搜索结果：BJR artificial intelligence

Medical imaging is fundamental to healthcare systems, aiding in the detection of internal abnormalities related to medical conditions beyond their physical symptoms. In low- and middle-income countries (LMICs), limited access to advanced imaging and scarcity of radiologists for image interpretation are evident. Upgrading available resources with artificial intelligence can expand the diagnostic capacity of LMICs to manage the growing prevalence and incidence of infectious diseases such as tuberculosis (TB). Chest X-rays can act as an effective triage tool for TB screening, and multiple models have been reported to improve the number of cases detected in high-burden settings. The case-finding strategies reported in literature have also demonstrated improved diagnostic accuracy and turnaround time post adoption of artificial intelligence (AI) for chest X-ray interpretation. AI assistance can help in identifying radiological involvement of TB, irrespective of their clinical symptoms. Furthermore, cost-effective, integrated workflows can also efficiently support LMICs by facilitating parallel diagnosis and appropriate linkage to care for multiple chest disorders through a unified pathway, thereby broadening the capabilities of chest X-ray based TB screening. By optimizing and strengthening LMIC health systems with AI, further scale-up and implementation can foster a supportive ecosystem for early disease diagnosis and decentralized care delivery.

Artificial intelligence (AI) can revolutionize clinical workflows in radiology but requires organizational change. An institutional strategy to develop and evaluate AI tools is outlined. A multidisciplinary AI board with a patient and public involvement and engagement group was created. A comprehensive framework was formed, comprising workstreams covering information governance; technical rigor, performance, and safety; economic considerations; and ethical and medical-legal aspects. In addition to recurring meetings, a workshop with clinicians, information technology specialists, and patient representatives helped to identify priority use cases. Technical infrastructure was enhanced to support the development, performance assessment, and deployment of AI tools. Primary areas for AI applications included training staff, vetting of image requests, quality assurance, image interpretation, and communicating imaging findings to patients. Potential barriers, gaps in evidence, and subsequent actions for AI implementation were outlined. Avenues for collaboration with industry and market-available solutions were outlined. A virtual Picture Archiving and Communication System server was developed and then connected to a deployment platform for performance evaluation of AI products. Establishing an institutional AI board and imaging AI sandbox has guided safe, effective AI implementation while creating an ideal setting for innovation and industry partnership. Our approach to the integration of imaging AI provides a pragmatic guide for other institutions.

Accurate prognostic tools in patients with chronic liver disease (CLD) have the potential to improve clinical outcomes and reduce health care costs. Imaging studies of CLD patients analysed using artificial intelligence (AI) algorithms for segmentation, detection and classification tasks have the potential to inform and improve prognostic models for liver-related outcomes. In this narrative review, we provide an overview of the strengths, weaknesses and approaches to inclusion of AI in prognostic models that use ultrasound (US), computed tomography (CT) and magnetic resonance imaging (MRI). We then use a prognostic endpoint-based approach to examine AI-based US, CT and MRI prognostic models in chronic liver disease (CLD). We highlight how AI has been applied to extract imaging features or build predictive models directly, and assess the limitations that currently hinder clinical translation. We also outline key challenges specific to prognostication in CLD and propose directions for future research.

Pericoronary adipose tissue (PCAT) is increasingly recognised as a biosensor of vascular inflammation. The guideline-driven widespread adoption of coronary computed tomography angiography (CCTA) as the first-line investigation for coronary artery disease (CAD) has created opportunities for evaluating the inflammatory burden through quantitative assessment of PCAT. Standardising raw PCAT imaging data for technical, anatomical, and biological variability provides the Fat-Attenuation Index (FAI) Score, which shows promise as a metric of coronary inflammation. Quantification of coronary inflammation has implications for the diagnosis, risk stratification, and monitoring of treatment in atherosclerotic cardiovascular disease. This review examines the anatomical and physiological basis of PCAT, highlighting the importance of standardising PCAT imaging for the implementation as a clinical biomarker, and reviews the role of artificial intelligence (AI) in enhancing precision and scalability. Emerging evidence on the modulation of FAI Score by therapeutic agents, including statins, biologics, and cardiometabolic drugs, and the potential utility of serial imaging in guiding clinical care is also discussed. With ongoing large-scale validation and emerging AI -based approaches, PCAT imaging is poised to complement traditional risk factors and plaque metrics; however, current evidence remains evolving, and the integration of inflammatory risk assessment could be useful to guide emerging anti-inflammatory treatments in personalised cardiovascular medicine.

