as tools to possibly predict delivery outcomes. We present here OptoCENTAL, a standardized platform based on multiple optical phantoms, from digital, through solid to liquid, for a comprehensive bench-testing, characterisation and validation of any photonics solution and instrumentation that aims at in vivo, clinical monitoring of the human placenta. Results: Exemplary applications of the OptoCENTAL platform on different types of optical systems, from wearable, continuous-wave devices to broadband and time-domain NIRS systems, demonstrate the flexibility of its procedures to be implemented with any setup, allowing users to compare performances across different solutions. The results also show the capability of OptoCENTAL to provide quantitative assessment of the major features required by any photonic solution for providing effective and efficient monitoring of the placenta, including basic instrument performances, quantification of monitoring accuracy, as well as depth sensitivity. OptoCENTAL represent the first-of-a-kind effort in standardising bench-testing and validation of optical imaging and spectroscopy methods in the framework of placental clinical applications, further advancing the translation of such modalities into the hospitals, as well as towards future certification and commercialisation of such technologies.
Rizal Maulana, Ádám Rák, Sándor Földi, György Cserey
Continuous monitoring of physiological signals is essential for the early detection of health problems. A measurement system that ensures high sensitivity, accuracy, and user comfort is needed. In this study, we designed and optimized a flexible piezoresistive yarn (FPY) sensor to achieve a high sensitivity and wide working range for detecting physiological signals. The representative sensor design was constructed by applying an FPY bonding pattern, utilizing tightly arranged triangular patterns and using minimal FPY. The prototype sensor operates in two measurement modes, strain and pressure, and was evaluated for measuring neck motion, finger bending, respiratory signals, and arterial blood pressure (ABP) waveforms. A qualitative evaluation, performed by comparing the characteristics of the measurement results of each physiological signal with those from related studies, indicates a high similarity in its morphological characteristics. Then, a quantitative evaluation through baseline drift analysis demonstrates that the FPY sensor displays high measurement stability. The ABP waveform measurement shows the most stable baseline, with a mean absolute error (MAE) of $0.0051 \pm 0.0029$ in terms of baseline drift, using normalized values from 0 to 1. Based on our results, the prototype sensor can be used as an innovative solution for physiological signal monitoring and can be further enhanced for personalized healthcare and sports applications.
Harry Penketh, Sonal Saxena, Michal Mrnka, Cameron P. Gallagher, Caitlin Lloyd, Diksha Garg, Christopher R. Lawrence, Nicholas E. Grant, John D. Murphy, David B. Phillips, Ian R. Hooper, Nick Stone, Euan Hendry
Accurate characterisation of margins in excised breast cancer tumours is critical to the success of surgical interventions, yet margin status is typically confirmed post-operatively using histopathology. Here we present a new approach to intraoperative margin assessment based on microwave single pixel imaging, demonstrating tissue phantom hydration mapping across large areas (~10 cm x 10 cm) at ~1 mm resolution. By leveraging the photo-induced change in microwave transparency of a silicon modulator placed under the sample, we map the microwave reflectivity and identify positive margins with deeply sub-wavelength resolution. We test the discriminatory capabilities of our approach using gelatine-based tumour phantoms with variations in water density representative of the margin and cancerous tissues of a resected tumour. We demonstrate the capability to identify, locate and quantify inadequate margins up to the typically targeted minimum thickness of 2 mm. Furthermore, using numerical modelling, we show that our approach is expected to be resilient to patient-specific tissue differences. Our technique has potential for future deployment as a real-time intraoperative tissue margin analysis tool.
