80 papers
This study demonstrates that a fully automated 3D nnU-Net framework can effectively segment head and neck cancer tumors from CT-only images, offering a cost-effective and scalable alternative to multimodal approaches that improves performance when trained on combined public and private datasets.
This prospective study demonstrates that multiparametric ultrasound time-harmonic elastography can effectively differentiate malignant from benign liver lesions by quantifying elevated tumor habitat viscosity and increased spatial stiffness heterogeneity, with diagnostic accuracy significantly improving for larger tumors.
This paper presents a practical, projection-based statistical framework that efficiently estimates noise-induced spectral bias in CT systems with high accuracy compared to Monte Carlo simulations while significantly reducing computational runtime, thereby enabling rapid, bias-aware optimization of spectral acquisition parameters.
This exploratory study demonstrates that photoacoustic imaging can non-invasively detect significant differences in muscle water, lipid, and hemoglobin ratios between patients with m.3243A>G mitochondrial myopathy and healthy controls, highlighting its potential as a novel biomarker for monitoring disease progression.
This retrospective multicenter study demonstrates that DeepMS, a deep learning model integrating focal lesion and normal-appearing white matter features from routine MRI, achieves high diagnostic accuracy for multiple sclerosis that surpasses current McDonald criteria biomarkers and retains efficacy even when focal lesions are masked.
This study reveals that increased functional connectivity within white matter networks, but not between white and gray matter, is significantly associated with social impairment severity in individuals with Autism Spectrum Disorder, offering new insights into the neural mechanisms underlying the disorder.
This study demonstrates that a transfer learning-based nnU-Net model, pretrained on TOF-MRA and fine-tuned on 7T 4D Flow MRI data, outperforms existing deep learning architectures in intracranial artery segmentation and provides the most accurate, automated hemodynamic quantification, thereby confirming that segmentation precision directly impacts the reliability of derived flow metrics.
This study establishes age- and sex-specific reference ranges for cardiac structure and function using cardiovascular magnetic resonance imaging data from a large, population-based German cohort to provide a normative framework for clinical interpretation and research on healthy aging.
This paper proposes a resource-efficient, two-stage conditional diffusion framework that generates synthetic PET scans from CT data to enable equitable oncology screening in rural settings by preserving critical metabolic biomarkers like SUV under simulated low-resource imaging conditions.
This prospective R-IDEAL Stage 0-2a study demonstrates that a 6-point quantitative Dixon sequence offers superior geometric accuracy, quantitative concordance, and reproducibility for PDFF and R2* estimation across 1.5T and 3T MR-Simulation and MR-Linac devices compared to 2- and 3-point methods, thereby validating its use for adaptive radiation therapy and bone marrow characterization in the pelvis and head and neck.
This study demonstrates the technical feasibility of implementing 3D-QALAS on a 1.5T MR-Linac to achieve whole-brain, 1 mm isotropic quantitative T1, T2, and PD mapping with high accuracy and reproducibility within a 7-minute acquisition time, paving the way for integrating these biomarkers into adaptive radiation therapy workflows.
This paper presents the first-in-human clinical demonstration of a novel ionizing radiation acoustic beam localization (iRABL) system that achieves sub-diffraction-limit spatial resolution and pulse-by-pulse imaging speed to enable real-time, high-accuracy mapping of proton dose deposition during treatment, marking a significant step toward image-guided "proton surgery."
This multi-institutional study demonstrates that converting chest radiographs from 16-bit to 8-bit depth does not significantly affect the performance of deep learning models in classifying sex, age, and obesity, suggesting that 8-bit images can be used for efficient data storage and processing without compromising diagnostic accuracy.
This study demonstrates that mask-aware embeddings derived from a medical foundational segmentation model (MedSAM2), when fused with clinical data, significantly improve the prediction of progression-free survival in multiple myeloma patients compared to clinical-only or radiomics baselines.
This paper introduces a lesion-centric latent phenotype learning pipeline for breast ultrasound that utilizes mask-weighted pooling of segmentation encoder latents and calibration to discover interpretable morphological phenotypes, achieving superior malignancy detection and descriptor consistency compared to radiomics and standard CNN baselines.
This study demonstrates that while activating external laser positioning systems (ELPS) during MRI simulation generally preserves quantitative values, it significantly degrades image quality—specifically causing a four-fold signal-to-noise ratio drop and geometric distortion errors when using the integrated body coil—necessitating clear clinical guidelines to avoid ELPS interference during imaging.
This prospective study demonstrates that magnetic resonance elastography (MRE) metrics, particularly tumor-to-normal brain tissue ratios of loss modulus and complex shear modulus, can effectively differentiate radiation necrosis from recurrent brain metastases by revealing that necrotic tissue is mechanically stiffer and more dissipative than tumor tissue.
By integrating structural MRI and cortical transcriptomics, this study identifies two distinct cortical ageing subtypes—a normative metabolic-immune decline and a compensatory stress-repair adaptation—linked to specific longevity gene expression profiles.
This study leverages the largest population-based clinical MRI cohort to date, demonstrating that individuals with mental disorders exhibit shared neuroanatomical alterations, such as smaller subcortical volumes and thinner cortices, compared to controls, thereby validating the utility of real-world clinical scans for patient stratification and future cross-disorder research.
This paper presents an intelligent handheld ultrasound system that combines an Actor-Critic reinforcement learning framework for automated probe navigation with YOLOv8n-based detection and hybrid segmentation algorithms to achieve expert-level accuracy in real-time carotid artery and thyroid examinations, significantly reducing operator dependency.