Self-Supervised Text-Vision Alignment for Automated Brain MRI Abnormality Detection: A Multicenter Study (ALIGN Study).

Should Spinal MRI Be Routinely Performed in Patients With Clinically Isolated Optic Neuritis?

Diagnosing Intracranial Hemorrhage (ICH) at an early stage is difficult since it affects the blood vessels in the brain, often resulting in death. We propose an ensemble of Convolutional Neural Networks (CNNs) combining Squeeze and Excitation-based Residual Networks with the next dimension (SE-ResNeXT) and Long Short-Term Memory (LSTM) Networks in order to address this issue. This research work primarily used data from the Radiological Society of North America (RSNA) brain CT hemorrhage challenge dataset and the CQ500 dataset. Preprocessing and data augmentation are performed using the windowing technique in the proposed work. The ICH is then classified using ensembled CNN techniques after being preprocessed, followed by feature extraction in an automatic manner. ICH is classified into the following five types: epidural, intraventricular, subarachnoid, intra-parenchymal, and subdural. A gradient-weighted Class Activation Mapping method (Grad-CAM) is used for identifying the region of interest in an ICH image. A number of performance measures are used to compare the experimental results with various state-of-the-art algorithms. By achieving 99.79% accuracy with an F-score of 0.97, the proposed model proved its efficacy in detecting ICH compared to other deep learning models. The proposed ensembled model can classify epidural, intraventricular, subarachnoid, intra-parenchymal, and subdural hemorrhages with an accuracy of 99.89%, 99.65%, 98%, 99.75%, and 99.88%. Simulation results indicate that the suggested approach can categorize a variety of intracranial bleeding types. By implementing the ensemble deep learning technique using the SE-ResNeXT and LSTM models, we achieved significant classification accuracy and AUC scores.

Prospective Clinical Trial Registration: A Prerequisite for Publishing Your Results.

Bone age assessment is of great significance to genetic diagnosis and endocrine diseases. Traditional bone age diagnosis mainly relies on experienced radiologists to examine the regions of interest in hand radiography, but it is time-consuming and may even lead to a vast error between the diagnosis result and the reference. The existing computer-aided methods predict bone age based on general regions of interest but do not explore specific regions of interest in hand radiography. This paper aims to solve such problems by performing bone age prediction on the articular surface and epiphysis from hand radiography using deep convolutional neural networks. The articular surface and epiphysis datasets are established from the Radiological Society of North America (RSNA) pediatric bone age challenge, where the specific feature regions of the articular surface and epiphysis are manually segmented from hand radiography. Five convolutional neural networks, i.e., ResNet50, SENet, DenseNet-121, EfficientNet-b4, and CSPNet, are employed to improve the accuracy and efficiency of bone age diagnosis in clinical applications. Experiments show that the best-performing model can yield a mean absolute error (MAE) of 7.34 months on the proposed articular surface and epiphysis datasets, which is more accurate and fast than the radiologists. The project is available at https://github.com/YameiDeng/BAANet/, and the annotated dataset is also published at https://doi.org/10.5281/zenodo.7947923.

Pharmacists in UK National Health Service (NHS) hospitals have a long tradition of involvement in the development of drug policy. This paper describes various approaches that have been employed in the development and implementation of drug policy in hospitals and examines the evidence for their effectiveness and acceptability in the context of a changing health service. A series of focused interviews was conducted with a range of staff, including doctors, nurses, pharmacists and managers at eight hospitals. Interview sites were selected on the basis of a national survey of clinical pharmacy roles and were broadly representative of UK NHS hospitals. Interview data were analysed using constant comparison and analytic induction. Three models used in the development of drug policy were identified: a 'traditional' model, in which a drug and therapeutics committee establishes a hospital-wide formulary which is implemented by pharmacists; a 'combined' model, in which there is much more emphasis on tailoring policies and feedback to specialties or clinical directorates; and a 'medical control' model, in which prescribing decisions are made by individual doctors without reference to explicit policies and with little active pharmacy involvement. Pharmacy involvement was seen as vital to the development of effective policies but hospital-specific factors influenced the choice of model at particular sites. Hospitals may be moving towards the 'combined' model which could have advantages in the current internal market within the NHS. However, evaluations of the various approaches to drug policy should help inform this decision.