Artificial intelligence (AI) is increasingly integrated into neuroradiology practice, with a growing number of FDA-cleared algorithms now supporting tasks ranging from acute triage to volumetric analysis. This review provides a structured overview of commercially available, FDA-regulated AI tools in neuroradiology, organized by clinical application. These include detection and prioritization of intracranial hemorrhage and large vessel occlusion, aneurysm identification on CTA, automated ASPECTS scoring, and brain tumor segmentation, as well as tools for image enhancement and quantitative analysis in neurodegenerative and demyelinating diseases. For each application, we describe the algorithm's intended function, summarize available performance data, and highlight areas where AI can add clinical value, such as reducing time to diagnosis, improving detection of subtle findings, or standardizing measurements. We also discuss key limitations, including reduced performance outside intended-use parameters and the need for broader validation. As AI tools continue to evolve, understanding their strengths, limitations, and optimal use cases is essential to their safe and effective deployment in neuroradiology.

We aimed to use an artificial intelligence (AI)-based pleural effusion segmentation model on baseline 18F-FDG positron emission tomography/computed tomography (PET/CT) images to investigate the prognostic value of PET/CT-derived parameters for overall survival (OS) among lung cancer patients with malignant pleural effusion (MPE). A total of 146 patients with MPEs were recruited. An integrated AI segmentation model combining 3D spatially weighted and 2D classical U-Net segmented pleural effusion for 18F-FDG PET/CT parameter extraction. Cox regression analyses revealed independent 12-month survival predictors. The area under the receiver operating characteristic curve (AUC) and DeLong's test were used to evaluate the discriminant power of the predictors and the LENT score. Bootstrap resampling was employed for internal validation. The patients comprised 81 males (55.5%) and had a mean age of 61.7 (SD = 11.5) years. The key survival predictors included maximum standardised uptake value (SUVmax), metabolic tumour volume (MTV), and total lesion glycolysis (TLG). The combined PET/CT parameters demonstrated a statistically significant advantage over that of the LENT score for 12-month survival prediction (AUC: 0.849, 95% confidence interval (CI) 0.795-0.903 vs. AUC: 0.732, 95%CI 0.660-0.796). The internal bootstrap validation had an AUC of 0.840 (95% CI: 0.671-0.922) and demonstrated a well-fitting calibration curve. The baseline 18F-FDG-PET/CT parameters extracted using the deep learning model performed excellent in predicting MPE survival and may complement existing MPE survival models and guide clinical stratified treatment. AI-integrated 18F-FDG-PET/CT radiomics improved prognostic assessment of MPE, facilitating personalised interventions stratified by survival expectations.

Artificial intelligence (AI), particularly deep learning (DL), is transforming the field of medical imaging and holds substantial promise for advancing breast cancer screening. This narrative review explores current and emerging AI applications in mammography screening, including image-based cancer detection, risk prediction, and workflow optimization, with attention to technical foundations, performance metrics, and clinical utility. Evidence indicates that AI may enhance diagnostic accuracy, enable more personalized risk assessment and screening strategies, and reduce radiologist workload, which has implications for accessibility, especially in resource-limited settings with radiologist shortages. However, real-world implementation of these tools remains challenging due to limitations in algorithm generalizability to diverse populations, calibration and reader response behaviour concerns, as well as regulatory, ethical and legal obstacles. While the potential impact is considerable, broader adoption will depend on prospective validation, transparent performance reporting, and strong governance mechanisms to maintain safety, equity, and public trust.

