MRI of the heart and abdominal organs provides unparalleled soft tissue contrast and quantitative biomarkers, yet remains highly susceptible to physiological motion. Contractions of the myocardium, respiratory excursions, peristalsis, vascular pulsatility, and unpredictable bulk patient movement generate artifacts that impair image quality, limit reproducibility, and may necessitate repeat scans. This review summarizes motion correction strategies in cardiac and abdominal MRI, emphasizing both clinical applications and methodological principles. Techniques to address motion can be broadly categorized into prospective and retrospective approaches. Prospective methods adjust acquisition in real time, for example through respiratory or cardiac gating, navigator echoes, or external sensors, while retrospective strategies apply corrections during or after reconstruction, using k-space binning, image registration, or model-based reconstructions. Rigid motion, such as translations or rotations of organs, can often be corrected efficiently, whereas non-rigid motion including myocardial contraction or peristalsis requires more sophisticated elastic registration or motion-compensated reconstruction. Application-specific challenges and solutions are highlighted across cardiac cine imaging, flow quantification, tagging, and quantitative mapping, as well as abdominal imaging of the liver, kidneys, and gastrointestinal tract. In each domain, examples are provided of how motion impacts diagnostic performance and how motion correction strategies can mitigate these effects. Strengths and limitations of current approaches are reviewed, from conventional breath-holding to advanced free-breathing motion-resolved imaging. Emerging trends include integration of artificial intelligence with motion-compensated reconstruction, advanced sensor technologies for real-time tracking, and hybrid approaches combining multiple strategies. While many methods remain research-focused, vendor-embedded solutions and open-source tools are increasingly available, narrowing the gap between technical advances and routine practice. Motion correction is poised to become a core feature of clinical MRI, enabling faster, more robust, and patient-friendly examinations that reduce repeat rates, improve diagnostic confidence, and expand access to high-quality imaging in challenging patient populations. EVIDENCE LEVEL: N/A. TECHNICAL EFFICACY: Stage 5.

Primary angle-closure glaucoma (PACG), an irreversible blinding disease characterized by retinal ganglion cell damage and optic nerve atrophy, exerts significant effects on brain functional networks. Using resting-state functional magnetic resonance imaging (rs-fMRI) data from 34 PACG patients and 34 matched healthy controls (HCs), we extracted four types of connectivity features-voxel-wise static functional connectivity (FC), dynamic functional connectivity (dFC), effective connectivity (EC), and dynamic effective connectivity (dEC)-via the AAL90 (Automated Anatomical Labeling 90) atlas following preprocessing. Elastic net feature selection was applied independently to each connectivity type to retain the top 10% most discriminative features. We evaluated the classification performance of ten machine learning models using individual feature types as well as their combined features, with the FC-based logistic regression (LR) model achieving optimal diagnostic efficacy (accuracy = 0.92, AUC = 0.96). SHapley Additive exPlanations (SHAP) of the model identified 20 critical connections, revealing abnormal patterns at both the region of interest (ROI)-level and network-level within brain networks such as the visual network (VSN), dorsal attention network (DAN), and sensorimotor network (SMN). Statistical group comparisons validated reduced connectivity (e.g., VSN-SMN, VSN-DAN) and enhanced DAN-thalamus connectivity in patients, while voxel-wise analyses of key regions confirmed diminished connectivity to visual areas. The results provide insights into how machine learning can be effectively employed to detect PACG-specific brain network disruptions and highlight potential neuroimaging biomarkers.