Intracranial hemorrhage (ICH) is an acute-stroke type leading to a high mortality rate. ICH diagnoses can keep patients out of life-threatening and longlasting consequences. These diagnoses are typically detected by analyzing images of Computed Tomography (CT) scan and symptoms of the patients. Consequently, the expertise of radiologists is significant in terms of detecting signals of ICH and making in-time decisions to save these patients. Over the past few decades, due to powerful-and-modern machine learning algorithms, several computer-aided detection (CAD) systems have been developed. They are becoming an effective part of routine clinical work to detect a few critical diseases, such as strokes, cancers, and so on. In this paper, we introduce a CAD that combines a deep-learning model and typical image processing techniques to determine whether patients suffer from ICH due to their CT images. The deep-learning model based on MobileNetV2 architecture was trained on RSNA (Radiological Society of North America) Intracranial Hemorrhage dataset. Then it was validated on a dataset of ICH-Vietnamese cases collected from Vinh Long Province Hospital, Vietnam. The experiment indicated that the classifier achieved AUC of 0.991, sensitivity of 0.992, and specificity of 0.807. After the deep-learning model identifies ICH-suspected slices, the Hounsfield Unit (HU) method is employed to highlight specific hemorrhage areas in these slices. DBSCAN is also used to remove noises detected by the HU method.

William G. Bradley, Jr, MD, PhD, FACR (1942-2017).

BackgroundCortical multiple sclerosis lesions are clinically relevant butinconspicuous at conventional clinical MRI. Double inversion recovery(DIR) and phase-sensitive inversion recovery (PSIR) are more sensitivebut often unavailable. In the past 2 years, artificial intelligence (AI)was used to generate DIR and PSIR from standard clinical sequences (eg,T1-weighted, T2-weighted, and fluid-attenuated inversion-recoverysequences), but multicenter validation is crucial for furtherimplementation.PurposeTo evaluate cortical and juxtacortical multiple sclerosis lesiondetection for diagnostic and disease monitoring purposes on AI-generatedDIR and PSIR images compared with MRI-acquired DIR and PSIR images in amulticenter setting.Materials and MethodsGenerative adversarial networks were used to generate AI-based DIR(n = 50) and PSIR (n = 43) images.The number of detected lesions between AI-generated images andMRI-acquired (reference) images was compared by randomized blindedscoring by seven readers (all with >10 years of experience inlesion assessment). Reliability was expressed as the intraclasscorrelation coefficient (ICC). Differences in lesion subtype weredetermined using Wilcoxon signed-rank tests.ResultsMRI scans of 202 patients with multiple sclerosis (mean age, 46 years± 11 [SD]; 127 women) were retrospectively collected from sevencenters (February 2020 to January 2021). In total, 1154 lesions weredetected on AI-generated DIR images versus 855 on MRI-acquired DIRimages (mean difference per reader, 35.0% ± 22.8;P < .001). On AI-generated PSIR images, 803lesions were detected versus 814 on MRI-acquired PSIR images (98.9%± 19.4; P = .87). Reliability was good for bothDIR (ICC, 0.81) and PSIR (ICC, 0.75) across centers. Regionally, morejuxtacortical lesions were detected on AI-generated DIR images than onMRI-acquired DIR images (495 [42.9%] vs 338 [39.5%]; P< .001). On AI-generated PSIR images, fewer juxtacortical lesionswere detected than on MRI-acquired PSIR images (232 [28.9%] vs 282[34.6%]; P = .02).ConclusionArtificial intelligence–generated double inversion-recovery andphase-sensitive inversion-recovery images performed well compared withtheir MRI-acquired counterparts and can be considered reliable in amulticenter setting, with good between-reader and between-centerinterpretative agreement.Published under a CC BY 4.0 license.Supplemental material is available for thisarticle.See also the editorial by Zivadinov and Dwyer in this issue.