To assess the cost-effectiveness of using artificial intelligence (AI)-derived software to assist reading CT scans of the chest to identify and analyse lung nodules compared to unaided reading in symptomatic, incidental and screening populations. Decision tree structures were developed in TreeAge Pro 2021. Structures were informed by British Thoracic Society clinical guidelines and clinical opinion. Results were presented as incremental cost-effectiveness ratios (ICERs) expressed as cost per quality-adjusted life-year (QALY) over a lifetime from the UK National Health Service and Personal Social Services perspective. For the symptomatic population, the unaided radiologist reading strategy dominated the AI-assisted reading strategy. In the incidental population, unaided radiologist reading was cost-effective with an ICER of approximately £1000 per QALY. Conversely, in the screening population, AI-assisted radiologist reading dominated unaided reading. The cause of AI assistance being cost-effective depended on the number of people who had undergone CT surveillance because of non-cancerous findings. Given the limitations in the quality and quantity of evidence to inform inputs, these results should be interpreted with caution. Current analyses based on limited evidence suggested that, in the symptomatic and incidental populations, unaided radiologist reading may be the more cost-effective strategy, while in the screening population, AI-assisted radiologist reading appeared to be the dominant strategy. Better quality evidence is required to have a definitive answer about their cost-effectiveness. This paper shows whether adding AI-derived software to radiologists' reading of CT scans to identify lung nodules offers good value for money.

Artificial intelligence (AI) holds great promise for advancing diagnostics and treatment in nuclear medicine. The rapid growth of AI over the past decade has been largely driven by advances in hardware components such as graphics processing units (GPUs) and the introduction of deep learning (DL) and convolutional neural networks (CNN). The integration of AI and medical imaging has the potential to revolutionize nuclear medicine by, for example, accelerating image acquisition, enhancing image quality, enabling advanced image generation, assisting image interpretation, and aiding treatment planning. Clinical applications have been demonstrated for most medical specialties, including oncology, neurology, and radionuclide therapy. The utilization of AI to provide automated, standardized procedures can help bring advanced imaging from major university centres to smaller local clinics, thus benefiting a broader range of patients. Additionally, AI has vast potential for predicting optimal treatment strategies, assessing risk, optimizing patient flow and outcomes, and even improving productivity, but these capabilities have yet to be fully utilized. The fraction of clinical AI applications in general healthcare reaching beyond the prototyping phase is reported to be as low as 2%. Indeed, in nuclear medicine, very few AI developments have reached commercial maturity. Currently, most AI applications in nuclear medicine follow the imaging flow from image acquisition and reconstruction, post-processing and image preparation, image analysis, and decision support for clinical interpretation. Below, we will briefly review selected areas and comment on challenges and opportunities for AI in nuclear medicine, with a special focus on the transition from development to clinical implementation.

Recent advances in artificial intelligence (AI) offer significant potential to address the growing bottleneck in radiology caused by an increasing volume of imaging studies amidst a global shortage of radiology professionals. This study presents a comprehensive review of the latest developments in AI, particularly in vision-language models for radiology report generation, providing radiologists with a current reference. We conducted a focused literature search for studies published from 2020 to 2024 and included 14 studies in our review specifically on chest X-ray datasets with limited coverage of 3D modalities, reflecting the early stage of research and ongoing methodological advances in report generation for volumetric imaging. We analysed the model architectures, report generation capabilities, training datasets, evaluation metrics, and performance of these models. Our review highlights the evolution of AI in radiology report generation and underscores the critical need for diverse datasets and standardized evaluation metrics. Despite rapid progress, current AI models are not yet capable of consistently producing high-quality reports and require further improvements in data diversity, model training, and evaluation metrics to achieve a level comparable to human experts.