Jason G Parker
An explicit positronium (Ps) source model was implemented in Geant4 to provide direct event-level control over annihilation channel selection, decay timing, and photon emission topology. The implementation supports direct annihilation, para-positronium (p-Ps), and ortho-positronium (o-Ps) branches with user-defined fractions, explicit routing of o-Ps events to two-photon (2-gamma) or three-photon (3-gamma) decay, exponential or fixed delay sampling, optional prompt-photon emission, and optional positron-range displacement. Event-level truth metadata were retained to support downstream validation and analysis. The implementation was evaluated in controlled Geant4 studies using native reference configurations, explicit branch-fraction sweeps, lifetime sweeps, timing benchmarks, and a frozen point-source downstream test harness. Observed 2-gamma and 3-gamma fractions followed the requested control parameters with the expected linear behavior, and measured mean delays reproduced the prescribed lifetime settings with near one-to-one agreement. Computational cost scaled linearly with event count, with modest overhead relative to native Geant4 operation. A minimal downstream validation framework was used to verify branch-consistent handling of pure and mixed datasets, including expected method-source compatibility and recovery of valid events under unified 2-gamma and 3-gamma routing. These results establish a practical and internally consistent code framework for explicit positronium modeling in Geant4 and provide a simple pathway for PET researchers to incorporate controlled Ps generation into existing simulation pipelines.
Qiyuan Shi, Yi Li
Vision Transformers (ViTs) have shown strong empirical performance on high-dimensional medical imaging data, yet their behavior under survival objectives and the interpretability of their attention mechanisms remain poorly understood. Under shallow ViTs, we design controlled experiments showing that token-level attention dynamics can recover outcome-relevant regions and that attention-based thresholding enables effective token pruning, improving both interpretability and predictive performance. We also study pretrained deep ViTs for survival analysis and propose a radiomics-guided hybrid model that integrates pixel-based embeddings with interpretable radiomic features through a multimodal Cox framework and contrastive alignment. Applied to a COVID-19 chest X-ray cohort with a composite ICU admission or mortality endpoint, the proposed approach achieves competitive discrimination while providing clinically meaningful attention maps and feature-group importance.
Tianyuan Wang, Daniël M. Pelt, Felix Lucka, Tristan van Leeuwen, K. Joost Batenburg
Traditional X-ray computed tomography (CT) scanning strategies typically select projection angles uniformly and allocate dose equally. In practice, however, CT scans often need to be fast, radiation-efficient, and adaptive. Sparse-view tomography addresses these requirements by reducing both the number of angles and the total dose budget. Under such constraints, angle selection and dose allocation should be information-driven, with more dose assigned to informative directions. To this end, we propose a dose-aware acquisition and reconstruction framework that combines a PWLS-PnP reconstruction backbone with an RL-based strategy for adaptive angle selection, explicitly accounting for angle-dependent photon statistics. Numerical experiments show that the proposed approach improves overall reconstruction quality and enhances defect detectability compared with conventional strategies, particularly when only a small number of projections or a constrained dose budget is available.
Fabian Bschorr, Pia Gebhard, Tobias Speidel, Volker Rasche
Purpose: Quasi-random Sobol-based sampling schemes exhibit deterministic structural artifacts when aggressively undersampled, particularly at low encoding densities required for accelerated 2D SPI/CSI. To address these limitations, two advanced undersampling strategies are investigated to mitigate deterministic behavior, improving image quality for time-constrained applications such as hyperpolarized MRI. Methods: An optimized Sobol sequence-derived point distribution with Heaviside-type density gradient center oversampling served as the initial sampling pattern. Undersampling was performed using two point-reduction algorithms: radius-adaptive stochastic undersampling (RAST), which applies a geometric, radius-dependent minimum-distance criterion, and Bayesian Information Gain Optimization (BINGO), that removes points based on their information gain to the reconstructed image. Phantom experiments were conducted on a 3 T clinical MRI system using up to 16-fold undersampling. Image quality was quantified using a performance score derived from RMSE, SSIM, and HFEN. Results: Both RAST and BINGO outperformed deterministic undersampling across all metrics. RAST achieved highest and most robust performance, with improvements up to 238% in the averaged metric score, while BINGO yielded improvements of 133% across matrix resolutions. Conclusion: The proposed strategies effectively reduce the number of encoding points in low-discrepancy 2D SPI point distributions while maintaining image quality under strong acceleration. RAST provides superior metric performance, whereas BINGO offers broad applicability, including suitability for non-linear encoding fields. These approaches support rapid acquisition workflows required for real-time and hyperpolarized applications.
