Diagnostic accuracy of vision-language models on Japanese diagnostic radiology, nuclear medicine, and interventional radiology specialty board examinations.

PURPOSE: The performance of vision-language models (VLMs) with image interpretation capabilities, such as GPT-4 omni (GPT-4o), GPT-4 vision (GPT-4V), and Claude-3, has not been compared and remains unexplored in specialized radiological fields, including nuclear medicine and interventional radiology. This study aimed to evaluate and compare the diagnostic accuracy of various VLMs, including GPT-4 + GPT-4V, GPT-4o, Claude-3 Sonnet, and Claude-3 Opus, using Japanese diagnostic radiology, nuclear medicine, and interventional radiology (JDR, JNM, and JIR, respectively) board certification tests. MATERIALS AND METHODS: In total, 383 questions from the JDR test (358 images), 300 from the JNM test (92 images), and 322 from the JIR test (96 images) from 2019 to 2023 were consecutively collected. The accuracy rates of the GPT-4 + GPT-4V, GPT-4o, Claude-3 Sonnet, and Claude-3 Opus were calculated for all questions or questions with images. The accuracy rates of the VLMs were compared using McNemar's test. RESULTS: GPT-4o demonstrated the highest accuracy rates across all evaluations with the JDR (all questions, 49%; questions with images, 48%), JNM (all questions, 64%; questions with images, 59%), and JIR tests (all questions, 43%; questions with images, 34%), followed by Claude-3 Opus with the JDR (all questions, 40%; questions with images, 38%), JNM (all questions, 42%; questions with images, 43%), and JIR tests (all questions, 40%; questions with images, 30%). For all questions, McNemar's test showed that GPT-4o significantly outperformed the other VLMs (all P < 0.007), except for Claude-3 Opus in the JIR test. For questions with images, GPT-4o outperformed the other VLMs in the JDR and JNM tests (all P < 0.001), except Claude-3 Opus in the JNM test. CONCLUSION: The GPT-4o had the highest success rates for questions with images and all questions from the JDR, JNM, and JIR board certification tests.

HAP-FAST: a feasibility study incorporating qualitative, mechanistic and costing sub-studies alongside a randomised pilot trial comparing chest x-ray to low-dose CT scan and empirical antibiotics to antibiotics guided by the BIOFIRE® FILM ARRAY® pneumonia plus panel in adults with suspected non-ventilator-associated hospital-cquired pneumonia.

INTRODUCTION: Non-ventilator-associated hospital-acquired pneumonia (nv-HAP) is the most common healthcare-associated infection (HCAI), is associated with high mortality and morbidity and places a major burden on healthcare systems. Diagnosis currently relies on chest x-rays to confirm pneumonia and sputum cultures to determine the microbiological cause. This approach leads to over-diagnosis of pneumonia, rarely identifies a causative pathogen and perpetuates unnecessary and imprecise antibiotic use. The HAP-FAST study aims to evaluate the feasibility of a randomised trial to evaluate the clinical impact of low-dose, non-contrast-enhanced thoracic CT scans and rapid molecular sputum analysis using the BIOFIRE® FILMARRAY® pneumonia plus panel (FAPP) for patients suspected with nv-HAP. METHODS AND ANALYSIS: The HAP-FAST feasibility study consists of a pilot randomised trial, a qualitative study, a costing analysis and exploratory analyses of clinical samples to investigate the immune-pathophysiology of HAP. Participants are identified and recruited from four acute hospitals in the Northwest of the UK. Using a Research Without Prior Consent model, the pilot trial will recruit 220 adult participants, with or without mental capacity, and with suspected HAP. HAP-FAST is a non-blinded, sequential, multiple assignment, randomised trial with two possible stages of randomisation: first, chest x-ray (CXR) or CT; second, if treated as nv-HAP, FAPP or standard microbiological processing alone (no FAPP). Pathogen-specific antibiotic guidance will be provided for FAPP results. Randomisation uses a web-based platform and followed up for 90 days. The feasibility of a future trial will be determined by assessing trial processes, outcome measures and patient and staff experiences. ETHICS AND DISSEMINATION: This study has undergone combined review by the UK NHS Research Ethics Committee and Health Research Authority. Results will be disseminated via peer-reviewed journals, via the funders' website and through a range of media to engage the public. TRIAL REGISTRATION NUMBER: NCT05483309.