Women continue to be disproportionately affected by a large burden of disease including cardiovascular disease, cancer, gynecologic disorders, osteoporosis, and maternal health complications contributing to significant morbidity and mortality. Traditional diagnostic tools and risk models often fail to account for sex-specific factors, leading to underdiagnosis and delayed care. Artificial intelligence (AI) is rapidly emerging as a transformative tool in women's health, offering new methods for opportunistic screening, early detection, and risk prediction across multiple conditions. This review explores the application of AI in radiology imaging with a focus on diseases that only affect women, and those that affect both men and women, focusing on the outcomes for women and how AI is affecting their care. We describe different applications of AI and summarize types of bias affecting these applications, with recommendation on strategies to mitigate these disparities. Neglecting women's health has profound economic, societal, and global health consequences. We hope by highlighting some transformative AI applications for women's health, we can promote their adoption to accelerate care, while factoring in various pitfalls that risk leaving women behind in the ongoing AI transformation. Specifically, we recommend a sociotechnical approach to AI development and deployment for women's health-factoring in the impact of complex social systems that have allowed persistent disparities and underinvestment in women's health.

Artificial intelligence (AI) models for diagnostic imaging face reproducibility challenges due to inconsistent reporting. Existing guidelines also lack specificity for imaging-based AI diagnostics, particularly regarding clinical usability and technical transparency. To address these gaps, the Completeness, Learnability, Applicability, Interpretability, Reproducibility, and Evaluation (CLAIRE) framework was developed as a practical reporting aid by a multidisciplinary team of clinicians and AI experts. CLAIRE was retrospectively validated on a subset of 10 imaging studies selected by theoretical saturation in medical and dental imaging. Internal validation demonstrated high reliability, with inter-rater agreement improving from Cohen's &#x3ba; 0.286 to 0.987 (p&#x2009;<&#x2009;0.01) after calibration, alongside a mean intra-rater reliability of 0.997 after a six-month washout period. This process yielded a 15-item structured checklist for standardising AI reporting, supported by an objective scoring system for quality categorisation and an editorial reference guide to facilitate systematic appraisal by reviewers and editors. CLAIRE aims to enhance clinician accessibility through plain-language technical summaries and assessments of real-world applicability. This proposal provides a unified and practical structure that improves reporting consistency, supports systematic assessment, and strengthens both reproducibility and clinical translation of AI-based imaging models.

Management guidelines for incidental aortic dilation detected on low-dose chest computed tomography (LDCT) lung cancer screening (LCS) are lacking. Therefore, this study aims to validate artificial intelligence (AI) software for automated aortic measurements and assess aortic dilation distribution in screening participants. Baseline LDCT scans from 2 tertiary centres (April 2017 to December 2023) were reviewed. In 100 randomly selected cases, radiologist- and AI-measured maximum aortic diameters (MADs) were compared at the ascending thoracic aorta (ATA), aortic arch (AACH), descending thoracic aorta (DTA), and abdominal aorta (AA). AI then analysed all scans, and coronary artery calcification (CAC) was assessed using the Agatston method to evaluate correlations with MADs. Overall, 1204 patients (99.2% men; mean age &#xb1; SD: 62.7&#x2009;&#xb1;&#x2009;5.4&#x2009;years) were included. Intraclass correlation coefficients between radiologists and AI were 0.950, 0.758, 0.933, and 0.931 for ATA, AACH, DTA, and AA, respectively. Mean maximum diameters were: ATA, 38.7&#x2009;&#xb1;&#x2009;3.7&#x2009;mm (33.4% &#x2265; 40&#x2009;mm, 18.5% &#x2265; 42&#x2009;mm, and 5.6% &#x2265; 45&#x2009;mm); AACH, 37.3&#x2009;&#xb1;&#x2009;3.3&#x2009;mm; DTA, 29.4&#x2009;&#xb1;&#x2009;2.9&#x2009;mm; and AA, 26.3&#x2009;&#xb1;&#x2009;2.3&#x2009;mm. MADs significantly correlated with CAC severity (P&#x2009;&#x2264;&#x2009;.001). AI software reliably measures MADs. Aortic dilation distribution may serve as a reference in LDCT LCS, and its association with CAC highlights the clinical importance of incorporating MADs into patient management. Validated AI software enables reliable MAD assessment; reported aortic dilation prevalence offers valuable reference data for LDCT LCS.