Background Accurately diagnosing lymph node metastasis (LNM) in patients with rectal cancer preoperatively can guide treatment but remains challenging. Purpose To develop and validate an enhanced MRI model incorporating the so-called dark-light-dark sign on contrast-enhanced T1-weighted MRI scans, tumor features, and European Society of Gastrointestinal and Abdominal Radiology (ESGAR) criteria for predicting pathologic LNM (pLNM) and to compare its efficacy to a nonenhanced MRI model without the dark-light-dark sign, ESGAR size and lymph node features alone, and lymph node short-axis diameter alone. Materials and Methods This two-center retrospective study included patients with rectal cancer who underwent primary surgery (center 1, January 2016 to December 2024; center 2, January 2016 to January 2019). The dark-light-dark sign was defined as the presence of an inner hypointense tumor (dark), an intermediate hyperintense rim (light), and an outer hypointense muscularis propria (dark). Enhanced and plain scan models were developed by using logistic regression. Diagnostic performance was compared with the receiver operating characteristic curve. Results The training set consisted of 922 patients, 354 of whom were pLNM positive (mean age, 60 years ± 12 [SD]; 211 men) and 568 of whom were pLNM negative (mean age, 61 years ± 11; 347 men); the external test set consisted of 149 patients, 73 of whom were pLNM positive (mean age, 61 years ± 12; 51 men) and 76 of whom were pLNM negative (mean age, 62 years ± 11; 45 men). The enhanced model yielded areas under the receiver operating characteristic curve (AUCs) of 0.89 (95% CI: 0.86, 0.90) and 0.83 (95% CI: 0.76, 0.90) in the training and external test sets, respectively, outperforming lymph node short-axis diameter alone, ESGAR alone, and the nonenhanced model (all P < .001). The enhanced model outperformed ESGAR features in subgroups with lymph node short-axis diameters of less than 5 mm (training set: AUC, 0.88 [95% CI: 0.85, 0.92]; external test set: AUC, 0.90 [95% CI: 0.80, 0.96]) and pathologic T2 stage (training set: AUC, 0.89 [95% CI: 0.84, 0.93]; external test set: AUC, 0.91 [95% CI: 0.80, 1.00]). Conclusion Integrating tumor characteristics-especially the dark-light-dark sign-with ESGAR status improved preoperative LNM diagnosis, particularly in patients with lymph nodes smaller than 5 mm or who had stage T2 tumors. © The Author(s). Published by the Radiological Society of North America under a CC BY 4.0 license. Supplemental material is available for this article.

Transcranial US is a noninvasive, portable, real-time imaging technique that enables targeted evaluation of the intracranial circulation and select brain parenchymal structures through cranial bone windows. Traditionally, transcranial Doppler (TCD) has been used in well-established clinical scenarios, including vasospasm monitoring, stroke risk assessment, and ancillary testing for brain death. TCD encompasses two main techniques: nonimaging TCD (niTCD) and TCD imaging (TCDi). niTCD is a "blind" US technique that relies on standardized depths for vessel localization, guided by characteristic flow waveforms. In contrast, TCDi is a triplex approach that integrates gray-scale (B-mode) US, color Doppler US, and spectral Doppler US into an image-guided examination, enabling direct visualization of the intracranial anatomy and improving diagnostic confidence and reproducibility. In parallel, transcranial B-mode sonography (TCS), incorporated into modern TCDi, has demonstrated promising applications in adult patients, both in the neurointensive care unit and in outpatient practice. In addition, contrast-enhanced US (CEUS) has emerged as a powerful adjunct, markedly enhancing the Doppler signal in patients with poor bone windows, and has expanded the scope of transcranial US toward new exploratory vascular and cerebral perfusion applications. This article provides a comprehensive update on transcranial US in adults, including gray-scale imaging, Doppler techniques, and CEUS applications. With a focus on evidence-based indications and practical techniques, this article aims to support broader awareness and clinical adoption of transcranial US in contemporary adult neuroimaging, with a particular emphasis on TCDi, which remains underrecognized and underused despite its demonstrated clinical value in select indications. © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license.