Benedetta Fantaci, Alejandro Frechilla, Matteo Frigelli, Philippe Büchler, Sabine Kling, Begoña Calvo
Accurate assessment of corneal mechanical properties is critical for understanding ocular biomechanics, predicting refractive surgery outcomes, and optimizing cross-linking (CXL) treatments. Conventional uniaxial tensile test is limited by non-physiological boundary conditions and simplified stress distributions. Inflation testing more closely reproduces the in vivo stress state but has traditionally lacked full-field deformation mapping. In this work, we present an integrated experimental-computational protocol combining inflation testing of freshly enucleated porcine eyes with high-resolution three-dimensional digital image correlation (3D-DIC). Fifteen corneas were analyzed across three cohorts: (i) de-epithelialized controls, (ii) CXL-treated (standard Dresden protocol), and (iii) anterior stromal ablation via femtosecond laser. Samples were subjected to controlled intraocular pressure (IOP) elevations up to 40 mmHg. The 3D-DIC approach provided dense, pointwise displacement and strain maps across the anterior surface, successfully quantifying the localized stiffening effects of CXL and the increased compliance induced by stromal ablation. These full-field kinematic data were integrated into a membrane-theory finite element framework to resolve principal in-plane strains, that were used for subsequent inverse modeling to derive anisotropic hyperelastic parameters of porcine corneal tissue. Overall, the method establishes an end-to-end route from physiologic loading to full-field strain mapping and constitutive parameter identification, enabling quantitative evaluation of treatment-induced biomechanical changes in the cornea.
Yetao He, Wenhan Guo, Deliang Wei, Evan Bel, Ji Yi, Yu Sun
Three-dimensional (3D) wide-field fluorescence microscopy is a widely used modality for volumetric imaging, but suffers from characteristic out-of-focus blur. Existing reconstruction methods either struggle to operate on high-dimensional volumes or fail to provide credibility characterization of the reconstruction. In this work, we introduce Volumetric Transport (VOLT), a 3D-native probabilistic framework for wide-field fluorescence microscopy reconstruction. VOLT combines a transport-based formulation that maps degraded measurements to clean volumes via stochastic interpolants with a 3D-native anisotropic network that separates lateral and axial processing. This design operates directly in voxel space and achieves improved scalability to large volumes without relying on slice-wise approximations. We develop both stochastic (SDE) and deterministic (ODE) variants within the same framework. We validate VOLT on simulated wide-field microscopy datasets. Our results show that VOLT significantly improves reconstruction quality in both lateral and axial directions while providing voxel-wise credibility estimates.
Marta I. Bracco, Jorge Malouf, Laurent Maimoun, Xavier Nogues, Jean Paul Roux, François DuBoeuf, Ludovic Humbert
Apr 19, 2026·q-bio.QM·PDF Background: Three-dimensional dual-energy X-ray absorptiometry reconstructs three-dimensional maps of the proximal femur's density distribution from standard hip scans, enabling the estimation of trabecular and cortical bone parameters. The aim of this study was to assess the agreement of these three-dimensional cortical and trabecular femur parameters across different series and models of Hologic densitometers. Methodology: The study cohort was composed of 103 women and men recruited from four clinical centers in Spain and France. Subjects had duplicated hip scans using different Hologic scanners from the Horizon, Discovery, and QDR4500 series. Analyses were performed using 3D-Shaper software. Inter-scanner agreement was evaluated using Deming regression and Bland-Altman analysis. Results: The parameters demonstrated strong inter-device agreement across all clinical centers and scanner models, with coefficients of determination greater than 0.91. Absolute biases were less than 2.5 mg$/$cm$^3$ for integral volumetric bone mineral density, less than 2.9 mg$/$cm$^3$ for trabecular volumetric bone mineral density, and less than 1.7 mg$/$cm$^2$ for cortical surface bone mineral density. No statistically significant bias was found between parameters obtained from different scanners. Furthermore, the observed bias was lower than the expected least significant change, indicating that inter-scanner variability across these devices is not clinically significant. Conclusions: This study demonstrated excellent agreement for standard and three-dimensional derived bone parameters at the hip across Hologic densitometers. These findings support their suitability for clinical use.