Increasing magic number and other trends in diagnostic radiology NRMP match data.

The magic number, or number of ranks needed to achieve a greater than 90 % chance of matching, has not been investigated for diagnostic radiology (DR). Somewhat reflective of a field's changing competitiveness, this individual metric can be useful for reassuring applicants or identifying a need to reach out to mentors. The NRMP's Charting Outcomes in the Match was accessed over the previous 10 cycles to assess changes to magic number and other match-related metrics. Over the last 10 cycles, there has been an increase in magic number for prospective radiologists. Based on the most 2022 recent report, the magic number was 14 compared to 5 and 2 in 2014 and 2016 respectively. Compared to the average US MD senior, those applying into DR were significantly more likely to match in 2014, 2016 and 2020 (p < 0.01 for all), and significantly less likely to match in 2018 and 2022 (p = 0.03 and p < 0.01, respectively). This trend has had important consequences for applicants and programs as the incentive to apply more widely grows. The increasing magic number demonstrates increasing competitiveness in the field, which might be due to a positive job market, changing medical student preferences, or increased access to radiology electives and mentors. The 2024 Charting Outcomes document will be the first to include data from a class almost entirely affected by the change to a pass/fail Step1 and the new preference signaling supplement. It is currently unclear how either change will affect the overall competitiveness of the field and the magic number.

Cavitating pulmonary lesion secondary to mycobacterium szulgai infection-case report of a rare pathogenic organism.

Mycobacterium szulgai (MS) is a species of non-tuberculous mycobacterium (NTM), which very rarely is identified as the causative pathogen of pulmonary infections. Due to its rarity, there are limitations in the existing literature regarding the diagnosis, investigation and treatment of MS pulmonary infection. Our case report provides further information regarding the clinical, microbiological and radiological findings associated with MS pulmonary infection with suggestions provided on its long term management.

Long-Term Survival and Factors Associated with Increased Mortality in Patients with Ocular Adnexal Lymphomas.

Orbital and ocular adnexal lymphoma (OAL) affects the orbit and the surrounding structures and can arise as several subtypes with variable prognoses. We performed an observational study on the relationship between OAL subtype, diagnostic features, and prognosis to offer valuable insights into imaging techniques, such as Positron Emission Tomography with 2-deoxy-2-[fluorine-18] fluoro-D-glucose integrated with Computed Tomography (18F-FDG PET-CT), in predicting outcomes. With this aim, we retrospectively reviewed 99 patients with OALs, recording demographics, cancer subtype, location and treatment, 18FDG avidity, and bone marrow positivity. We divided patients into Group 1 (those presenting with extranodal marginal zone lymphoma-EMZL) and Group 2, including all other subtypes. The primary outcome was long-term cancer-specific survival (CSS) based on key predictors, performed through Kaplan-Meier curves and the log-rank test, with a p < 0.05 significance threshold. The mean patient age was 67 years (57-75.5). The most frequent histopathologic subtypes were EMZL lymphoma in 69 patients (69.7%), small lymphocytic lymphoma (11.1%) and diffuse-large B-cell lymphoma (10.1%). Patients of Group 1 showed a better prognosis (CSS = 80%) compared to those of Group 2 (CSS = 60%) (p = 0.01). In patients with high-grade lymphoma, the occurrence of 18FDG avidity (p = 0.003) and bone marrow positivity (p = 0.005) were related to a worse prognosis. In our group, EMZL was the most prominent subtype of OALs and exhibited the best prognosis, low 18FDG avidity, and bone marrow negativity. By observing specific patterns in radiological findings, it is possible to increase our understanding of disease progression, treatment response, and the overall prognosis in OAL patients.

A comparative evaluation of polymeric materials as tissue equivalent phantoms in diagnostic radiology.