Purpose To develop and validate a deep learning (DL) model for automated assessment of coronary stenosis in vessels with heavily calcified plaques at coronary CT angiography (CCTA), using quantitative coronary angiography as the reference standard. Materials and Methods A total of 10 101 CCTA examinations (June 2017-December 2020) from three tertiary hospitals in China were retrospectively collected for DL model development. External testing dataset 1 included 442 CCTA examinations (Agatston score > 300) from two independent hospitals (January 2021-May 2022) for performance evaluation. The separate external testing dataset 2 of 120 CCTA examinations was used for a reader study assessing whether DL assistance improved diagnostic accuracy among junior, attending, and senior radiologists. External testing dataset 3 included 150 prospectively collected CCTA examinations (June-July 2023) that were analyzed to compare model performance against clinical reports, simulating real-world deployment. Model diagnostic performance was assessed using receiver operating characteristic analysis, with quantitative coronary angiography as the reference. Results In external testing dataset 1, specificities for detecting 50% or more stenosis were 78%, 72%, and 48% and the areas under the receiver operating characteristic curve (AUC) were 0.89, 0.90, and 0.87 at the segment, vessel, and patient levels, respectively. In external testing dataset 2, DL assistance improved radiologist specificity by 7%-11% (P < .001) with improving AUC and increased interreader agreement (Δκ = 0.155-0.228; P < .05). In external testing dataset 3, the model demonstrated 53% specificity and a higher AUC versus clinical reports (0.91 vs 0.76; P < .001). Conclusion The proposed DL model accurately detected coronary stenosis of heavily calcified plaques at CCTA and improved diagnostic performance of radiologists. Keywords: CT Angiography, Cardiac, Heart, Arteriosclerosis, Calcifications, Calculi, Quantification, Diagnosis Supplemental material is available for this article. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license. See also commentary by Maiter and Alabed in this issue.

BackgroundCurrent clinical imaging modalities such as CT and MRI provide resolutionadequate to diagnose cardiovascular diseases but cannot depict detailedstructural features in the heart across length scales. Hierarchicalphase-contrast tomography (HiP-CT) uses fourth-generation synchrotronsources with improved x-ray brilliance and high energies to providemicron-resolution imaging of intact adult organs with unprecedenteddetail.PurposeTo evaluate the capability of HiP-CT to depict the macro- to microanatomyof structurally normal and abnormal adult human hearts ex vivo.Materials and MethodsBetween February 2021 and September 2023, two adult human donor heartswere obtained, fixed in formalin, and prepared using a mixture ofcrushed agar in a 70% ethanol solution. One heart was from a 63-year-oldWhite male without known cardiac disease, and the other was from an87-year-old White female with a history of multiple known cardiovascularpathologies including ischemic heart disease, hypertension, and atrialfibrillation. Nondestructive ex vivo imaging of these hearts withoutexogenous contrast agent was performed using HiP-CT at the EuropeanSynchrotron Radiation Facility.ResultsHiP-CT demonstrated the capacity for high-spatial-resolution, multiscalecardiac imaging ex vivo, revealing histologic-level detail of themyocardium, valves, coronary arteries, and cardiac conduction systemacross length scales. Virtual sectioning of the cardiac conductionsystem provided information on fatty infiltration, vascular supply, andpathways between the cardiac nodes and adjacent structures. HiP-CTachieved resolutions ranging from gross (isotropic voxels ofapproximately 20 µm) to microscopic (approximately 6.4-µmvoxel size) to cellular (approximately 2.3-µm voxel size) inscale. The potential for quantitative assessment of features in healthand disease was demonstrated.ConclusionHiP-CT provided high-spatial-resolution, three-dimensional images ofstructurally normal and diseased ex vivo adult human hearts. Whole-heartimage volumes were obtained with isotropic voxels of approximately 20µm, and local regions of interest were obtained with resolutiondown to 2.3–6.4 µm without the need for sectioning,destructive techniques, or exogenous contrast agents.© The Author(s) 2024. Published by the Radiological Society of North America under a CC BY 4.0 license.Supplemental material is available for thisarticle.See also the editorial by Bluemke and Pourmorteza in this issue.

Enhanced myometrial vascularity (EMV) is a common postpregnancy (ie, postpartum, postpregnancy termination, and postpregnancy loss) sonographic finding that represents involuting myometrial hypervascularity of the placental bed in the early puerperium and may persist if retained products of conception (RPOC) are present. On color Doppler US images, EMV is characterized by tortuous and dilated myometrial vessels with high-velocity and turbulent flow. These features can be easily mistaken for an arteriovenous malformation (AVM). However, despite sharing overlapping flow patterns, AVM and EMV are fundamentally distinct. AVMs are vascular congenital malformations, whereas EMV is a dynamic finding in the postpregnancy state related to persistent or involuting uteroplacental circulation. Most EMVs that come to medical attention regress spontaneously after expulsion of the associated RPOC. AVMs, on the other hand, are not associated with pregnancy, do not regress spontaneously, and are difficult to treat, often requiring repeated embolization sessions with liquid embolic agents. By becoming familiar with EMV, radiologists may better serve patients and referring clinicians with guidance for further investigation or treatment (when appropriate), thereby preventing unnecessary interventions, hysterectomies, or potential complications associated with such treatments. The authors clarify the definitions of EMV, AVM, and other pertinent terms, discuss the physiology of the uteroplacental circulation and postpregnancy EMV, address the pathogenesis of persistent EMV associated with RPOC, and provide an overview of the imaging features and management of EMV. ©The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license. Supplemental material is available for this article. See the invited commentary by Kirsch and Ponder in this issue.