Paul A. Constable, Dorothy A. Thompson, Irene O. Lee, Lynne Loh, Aleksei Zhdanov, Mikhail Kulyabin, Andreas Maier
The LEOPs (Light-ERG-Oscillatory Potentials) dataset provides light-adapted (LA) electroretinogram (ERG) and Oscillatory Potentials (OPs) waveforms for typically developing Control, Autism Spectrum Disorder (ASD) and ASD + Attention Deficit Hyperactivity Disorder (ADHD) childhood and adolescent populations. The ERGs were recorded in the Right And Left eyes with skin electrodes using the handheld RETeval device at two sites in Australia and the United Kingdom. The LEOPs dataset includes 5309 single flash ERG and 4434 OPs waveforms as well as images selected from each participant showing the position of the skin electrode. The LEOPs dataset is constructed from recordings using a 9 step randomized flash series from $-0.37$ to $1.20$~$Td.s$, a 2 step at 113 and 446 $Td.s$ flash strengths (2500 Control, 1730 ASD and 451 ASD + ADHD samples), as well as the $85$~$Td.s$ (Light Adapted 3 $cd.s.m^{-2}$ (LA3)) equivalent International Society of Clinical Electrophysiology of Vision (ISCEV) Standard flash with 435 Control, 176 ASD and 37 ASD + ADHD waveform samples. Code for the stimulus is provided along with participant demographics, date and time of testing, and where available diagnostic scores for the ASD and ASD + ADHD groups, alongside iris color, electrode position with image files and time domain values for the ERG and summed values for the OPs. The repository contains excel file, exported JSON files on the patient level that are more suitable for machine learning tasks, images of electrode position for each recording and the protocol files for use with the RETeval.
Grzegorz Bauman, Pavlos Panos, Philipp Latzin, Oliver Bieri
Purpose: To develop a robust deep learning framework for non-contrast-enhanced functional lung MRI, overcoming the limitations of spectral decomposition in the presence of physiological non-stationarity. Methods: We introduce VQ-Wave (Ventilation/Q-perfusion Waveform-based Assessment of Variable Evolutions), a physics-driven spatio-temporal inception neural network trained on synthetic signal models to estimate ventilation and perfusion parameters. By processing local spatial context alongside temporal evolution, the network learns to decouple physiological signals from noise. The training generator simulated non-stationary dynamics, including amplitude modulations, frequency drifts, and noise. Performance was validated against matrix pencil (MP) decomposition using numerical phantoms and in-vivo lung MRI acquired in four healthy volunteers and two children with cystic fibrosis (CF) at 1.5T. Results: In numerical benchmarks, VQ-Wave demonstrated superior robustness to non-stationarity, maintaining low global and regional error rates where MP exhibited stochastic instability due to spectral leakage. In-vivo, VQ-Wave accurately captured functional defects in patients with CF yielding ventilation and perfusion maps with high quantitative stability (mean variation < 12%) even when scan time was reduced from 45s to 15s. Conversely, under irregular physiology and short scan lengths, MP decomposition severely degraded, exhibiting systematic amplitude instability, overestimation bias, and regional signal dropouts. Conclusion: VQ-Wave offers a robust, physics-driven neural network-based alternative to spectral decomposition. By effectively handling physiological irregularity and noise, it enables reliable functional lung imaging with substantially shortened acquisition protocols.