In this study tissue equivalency of the polymeric materials was investigated by comparing with ICRP 110 Male Adult Computational Phantom tissues. For this purpose, radiological properties of polyamide (PA), high density polyethylene (HDPE), ultra-high molecular weight polyethylene (UHMWPE), polypropylene (PP), polyvinyl chloride (PVC), polytetrafluoroethylene (PTFE), polyethylene terephthalate (PET), polyoxymethylene (POM) and polyurethane foam (PU FOAM) were evaluated in the diagnostic energy range (15-150 keV). The radiological properties of the materials and ICRP 110 Male and Female Adult Computational Phantom tissues were calculated with Phy-X/PSD software. No major differences were seen except for sex-specific organs, and comparisons were made using an adult male phantom. To confirm the results experimentally, a chest phantom was designed with the polymeric materials. The phantom was scanned by Siemens SOMATOM Edge CT device with tube voltage of 120 kVp and Hounsfield Unit (HU) values were measured. In addition, HU values were calculated using theoretical relationships and significant agreement was obtained between measured and calculated HUs. It was determined that PA, PP, UHMWPE and HDPE were equivalent to muscle and adipose tissue, PVC and PTFE were equivalent to mineral bone, PET and POM were equivalent to spongiosa bone and PU FOAM was equivalent to lung tissue.

Integrating Text and Image Analysis: Exploring GPT-4V's Capabilities in Advanced Radiological Applications Across Subspecialties.

This study demonstrates that GPT-4V outperforms GPT-4 across radiology subspecialties in analyzing 207 cases with 1312 images from the Radiological Society of North America Case Collection.

GPT-4 Turbo with Vision fails to outperform text-only GPT-4 Turbo in the Japan Diagnostic Radiology Board Examination.

PURPOSE: To assess the performance of GPT-4 Turbo with Vision (GPT-4TV), OpenAI's latest multimodal large language model, by comparing its ability to process both text and image inputs with that of the text-only GPT-4 Turbo (GPT-4 T) in the context of the Japan Diagnostic Radiology Board Examination (JDRBE). MATERIALS AND METHODS: The dataset comprised questions from JDRBE 2021 and 2023. A total of six board-certified diagnostic radiologists discussed the questions and provided ground-truth answers by consulting relevant literature as necessary. The following questions were excluded: those lacking associated images, those with no unanimous agreement on answers, and those including images rejected by the OpenAI application programming interface. The inputs for GPT-4TV included both text and images, whereas those for GPT-4 T were entirely text. Both models were deployed on the dataset, and their performance was compared using McNemar's exact test. The radiological credibility of the responses was assessed by two diagnostic radiologists through the assignment of legitimacy scores on a five-point Likert scale. These scores were subsequently used to compare model performance using Wilcoxon's signed-rank test. RESULTS: The dataset comprised 139 questions. GPT-4TV correctly answered 62 questions (45%), whereas GPT-4 T correctly answered 57 questions (41%). A statistical analysis found no significant performance difference between the two models (P = 0.44). The GPT-4TV responses received significantly lower legitimacy scores from both radiologists than the GPT-4 T responses. CONCLUSION: No significant enhancement in accuracy was observed when using GPT-4TV with image input compared with that of using text-only GPT-4 T for JDRBE questions.

Abnormal Uterine Bleeding: A Pictorial Review on Differential Diagnosis and Not-So-Common Cases of Interventional Radiology Management.

Abnormal uterine bleeding (AUB) poses a multifaceted challenge in women's health, necessitating an integrated approach that addresses its diverse etiologies and clinical presentations. The International Federation of Gynecology and Obstetrics PALM-COEIN classification system provides a systematic approach to the diagnosis of AUB in non-pregnant women, based on clinical and imaging-based categorization of causes into structural (Polyps, Adenomyosis, Leiomyomas and Malignancy; PALM), and non-structural causes (Coagulopathies, Ovulatory disorders, primary Endometrial disorders, Iatrogenic and Not otherwise classified; COEIN). On the other hand, placental disorders, uterine rupture, ectopic pregnancy and retained products of conceptions are the main causes of uterine bleeding during pregnancy and in the peripartum period. Ultrasound is usually the first-line imaging technique for the differential diagnosis of causes of AUB. Computed Tomography may be useful if ultrasound findings are unclear, especially in emergency settings. Magnetic resonance imaging, when indicated, is an excellent second-line diagnostic tool for a better non-invasive characterization of the underlying cause of AUB. This pictorial review aims to illustrate the main causes of AUB from the point of view of diagnostic imaging and to show not-so-common cases that can be treated by means of interventional radiology.