Decisions to perform abdominal aortic aneurysm (AAA) repair are dependent on aneurysm size, but variation in size measurement leads to inconsistency in management. AI-assisted systems have potential to improve repeatability of measuring aortic dimensions. This study compared the repeatability and agreement in clinical decision-making between using artificial intelligence (AI)-automated and traditional semi-automated methods for measuring abdominal aortic aneurysm (AAA) size. Computed tomography angiogram scans from 142 patients who had scans at baseline (n = 142), 1 year (n = 100), 2 years (n = 56) and 3 years (n = 4) were analysed using semi-automated and AI-assisted automated systems. Three observers measured maximal AAA diameter and volume twice with each method. Intra- and inter-observer repeatability were assessed using reproducibility coefficients (RC). Measurements were used to decide if AAA repair was required according to clinical guidelines, and agreement was evaluated using Kappa coefficients (K). The ability of AI-assisted measurements to predict actual requirement for AAA repair was assessed using Cox proportional hazard analysis. AI-assisted measurements had perfect intra- and inter-observer repeatability (RC = 0) which were significantly superior to traditional measurements (RC for diameter: 1.9-5.6 mm; volume: 7.8-22.6 cm³, p < 0.001). Agreement about AAA repair was superior using AI-assisted (K = 1) than traditional (K = 0.55-0.70) measurements. Baseline AI-assisted measurements predicted actual requirement for AAA repair (Hazard ratio, HR, per mm diameter increase 1.12, 95% confidence intervals, CI, 1.03-1.22, HR per cm³ volume increase 1.02, 95% CI 1.01-1.02, p < 0.001). The findings suggest AI-assisted measurement of AAA size would enhance the consistency of decisions about AAA repair.

Pretreatment determination of histological differentiation grade is critical for prognostic evaluation in laryngeal and hypopharyngeal squamous cell carcinoma (LHSCC). This study aimed to develop a contrast-enhanced CT (CECT)-based Vision Transformer (ViT) model for noninvasive evaluation of histological grades in LHSCC. In this retrospective multicenter study, a total of 1,648 LHSCC patients who underwent CECT scans were enrolled from three hospitals in this study. Participants were divided into a training cohort (n = 1,239), an internal validation cohort (n = 310) from one hospital, and an external validation cohort (n = 99) from the other two hospitals. The diagnostic model integrates a pre-trained ViT for CECT feature extraction and an XGBoost classifier for prediction. The model's predictive performance was evaluated using the area under the curve (AUC), decision curve analysis (DCA), and calibration curve. The ViT model achieved AUCs of 0.887 (95%CI: 0.848-0.927) in internal validation and 0.796 (95%CI: 0.693-0.899) in external validation cohorts, significantly outperforming the conventional radiomics model (AUCs: 0.775, 95%CI: 0.714-0.837 and 0.544, 95%CI: 0.388-0.699; p < 0.001 and 0.002, respectively). Clinically, DCA demonstrated superior clinical utility, while calibration curves showed excellent prediction reliability. Gradient-weighted Class Activation Mapping visualization identified CT image regions most influential for the model's predictions, providing interpretability for clinical decision-making. The ViT-based deep learning model developed in this study using CECT demonstrated excellent predictive performance for histological grading of LHSCC, with promising application for patient prognosis assessment.

X-ray phase contrast imaging (XPCI), when implemented in micro-computed tomography (micro-CT) mode, offers high-contrast 3D imaging of weakly-attenuating material samples. In the so-called single-mask edge illumination approach, a mask with periodically spaced transmitting apertures is used to split the x-ray beam into narrow beamlets; when the beamlets are aligned with the boundaries ('edges') between detector pixels, their refraction-induced deviation can be detected and used to form images. A shortcoming is that the mask reduces the x-ray flux, necessitating longer exposures and therefore longer acquisition times. We show that the demand on exposure time can be relaxed by integrating the deep learning-based denoising technique Noise2Inverse into the image processing workflow. The applicability of Noise2Inverse to single-mask edge illumination XPCI micro-CT is demonstrated, and its performance at severe noise levels is explored. Taking advantage of the distinct imaging system characteristics, we also propose an adaptation of Noise2Inverse, called Noise2Phase, which does not rely on splitting the CT dataset by projections.