Liver diseases consistently plague people's daily lives as a result of their high morbidity and mortality rates. Ultrasound (US), favoured by its flexibility, free of radiation, cost-effectiveness, and real-time capabilities, has been commonly employed as one of the first-line imaging tools for hepatic conditions. Artificial intelligence (AI) algorithms are increasingly applied to automatically identify intricate patterns and perform quantitative analyses in US imaging, potentially reducing radiologists' workload and improving diagnostic efficiency. AI-based US has been of substantial assistance in detecting, diagnosing, screening as well as monitoring of various liver diseases, and has attracted extensive attention among the medical community. In this review, we present a general introduction to AI in medical imaging; we next review its rapidly evolving applications in liver US, covering evaluation of hepatic steatosis severity, assessment of liver fibrosis, identification of focal hepatic lesions, preoperative prediction of high-risk pathological characteristics, assessment of postoperative prognosis, and the analysis of the model of integrated application of multi-omics data; finally, we present an outlook on the clinical applications of AI-based US in the liver diseases.

The aim of this systematic review and meta-analysis is to evaluate the diagnostic accuracy of artificial intelligence (AI) based algorithms in detecting pneumoperitoneum on medical imaging. Online databases were searched until June 2024. Statistical analyses were conducted using Open Meta-Analyst software and STATA 17.0. The analysis included overall sensitivity, specificity, diagnostic odds ratio (DOR), and area under the curve (AUC). Meta-regression and subgroup analyses were conducted to identify sources of heterogeneity among the included studies. Among the 14 AI-based radiograph models analysed, AI demonstrated high diagnostic accuracy for pneumoperitoneum, with a sensitivity of 83.6% (95% CI: 80.2%-86.4%), specificity of 92.9% (95% CI: 88.3%-95.8%), negative likelihood ratio of 0.18, and positive likelihood ratio of 11.76 (all P&#x2009;<&#x2009;.001). Deep learning models showed higher sensitivity (83.7%) but slightly lower specificity (91.2%) compared to machine learning models (sensitivity 77%, specificity 98%). The AUC was 0.93, with a DOR of 76. Meta-regression revealed larger sample sizes significantly improved specificity. Deeks' funnel plot showed no publication bias. AI models are effective in diagnosing pneumoperitoneum. The high accuracy of these models enhances the potential for rapid and precise detection, thereby improving patient management. Future prospective multicentre studies with larger sample sizes and comparisons of various models are highly anticipated. This is the first meta-analysis to evaluate AI's diagnostic accuracy for pneumoperitoneum, revealing high sensitivity and specificity, comparing deep learning and machine learning performance, and highlighting AI's potential to enhance early diagnosis and prioritization in clinical workflows.

Radiology is undergoing a major shift with the growing use of artificial intelligence (AI), and more change is expected with the emergence of agentic AI-systems that can initiate, manage, and coordinate tasks. So far, most discussions about AI's impact on radiology follow 2 main approaches. The first, the "displacement" approach, tries to predict which jobs are most at risk of being replaced by AI. This narrative often warns that radiologists may be displaced. The second, the automation-versus-augmentation approach, looks within jobs to identify which tasks are likely to be fully automated (automation) and which will be improved by AI working alongside humans (augmentation). This paper introduces a third approach: reconfiguration. Instead of focusing on job loss or task replacement, the reconfiguration model looks at how AI changes the way tasks connect, how responsibilities shift, and how professional roles evolve. Drawing on recent research and developments in AI, this paper advances the reconfiguration approach and articulates why it offers a clearer way to understand-and help shape-the future of work in radiology. This paper offers a forward-looking reflection on the shifting nature of radiological work-clinically, educationally, and organizationally-as AI systems become increasingly integrated into practice.