Pulmonary artery involvement in Behçet's disease: A challenging case and comprehensive management approach.

A 26-year-old male with Behçet's disease (BD) presented with recurrent oral and genital ulcers, bilateral pneumonia, and a left lower pulmonary artery aneurysm. Endovascular coil embolization was initially performed, followed by treatment with prednisone, colchicine, and azathioprine. Despite treatment, disease progression occurred, requiring additional embolization, intravenous pulse methylprednisolone, and cyclophosphamide. Ultimately, a combination of medical and endovascular interventions resolved the pulmonary arterial aneurysms. This case highlights BD's systemic inflammatory nature and vascular complications like pulmonary artery aneurysms. It emphasizes the importance of early detection and individualized, multidisciplinary care for such complications.

Workflow interruptions in an era of instant messaging: A detailed analysis.

INTRODUCTION: The complex practice environment and responsibilities incumbent on diagnostic radiologists creates a workflow susceptible to disruption. While interruptions have been shown to contribute to medical errors in the healthcare delivery environment, the exact impact on highly subspecialized services such as diagnostic radiology is less certain. One potential source of workflow disruption is the use of a departmental instant messaging system (Webex), to facilitate communications between radiology faculty, residents, fellows, and technologists. A retrospective review was conducted to quantify the frequency of interruption experienced by our neuroradiology fellows. MATERIALS AND METHODS: Data logs were gathered comprising all instant messages sent and received within the designated group chats from July 5-December 31, 2021, during weekday shifts staffed by neuroradiology fellows. Interruptions per shift were calculated based on month, week, and day of the week. RESULTS: 14,424 messages were sent across 289 total shifts. The 6 fellows assigned to the main neuroradiology reading room sent 3258 messages and received 10,260 messages from technologists and other staff. There was an average of 50 interruptions per shift when examined by month (range 48-53), and 52 interruptions per shift when examined by day of the week (range 40-60). CONCLUSION: Neuroradiology fellows experience frequent interruptions from the departmental instant messaging system. These disruptions, when considered in conjunction with other non-interpretative tasks, may have negative implications for workflow efficiency, requiring iterative process improvements when incorporating new technology into the practice environment of diagnostic radiology.

The impact of large language models on radiology: a guide for radiologists on the latest innovations in AI.

The advent of Deep Learning (DL) has significantly propelled the field of diagnostic radiology forward by enhancing image analysis and interpretation. The introduction of the Transformer architecture, followed by the development of Large Language Models (LLMs), has further revolutionized this domain. LLMs now possess the potential to automate and refine the radiology workflow, extending from report generation to assistance in diagnostics and patient care. The integration of multimodal technology with LLMs could potentially leapfrog these applications to unprecedented levels.However, LLMs come with unresolved challenges such as information hallucinations and biases, which can affect clinical reliability. Despite these issues, the legislative and guideline frameworks have yet to catch up with technological advancements. Radiologists must acquire a thorough understanding of these technologies to leverage LLMs' potential to the fullest while maintaining medical safety and ethics. This review aims to aid in that endeavor.

Quantitative assessments of retinal macular structure among rural-dwelling older adults in China: a population-based, cross-sectional, optical coherence tomography study.