Quantification of the calcific burden is valuable in percutaneous coronary intervention (PCI) planning and in research to assess its changes after pharmacotherapies targeting plaque progression. In intravascular ultrasound (IVUS) images this analysis is currently performed manually and time consuming. To overcome these limitations, we introduce a deep-learning (DL) method for seamless detection of the calcific tissue. IVUS images from 197 vessels were analysed by an expert who identified the presence of calcium, and these estimations were used to train a DL model for fast detection of calcific deposits. The output of the model was tested in a set of 30 vessels against the estimations of the two experts. Comparison was performed at a frame-, lesion- and segment level. In total 26,211 frames were included in the training and 5,138 in the test set. The estimations of the DL method for the presence of calcium were similar to the experts (kappa 0.842 and 0.848, p < 0.001), while the correlation between the DL approach and the two experts for the arc of calcium (0.946 and 0.947, p < 0.001) and calcific area (0.745 and 0.706, p < 0.001) were high. Lesion- (0.971 and 0.990, p < 0.001) and segment-level analysis (0.980 and 0.981, p < 0.001) demonstrated a high correlation between the method and the two experts for calcific burden. The proposed DL method is able to accurately detect the calcific tissue and quantify its burden. These features render it useful in research and are expected to facilitate its application in the clinical workflows to guide PCI.

Normal pressure hydrocephalus (NPH) is a potentially treatable neurodegenerative disorder that remains underdiagnosed due to its clinical overlap with other conditions and the labor-intensive nature of manual imaging analyses. Imaging biomarkers, such as the callosal angle (CA), Evans Index (EI), and Disproportionately Enlarged Subarachnoid Space Hydrocephalus (DESH), play a crucial role in NPH diagnosis but are often limited by subjective interpretations. To address these challenges, we developed a fully automated and robust deep learning framework for measuring the CA directly from raw T1 MPRAGE scans. Our method integrates two complementary modules. First, a BrainSignsNET model is employed to accurately detect key anatomical landmarks, notably the anterior commissure (AC) and posterior commissure (PC). Preprocessed 3D MRI scans, reoriented to the Right Anterior Superior (RAS) system and resized to standardized cubes while preserving aspect ratios, serve as input for landmark localization. After detecting these landmarks, a coronal slice, perpendicular to the AC-PC line at the PC level, is extracted for subsequent analysis. Second, a UNet-based segmentation network, featuring a pretrained EfficientNetB0 encoder, generates multiclass masks of the lateral ventricles from the coronal slices which then used for calculation of the CA. Training and internal validation were performed using datasets from the Baltimore Longitudinal Study of Aging (BLSA) and BIOCARD, while external validation utilized 376 clinical MRI scans from Johns Hopkins Bayview Hospital, as well as PENS trial. Our framework achieved high concordance with manual measurements, demonstrating a strong correlation (r = 0.98, p < 0.001) and a mean absolute error (MAE) of 3.26 (SD 1.89) degrees. Moreover, error analysis confirmed that CA measurement performance was independent of patient age, gender, and EI, underscoring the broad applicability of this method. These results indicate that our fully automated CA measurement framework is a reliable and reproducible alternative to manual methods, outperforms reported interobserver variability in assessing the CA, and offers significant potential to enhance early detection and diagnosis of NPH in both research and clinical settings.

Chronic obstructive pulmonary disease is a common respiratory disease. The severity of acute exacerbation of chronic obstructive pulmonary disease is related to disease progression and risk of death. However, the existing grading standards mainly depend on indicators, such as respiratory rate, whether to apply assisted respiratory muscles, and changes in consciousness state, and only reflect the subjective judgment. Imaging omics can extract muscle characteristic data for more complex analysis, which helps to provide a more objective and accurate method to assess the severity of disease for clinic. The purpose of this study is to construct a severity prediction model based on the combination of chest CT muscle imaging features and clinical data in hospitalized patients with AECOPD. 234 hospitalized patients with AECOPD were retrospectively included, divided into 79 grade I, 74 grade II, and 81 grade III. Clinical data and chest CT images were collected. Construction of clinical feature model combined with muscle imaging omics model based on Python machine learning platform. The number of hospitalizations for acute exacerbation, disease course, risk of acute exacerbation in stable stage, white blood cell count, neutrophil count, creatinine, and N-terminal B-type natriuretic peptide precursor were statistically different among hospitalized patients with AECOPD in the last year (all P < 0.05). The best model to predict the severity of AECOPD by cascade probability combination method is Xgboost model with AUC of 0.890. The disease grading prediction model of AECOPD inpatients constructed based on clinical data and muscle imaging omics characteristics has good performance, and has great potential in assisting clinicians to more accurately stratify the risk of AECOPD inpatients.