Iram Barbaro Rivas Ortiz, Sahar Ranjbar, Piergiorgio Cerello, Emanuele Maria Data, Mohammad Fadavi Mazinani, Miguel David Fernandez Moreira, Veronica Ferrero, Simona Giordanengo, Felix Mas Milian, Diango Manuel Montalvan Olivares, Francesco Pennazio, Marco Pullia, Roberto Sacchi, Roberto Cirio, Simone Savazzi, Anna Vignati, Elisa Fiorina
Prompt Gamma Timing (PGT) is a promising technique for in vivo range verification in particle therapy, exploiting the time-of-flight between primary particles and prompt gamma rays emitted by nuclear interactions. PGT distribution is highly sensitive to beam energy and target density, which, under controlled detector positioning, enables real-time monitoring of particle range, detection of morphological changes, and support for adaptive treatment strategies. This study investigates for the first time the application of PGT in carbon ion therapy. Measurements were performed using a dedicated detection system composed of a silicon strip sensor for primary ion timing and a LaBr3(Ce) read out by a SiPM for secondary radiation. Carbon ion beams with energies of 166.41, 268.86, and 398.84 MeV/u irradiated a homogeneous 30.0 cm PMMA target at CNAO. The secondary radiation detector was positioned at four off-beam positions to assess the robustness of the PGT technique. Simulations based on Geant4 were conducted for all configurations to evaluate agreement and predictive capability. A bin-by-bin comparison of experimental and simulated PGT intensities demonstrated strong agreement within the 95% confidence interval, with no incompatible bins at 166.41 MeV/u, at most 1% at 268.86 MeV/u, and up to 8% at 398.84 MeV/u, depending on detector position. Photons were identified as the dominant contribution to the detected signals, particularly for detector positions upstream with respect to the primary particle beam, minimizing signal contamination from neutrons and charged fragments. The validated experimental-simulation framework confirms the capability of the proposed PGT system to resolve energy-dependent differences and highlights its potential for detecting clinically relevant changes in the particle beam range, supporting further development toward real-time monitoring in carbon ion therapy.
Neda Valizadeh, Robabeh Rahimi, Ramin Abolfath
{\bf Purpose}: To develop a geometry-governed diffusion framework that explains differential tissue response under FLASH ultra-high dose rate (UHDR) irradiation by explicitly accounting for structural heterogeneity and anomalous transport in biological tissues. {\bf Methods}: We formulate a generalized diffusion--reaction model on fractal substrates to describe molecular transport in heterogeneous media. Tissue architecture is characterized by a fractal (Hausdorff) dimension \(D\), while scale-dependent transport inefficiency and memory effects are captured by a fractional parameter \(θ\). Analytical solutions for radially symmetric geometries are derived and compared with classical normal (Euclidean) diffusion and a Gaussian reference model under identical physical conditions. Transport behavior is quantified through transient probability distributions and steady-state spatial profiles. {\bf Results}: The model reveals systematic suppression of long-range transport and enhanced localization as tissue structural complexity increases. Increasing \(θ\) leads to subdiffusive dynamics, reduced effective diffusion lengths, and persistent non-Gaussian concentration profiles, even in the steady state. While increasing \(D\) alone enhances spatial accessibility, fractional dynamics dominate transport behavior when \(θ>0\), counteracting geometric connectivity. These effects produce a separation between regimes characterized by efficient inter-track overlap and rapid homogenization, and regimes marked by isolated, long-lived reactive domains.
Marco Schlimbach, Moritz Rempe, Jessica Mnischek, Lukas T. Rotkopf, Jens Weingarten, Jens Kleesiek, Kevin Kröninger
Objective. Standard Magnetic Resonance Imaging (MRI) reconstruction pipelines discard phase information captured during acquisition, despite evidence that it encodes tissue properties relevant to tumor diagnosis. Current machine learning approaches inherit this limitation by operating exclusively on reconstructed magnitude images. The aim of this study is to build a generative framework which is capable of jointly modeling magnitude and phase information of complex-valued MRI scans. Approach. The proposed generative framework combines a conditional variational autoencoder, which compresses complex-valued MRI scans into compact latent representations while preserving phase coherence, with a flow-matching-based generative model. Synthetic sample quality is assessed via a real-versus-synthetic classifier and by training downstream classifiers on synthetic data for abnormal tissue detection. Main results. The autoencoder preserves phase coherence above 0.997. Real-versus-synthetic classification yields low AUROC values between 0.50 and 0.66 across all acquisition sequences, indicating generated samples are nearly indistinguishable from real data. In downstream normal-versus-abnormal classification, classifiers trained entirely on synthetic data achieve an AUROC of 0.880, surpassing the real-data baseline of 0.842 on a publicly available dataset (fastMRI). This advantage persists on an independent external test set from a different institution with biopsy-confirmed labels. Significance. The proposed framework demonstrates the feasibility of jointly modeling magnitude and phase information for normal and abnormal complex-valued brain MRI data. Beyond synthetic data generation, it establishes a foundation for the usage of complete brain MRI information in future diagnostic applications and enables systematic investigation of how magnitude and phase jointly encode pathology-specific features.