OBJECTIVES: To quantitatively assess and compare retinal macular structures of rural-dwelling older adults in China using two different optical coherence tomography (OCT) scanners and to examine their associations with demographic, lifestyle, clinical and ocular factors. DESIGN, SETTING AND PARTICIPANTS: This population-based, cross-sectional study included 971 participants (age ≥60 years) derived from the Multimodal Interventions to Delay Dementia and Disability in Rural China study. We collected data on demographics, lifestyle factors, clinical conditions (eg, cardiovascular disease (CVD)) and ocular factors (eg, visual acuity and spherical equivalent). We used two models of spectral-domain OCT to measure macular parameters in nine Early Treatment Diabetic Retinopathy Study subfields. Data were analysed using the multiple general linear models. RESULTS: Spectralis OCT demonstrated higher macular thickness but a lower macular volume than Primus 200 OCT (p<0.05). Nasal quadrant of the inner and outer subfields was the thickest, followed by superior quadrant. Adjusting for multiple potential confounding variables, older age was significantly correlated with lower average inner and outer macular thicknesses and overall macular volume. Men had higher macular parameters than women. The presence of CVD was correlated with lower central macular thickness (ß=-6.83; 95% CI: -13.08 to -0.58; p=0.032). Middle school or above was associated with higher average inner macular thickness (ß=7.85; 95% CI: 1.14 to 14.55; p=0.022) and higher spherical equivalent was correlated with lower average inner macular thickness (ß=-1.78; 95% CI: -3.50 to -0.07; p=0.042). CONCLUSIONS: Macular thickness and volume assessed by Spectralis and Primus 200 OCT scanners differ. Older age and female sex are associated with lower macular thickness and volume. Macular parameters are associated with education, CVD and spherical equivalent. TRIAL REGISTERATION NUMBER: MIND-China study (ChiCTR1800017758).

Effects of low-dose ionizing radiation on the molecular pathways linking neurogenesis and autism spectrum disorders in zebrafish embryos.

Prenatal exposure to environmental factors may play an important role in the aetiopathogenesis of autism spectrum disorder (ASD). We aim to investigate the potential effects of low-dose x-rays from dental diagnostic x-rays on neurodevelopment and molecular mechanisms associated with ASD in developing zebrafish embryos. Zebrafish embryos were divided into four groups and exposed using a dental x-ray unit: control, 0.08, 0.15 and 0.30 seconds, which are exemplary exposure settings for periapical imaging. These exposure times were measured as 7.17, 23.17 and 63.83 mSv using optical stimulated luminescence dosimeters. At the end of 72 hours post-fertilization, locomotor activity, oxidant-antioxidant status, and acetylcholine esterase (AChE) activity were analyzed. Expression of genes related to apoptosis (bax, bcl2a, p53), neurogenesis (α1-tubulin, syn2a, neurog1, elavl3) and ASD (eif4eb, adsl2a, shank3) was determined by RT-PCR. Even at reduced doses, developmental toxicity was observed in three groups as evidenced by pericardial edema, yolk sac edema and scoliosis. Deleterious effects of dental x-rays on neurogenesis through impaired locomotor activity, oxidative stress, apoptosis and alterations in genes associated with neurogenesis and ASD progression were more pronounced in the 0.30s exposure group. Based on these results we suggest that the associations between ASD and low-dose ionizing radiation need a closer look.

A rare duo: Pulmonary arterial sling with a pig bronchus.

Pulmonary artery sling is an incomplete vascular ring, where the left pulmonary artery originates from the right pulmonary artery, leading to airway constriction. A tracheal bronchus is an anatomical variation in which an extra bronchus originates from the trachea, frequently resulting in respiratory symptoms or complications. We report a 6-week-old female patient with a pulmonary artery sling coursing around the distal trachea and a concurrent tracheal bronchus.

AI-based X-ray fracture analysis of the distal radius: accuracy between representative classification, detection and segmentation deep learning models for clinical practice.

OBJECTIVES: To aid in selecting the optimal artificial intelligence (AI) solution for clinical application, we directly compared performances of selected representative custom-trained or commercial classification, detection and segmentation models for fracture detection on musculoskeletal radiographs of the distal radius by aligning their outputs. DESIGN AND SETTING: This single-centre retrospective study was conducted on a random subset of emergency department radiographs from 2008 to 2018 of the distal radius in Germany. MATERIALS AND METHODS: An image set was created to be compatible with training and testing classification and segmentation models by annotating examinations for fractures and overlaying fracture masks, if applicable. Representative classification and segmentation models were trained on 80% of the data. After output binarisation, their derived fracture detection performances as well as that of a standard commercially available solution were compared on the remaining X-rays (20%) using mainly accuracy and area under the receiver operating characteristic (AUROC). RESULTS: A total of 2856 examinations with 712 (24.9%) fractures were included in the analysis. Accuracies reached up to 0.97 for the classification model, 0.94 for the segmentation model and 0.95 for BoneView. Cohen's kappa was at least 0.80 in pairwise comparisons, while Fleiss' kappa was 0.83 for all models. Fracture predictions were visualised with all three methods at different levels of detail, ranking from downsampled image region for classification over bounding box for detection to single pixel-level delineation for segmentation. CONCLUSIONS: All three investigated approaches reached high performances for detection of distal radius fractures with simple preprocessing and postprocessing protocols on the custom-trained models. Despite their underlying structural differences, selection of one's fracture analysis AI tool in the frame of this study reduces to the desired flavour of automation: automated classification, AI-assisted manual fracture reading or minimised false negatives.