Cochlear implants substantially improve spoken language in children with severe to profound sensorineural hearing loss, yet outcomes remain more variable than in children with healthy hearing. This variability cannot be reliably predicted for individual children using age at implant or residual hearing. Development of an artificial intelligence clinical tool to predict which patients will exhibit poorer improvements in language skills may enable an individualized approach to improve language outcomes. To compare the accuracy of traditional machine learning (ML) with deep transfer learning (DTL) algorithms to predict post-cochlear implant spoken language development in children with bilateral sensorineural hearing loss using a binary classification model of high vs low language improvers. This multicenter diagnostic study enrolled children from English-, Spanish-, and Cantonese-speaking families across 3 independent clinical centers in the US, Australia, and Hong Kong. A total of 278 children with cochlear implants were enrolled from July 2009 to March 2022 with 1 to 3 years of post-cochlear implant outcomes data. All children underwent pre-cochlear implant 3-dimensional volumetric brain magnetic resonance imaging (MRI). ML and DTL algorithms were trained to predict high vs low language improvers in children with cochlear implants using neuroanatomical features from presurgical brain MRI. Data were analyzed from August 2023 to April 2025. Cochlear implants. The accuracy, sensitivity, and specificity of prediction models based on brain neuroanatomic features using traditional ML and DTL learning. Of 278 children, 137 (49.3%) were female, and the mean (SD) age at cochlear implant was 25.7 (18.8) months. DTL prediction models using bilinear attention-based fusion strategy achieved an accuracy of 92.39% (95% CI, 90.70%-94.07%), sensitivity of 91.22% (95% CI, 89.98%-92.47%), specificity of 93.56% (95% CI, 90.91%-96.21%), and area under the curve of 0.98 (95% CI, 0.97-0.99). DTL outperformed traditional ML models in all outcome measures. The results of this diagnostic study suggest that DTL prediction of language improvement on the individual child level using neuroanatomic features demonstrates greater accuracy, sensitivity, and specificity than traditional ML prediction. DTL was substantially improved by direct capture of discriminative and task-specific information that are advantages of representation learning enabled by this approach vs ML. The results support the feasibility of a single DTL prediction model for language prediction for children served by cochlear implant programs worldwide. Prediction of low improvement may enable targeted early and customized intervention to improve language.

This review aims to summarize the research progress of magnetic resonance imaging (MRI)-based deep learning ( DL) in meningiomas, analyze its advantages, limitations, and key issues in clinical translation, provide technical references for relevant medical researchers and clinicians, thereby promoting the faster and more standardized application of DL in clinical diagnosis and treatment, and ultimately benefiting patients. The early detection and accurate grading and classification of meningiomas are crucial for formulating personalized treatment plans. DL has achieved breakthrough progress in the field of meningioma imaging analysis. By adopting objective and quantitative analysis methods, it effectively overcomes the limitation of traditional diagnostic methods that rely on subjective human visual judgment, opening up broad prospects for the precise diagnosis and treatment of meningiomas. The literature search and selection process for this review was conducted as follows: Search period: 1 January 2019 to 31 October 2024; Databases searched: PubMed, Web of Science, and Embase; Search string: (("meningioma" OR "meningiomas") AND ("magnetic resonance imaging" OR "MRI") AND ("deep learning" OR "convolutional neural network" OR "CNN" OR "transformer" OR "neural network" OR "neural networks")). The application of DL in meningioma research marks that medical imaging diagnosis has entered a new intelligent stage. By providing doctors with more objective and accurate diagnostic basis, it facilitates the formulation of personalized treatment plans, thereby improving patients' treatment outcomes and quality of life. The continuous breakthroughs of DL in the field of meningiomas indicate that the future of medical imaging diagnosis will be more intelligent and precise.