Gia-Bao Ha, Lucas Takanori Sanchez Shiromizu, Jaehyeon Song, Zhuyun Xie, Henry Crandall, Dinali Assylbek, Alexandra Boyadzhiev, Huanan Zhang, Fernando Guevara Vasquez, Ramakrishna Mukkamala, Michael Widlansky, Shamim Nemati, Jesse Capecelatro, C. Alberto Figueroa, Benjamin Sanchez
Continuous ambulatory monitoring of peripheral vascular perfusion could enable earlier detection of vascular dysfunction in individuals with diabetes mellitus and more timely management of cardiovascular disease. Clinical imaging modalities provide high-fidelity vascular information but are impractical for ambulatory use, whereas most wearable devices are limited to single-modality sensing and do not provide imaging. Electrical bioimpedance has the potential to bridge this gap by enabling rapid spatial and temporal imaging while remaining sensitive to hemodynamic changes. Here, we introduce a wearable ring with 8 electrodes and 32-channel bioimpedance sensing for finger blood flow imaging. In 96 healthy participants measured at rest and during autonomic maneuvers, we resolve conductivity images in the digital arteries associated with pulsatile blood flow and train neural network models for continuous cuffless blood pressure waveform estimation. We demonstrate the feasibility of bioimpedance imaging in a ring form factor, supporting its potential for ambulatory cuffless hemodynamic monitoring.
Moritz Zaiss, Amr Aly, Jonathan Endres, Tobias Dornstetter, Simon Weinmüller, Andreas Maier
Purpose: Novel MR sequence developments still today allow generation of new diagnostic tools or novel imaging biomarkers. Programming MRI pulse sequences, however, is time-consuming and requires deep expertise in sequence design, restrictions by hardware constraints and MRI physics; even small modifications often require substantial debugging and validation. LLMs can assist when given structured prompts and error feedback, but many generated sequences still exhibit physical inconsistencies. We present Agent4MR, an agent-based framework that automatically generates and refines PyPulseq sequences using a structured, physics-aware validation report. These agents can perform also autonomous research. Methods: We evaluated Agent4MR on a spin-echo EPI task across three state-of-the-art LLMs and compared it to a context-only baseline (LLM4MR) and to a human developer with the same tools. We tested an MR autoresearch on a fluid-suppressed spin-echo EPI challenge for three different model generations. Results: Across all models, Agent4MR consistently produced artifact-free, physically valid sequences in a single user interaction, reducing the number of required interactions below the human baseline while maintaining correct timing and k-space coverage. Autonomous agents could then improve a sequence to match a given target contrast in an autoresearch approach. Conclusion: An appropriate agentic harness with physics-based validation can turn general-purpose LLMs into reliable MRI sequence developers and may ultimately enable non-experts to refine or innovate MR sequences guided by biological or clinical questions, or let swarms of agents realize sequence programming for them. Keywords: MRI; pulse sequence; PyPulseq; large language models; agents; autoresearch, sequence development.