Background MRI has helped to identify mesorectal fascia (MRF) positivity as a poor prognostic factor in patients with locally advanced rectal cancer. However, outcomes in patients with MRF positivity may vary. Purpose To investigate the prognosis of the four types of MRF-positive classifications and their genetic and protein-level interpretations. Materials and Methods In this multicenter retrospective study (January 2009 to May 2023), patients with locally advanced rectal cancer who underwent rectal MRI were analyzed and categorized as follows: MRF caused by tumor invasion (tumor invasion vs no tumor invasion), extramural vascular invasion (EMVI; EMVI vs no EMVI), lymph node metastasis (LNM; LNM vs no LNM), and tumor deposits (TD; TD vs no TD). Disease-free survival (DFS) was estimated using the Kaplan-Meier method and compared using the log-rank test. Prognostic factors were evaluated using multivariable Cox proportional hazard models. Tumor-associated genes were sequenced, tumor mutational burden (TMB) was calculated, and immunohistochemistry was performed to quantify the expression of Ki-67, as well as the infiltration densities of CD3+ and CD8+ T cells. Results A total of 959 patients from three centers were included (median age, 56 years [IQR, 48-63 years]; 622 men). Kaplan-Meier analysis showed MRF with TD and MRF with EMVI were associated with worse DFS compared with MRF without TD (median DFS, 18 vs 99 months, respectively; P < .001) and MRF without EMVI (median DFS, 39 vs 99 months, respectively; P = .01). At multivariable Cox regression analysis, MRF with TD and MRF with EMVI were independent risk factors for poorer DFS (hazard ratio, EMVI vs TD: 1.95 [P = .006] vs 2.55 [P < .001], respectively). TMB and T cell infiltration (CD3+ and CD8+) were lower in patients with MRF and EMVI compared with MRF without EMVI (log-transformed TMB: -0.49 vs -0.44, respectively [P = .04]; mean counts, CD3+: 103.4 vs 251.4, respectively; mean counts, CD8+: 72.3 vs 183.0, respectively [P < .001 for all]). Ki-67 expression was higher in patients with MRF with TD versus MRF without TD (median, 70% vs 60%, respectively; P = .01). Conclusion MRF caused by TD or EMVI at pretreatment rectal MRI predicted poor prognosis in patients with locally advanced rectal cancer. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Horvat and Kim in this issue.

To monitor the effects of dietary treatment in adult-onset adrenoleukodystrophy (ALD) by means of somatosensory evoked potentials (SEPs) and motor evoked potentials (MEPs). SEPs and MEPs have proved useful in revealing signs of progressively severe, central dying-back axonopathy in early stages of adult-onset ALD. Eight patients with adult-onset ALD underwent clinical examination, brain and spine MRI, and SEP and MEP studies before and after 3 years of Lorenzo's oil dietary therapy. Before treatment, brain MRI was normal in five patients. Three of these patients had pure spinal SEP abnormalities and in the remaining two patients SEPs showed signs of involvement of both the spinal and cerebral somatosensory tracts. After treatment, the three patients with pure spinal abnormalities showed clinical and neurophysiologic worsening, whereas the two patients with a more advanced stage of disease (exhibited by SEPs) showed substantially unchanged clinical and neurophysiologic features. The patients with abnormal brain MRI at the onset of treatment showed clinical and neurophysiologic worsening. Lorenzo's oil therapy had no effect on patients with evidence of inflammatory brain lesions. Moreover, in patients without clear signs of inflammatory damage, this treatment does not modify significantly the natural course of the disease. However, because effective treatments should begin before the onset of severe neurologic symptoms, SEPs and MEPs should be considered to evaluate the effectiveness of other experimental treatments in the patient with a negative brain MRI.