To test the feasibility and quantify the performance of low-dose CT urography (CTU) with artificial intelligence iterative reconstruction (AIIR) for bladder cancer (BC) evaluation. A total of 122 patients undergoing CTU examination were prospectively enrolled, where the routine-dose scan (120&#x2009;kVp, ref 100&#x2009;mAs) at corticomedullary phase (CMP) was followed immediately by a low-dose scan (120&#x2009;kVp, ref 20&#x2009;mAs). Routine-dose images were reconstructed with hybrid iterative reconstruction (HIR, RD-HIR), while low-dose images were with AIIR (LD-AIIR) and HIR (LD-HIR). The image quality was first evaluated regarding streak artefacts around the bladder and then in contrast-to-noise ratio (CNR) for various manifestations of bladder wall. The diagnostic performance of BC was characterized using receiver operating characteristic (ROC) analysis, in respect to the clinical diagnostic report. The effective dose at low-dose CMP was 80.2% lower than routine-dose scan (7.6&#x2009;&#xb1;&#x2009;1.2 vs 1.5&#x2009;&#xb1;&#x2009;0.3&#x2009;mSv). Nineteen cases in LD-HIR were deemed clinically unacceptable for presenting severe artefacts around the bladder, while found well above the basic requirement in LD-AIIR. The highest CNR was found in LD-AIIR in all scenarios (all P&#x2009;<&#x2009;.001). The area under ROC curve in LD-AIIR was comparable to RD-HIR (0.988 vs 0.990, P&#x2009;=&#x2009;.172) and significantly higher than LD-HIR (0.988 vs 0.831, P&#x2009;<&#x2009;.001). The low-dose AIIR protocol allows for a profound dose reduction (80.2%) while maintaining reliable diagnosis of BC on corticomedullary phase CTU images. Corticomedullary phase CTU with AIIR permits 80.2% dose reduction while preserving reliable BC diagnosis.

Healthcare systems are now funding implementation of artificial intelligence (AI) algorithms in radiology, which will change the experience of care for patients. Currently, there is still limited evidence of patient attitudes to AI implementation in healthcare. We aimed to determine current attitudes to AI of people attending hospital for diagnostic imaging. This prospective study was conducted at a tertiary hospital network. Following ethical approval and informed consent, an 18-item questionnaire was administered to patients attending for outpatient imaging, assessing their views on AI. Factor analysis was undertaken to identify themes. In total, 162 people completed the questionnaire; 56% of whom were female (91/162). Most people thought that AI in healthcare would be useful (78%) and should be used (64%). People felt strongly that doctors should be responsible for decisions involving AI (71%). Three latent factors were identified: "utility and safety," "interaction," and "comparability to doctors." People were positive about the utility and safety of AI, were concerned about a loss of personal interaction, and compared AI unfavorably to doctors. There was strong opposition to autonomous AI decisions. The findings of this study map current patient acceptability of AI in healthcare and should inform strategies to balance AI ethical implementation with delivering value to patients and the healthcare system. This study provides insights into current patient attitudes to AI in healthcare in a UK setting where AI tools are actively being deployed, building on prior European surveys and guiding ongoing AI design and implementation.

Artificial intelligence (AI) has shown promise for estimating volumetric breast density values from processed, "for presentation," mammograms. However, previous evaluations have typically used small datasets or focused on a single vendor. In this study, we aimed to improve volumetric breast density estimation from processed mammograms for the three main UK vendors with a combination of improved training methods and the utilization of up-to-date data from the large OPTIMAM Mammography Image Database (OMI-DB). Paired processed/unprocessed mammograms were obtained from OMI-DB. Ground-truth, image-level density values were calculated by passing unprocessed images through a commercial density estimation tool. AI tools, comprising feed-forward convolution neural networks, were then trained to reproduce these values from the corresponding processed mammograms. Patient-level AI predictions for volumetric breast density demonstrated strong correlation with ground-truth values derived from unprocessed image counterparts (r&#x2009;=&#x2009;0.954-0.976). Models trained on less prevalent manufacturers performed worse (r&#x2009;=&#x2009;0.954 compared to 0.976 for the most prevalent manufacturer), highlighting the importance of collecting larger training datasets in future. Error levels were higher in patients with dense breasts. Model performance was generally consistent across screening sites but correlated with patient age, possibly due to the correlation of age and breast density. The presented models demonstrated good performance overall and were generally consistent across screening sites. The presented AI tools provide a means of estimating breast density from processed mammograms, enabling further research into breast cancer epidemiology and risk where only processed mammograms are available.