Fan Xiao, Nikolaos Delopoulos, Niklas Wahl, Lennart Volz, Lina Bucher, Matteo Maspero, Miguel Palacios, Muheng Li, Samir Schulz, Viktor Rogowski, Ye Zhang, Zoltan Perko, Christopher Kurz, George Dedes, Guillaume Landry, Adrian Thummerer
Purpose: Accurate dose calculation is essential in radiotherapy for precise tumor irradiation while sparing healthy tissue. With the growing adoption of MRI-guided and real-time adaptive radiotherapy, fast and accurate dose calculation on CT and MRI is increasingly needed. The DoseRAD2026 dataset and challenge provide a public benchmark of paired CT and MRI data with beam-level photon and proton Monte Carlo dose distributions for developing and evaluating advanced dose calculation methods. Acquisition and validation methods: The dataset comprises paired CT and MRI from 115 patients (75 training, 40 testing) treated on an MRI-linac for thoracic or abdominal lesions, derived from the SynthRAD2025 dataset. Pre-processing included deformable image registration, air-cavity correction, and resampling. Ground-truth photon (6 MV) and proton dose distributions were computed using open-source Monte Carlo algorithms, yielding 40,500 photon beams and 81,000 proton beamlets. Data format and usage notes: Data are organized into photon and proton subsets with paired CT-MRI images, beam-level dose distributions, and JSON beam configuration files. Files are provided in compressed MetaImage (.mha) format. The dataset is released under CC BY-NC 4.0, with training data available from April 2026 and the test set withheld until March 2030. Potential applications: The dataset supports benchmarking of fast dose calculation methods, including beam-level dose estimation for photon and proton therapy, MRI-based dose calculation in MRI-guided workflows, and real-time adaptive radiotherapy.
B. Kaoui, A. Bou Orm, P. Navet, J. Baish, L. L. Munn
Bicuspid valves with crescent-shaped leaflets are found in lymphatic vessels and veins, where their primary function is to prevent reflux and ensure unidirectional flow toward the heart. These valves are passive, and their functionality emerges spontaneously from a complex interplay between the properties of the valve leaflets and the flow patterns developing within the vessel sinus region surrounding the valve. The main function of the valves is to limit retrograde flow, or reflux, but the optimal valve structure has not been well-characterized. Here we investigate numerically how the length of the leaflets affects the valve efficiency in preventing reflux. The valves are subjected to backward flow, akin to that imposed by gravity. We report the flux through the valve orifice as a function of key parameters: valve length, leaflet length, and leaflet rigidity. We monitor the transition in the flow regime - from reflux to complete flow blockage - by varying only the leaflet length. The transition threshold is found to depend strongly on the valve shape and stiffness. We captured these control parameters numerically to evaluate the ability of the valve to close and prevent reflux. This study allowed us to explain reflux observed experimentally in certain incompetent abnormal and immature valves, particularly those with shorter leaflets.
Paula Arguello, Berk Tinaz, Mohammad Shahab Sepehri, Maryam Soltanolkotabi, Mahdi Soltanolkotabi
Deep learning underpins a wide range of applications in MRI, including reconstruction, artifact removal, and segmentation. However, progress has been driven largely by public datasets focused on brain and knee imaging, shaping how models are trained and evaluated. As a result, careful studies of the reliability of these models across diverse anatomical settings remain limited. In this work, we introduce MosaicMRI, a large and diverse collection of fully sampled raw musculoskeletal (MSK) MR measurements designed for training and evaluating machine-learning-based methods. MosaicMRI is the largest open-source raw MSK MRI dataset to date, comprising 2,671 volumes and 80,156 slices. The dataset offers substantial diversity in volume orientation (e.g., axial, sagittal), imaging contrasts (e.g., PD, T1, T2), anatomies (e.g., spine, knee, hip, ankle, and others), and numbers of acquisition coils. Using VarNet as a baseline for accelerated reconstruction task, we perform a comprehensive set of experiments to study scaling behavior with respect to both model capacity and dataset size. Interestingly, models trained on the combined anatomies significantly outperform anatomy-specific models in low-sample regimes, highlighting the benefits of anatomical diversity and the presence of exploitable cross-anatomical correlations. We further evaluate robustness and cross-anatomy generalization by training models on one anatomy (e.g., spine) and testing them on another (e.g., knee). Notably, we identify groups of body parts (e.g., foot and elbow) that generalize well with each other, and highlight that performance under domain shifts depends on both training set size, anatomy, and protocol-specific factors.