Rising use of diagnostic imaging in Australia: An analysis of Medicare-funded radiology services between 2000 and 2021.

INTRODUCTION: The use of diagnostic imaging services is increasing worldwide. This has important impacts on healthcare resource allocation and potential risks to the population. This study aimed to quantify trends in medical imaging in Australia over the past two decades. METHODS: Data were extracted from the Australian Medicare Benefits Schedule (MBS) between 2000 and 2021. Simple linear regression analyses were performed to assess changes in absolute utilisation and utilisation rate per 100,000 population of total imaging services as well as by each imaging modality. Logistic regression analysis was performed to assess changes in total imaging services as a proportion of total Medicare services over time. Chi-squared test was used to assess for change in modality composition of imaging services. RESULTS: There were 436,255,500 imaging studies performed between 2000 and 2021. The absolute utilisation of total imaging services increased annually by an average of 864,404 (95% CI: 808,235-920,573, p < 0.001). For each consecutive year, the proportion of total Medicare services attributed to total imaging services increased by 0.01% (95% CI: 0.01-0.01, p < 0.01). There was also a statistically significant increase in the utilisation rates of imaging services per 100,000 population for each imaging modality. The number of imaging services per radiologist increased on average by 74 (95% CI: 26-122, p < 0.05) annually. CONCLUSION: The utilisation of diagnostic imaging services has increased in Australia between 2000 and 2021, outpacing the population growth, total healthcare services, and the radiologist workforce.

Danish Prostate Cancer Consortium Study 1 (DPCC-1) protocol: Multicentre prospective validation of the urine-based three-microRNA biomarker model uCaP.

INTRODUCTION: The primary objective of the Danish Prostate Cancer Consortium Study 1 (DPCC-1) is to provide validation for a novel urine-based microRNA biomarker, called uCaP, for a diagnosis of prostate cancer. METHODS AND ANALYSIS: Eligible participants are biopsy naïve men aged ≥18 years with prostate-specific antigen (PSA) levels ≥3 ng/mL, who are referred to prostate MRI due to suspicion of PC at one of the following three major urology/uroradiology centers: Aarhus University Hospital, Herlev & Gentofte University Hospital, or Odense University Hospital, where MRI and targeted biopsy are implemented in clinical use. Exclusion criteria include previous diagnosis of urogenital cancer, contraindication to MRI, gender reassignment treatment or PSA level >20 ng/mL. The participants will be asked to donate a urine sample in connection with their MRI. The study is observational, uses a diagnostic accuracy testing setup and will integrate into the current diagnostic pathway.We will measure the levels of the three microRNAs in the uCaP model (miR-222-3 p, miR-24-3 p and miR-30c-5p) in extracellular vesicle-enriched cell-free urine samples, to assess if uCaP can improve specificity and retain sensitivity for International Society of Urological Pathology Grade Group ≥2 PC, when used as a reflex test to PSA ≥3 ng/mL. We hypothesise that uCaP can improve selection for prostate MRI and reduce the number of unnecessary scans and biopsies. ETHICS AND DISSEMINATION: This study is approved by the Central Denmark Region Committee on Health Research Ethics (reference number: 1-10-72-85-22). All participants will provide written informed consent. Study results will be published in peer-reviewed journals and presented in scientific meetings. TRIAL REGISTRATION NUMBER: NCT05767307 at clinicaltrials.gov.