Faculty Bibliography

BACKGROUND AND AIMS/UNASSIGNED:Pancreatic cystic lesions (PCLs) are increasingly detected because of the widespread use of imaging techniques. Among them, mucinous PCLs carry a higher malignancy risk, with intraductal papillary mucinous neoplasms (IPMNs) being the most frequent subtype. Accurate stratification based on the degree of dysplasia-low-grade dysplasia (LGD) versus high-grade dysplasia or carcinoma (HGD/C)-is essential to guide clinical management and avoid unnecessary surgical interventions. This study aimed to develop and evaluate a deep learning model for stratifying IPMNs into HGD/C and LGD using endoscopic ultrasound (EUS) images. METHODS/UNASSIGNED:This multicenter study included EUS images collected from 5 centers across Spain, Brazil, and the United States. Ground truth classification of IPMNs was established through cytologic and biochemical analysis of cyst fluid, EUS-guided through-the-needle biopsy, or surgical specimens. A deep learning model was trained to distinguish LGD from HGD/C. Model performance was assessed on the basis of sensitivity, specificity, accuracy, and area under the precision-recall curve. RESULTS/UNASSIGNED:A total of 51,046 EUS images were extracted from 30 examinations performed at 5 centers in Portugal, Spain, Brazil, and the United States. The model distinguished IPMNs with HGD/C from those with LGD with a sensitivity of 95.7%, a specificity of 88.7%, and an overall accuracy of 87.2%. The area under the receiver operating characteristic curve was 0.951. CONCLUSIONS/UNASSIGNED:To our knowledge, this is one of the first studies to evaluate the potential of an artificial intelligence model for dysplasia grading of IPMNs. Prospective validation of our model is necessary to ensure clinical benefit.

Pediatric autoimmune hemolytic anemia (AIHA) is a heterogeneous disease with significant morbidity due to the underlying condition and its treatment. Evidence-based guidelines for evaluation and management are lacking. Data from 399 patients with AIHA followed at 15 pediatric centers were collected to identify factors associated with secondary diagnoses, recurrent/chronic course, therapeutic efficacy, and mortality. Most had AIHA associated with secondary diagnoses including Evans syndrome (37%, 142/385), other autoimmunity (22%, 86/392), and inborn errors of immunity (IEI, 18%, 68/379). Of 305 patients tested, 82% had abnormal functional immune results. Genetic testing for an IEI was sent in 31% (109/348) with pathogenic findings identified in 32% of those tested. Patients with IEI or other autoimmunity more frequently had abnormal immunoglobulin and complement testing. Prevalence of IEI was not different between those presenting with or without infection. The median number of treatments for the first AIHA episode was 2 (range: 0-17). Of those with warm AIHA, 31% received steroid-sparing therapy during the first episode. Patients with recurrent AIHA (42%) had a higher rate of abnormal immune tests (OR=2.29, p=0.012), Evans syndrome (OR= 4.85; p<0.001), IEI (OR=3.88, p<0.001), and other autoimmune disorders (OR=3.29; p<0.001). With median follow up of 4.9 years (range: 0-19.4 years), 72/257 (28%) with warm AIHA continued to have active disease on treatment. Of the 399 patients, 10 died, all of whom had secondary diagnoses. Expansive immune evaluation, monitoring, and targeted treatments directed at immune diagnoses are needed for pediatric AIHA, highlighting the need for evidence-based pediatric AIHA guidelines.

Artificial intelligence (AI) has rapidly emerged as a transformative force in radiology, offering enhanced diagnostic accuracy, workflow optimization, and the potential to alleviate rising imaging demands. As radiology remains inherently dependent on pattern recognition and high-volume data interpretation, it represents an ideal domain for AI integration. This narrative review synthesizes current evidence on the clinical impact of AI across multiple dimensions of radiologic practice, including diagnostic performance, workflow efficiency, patient perspectives, and trainee education. AI systems have demonstrated performance approaching or exceeding that of radiologists in high-prevalence tasks, particularly in chest imaging and breast cancer screening, while also improving triage and reducing report turnaround times. However, these benefits are accompanied by significant challenges. Automation bias, over-reliance on algorithmic output, and anthropomorphic framing may compromise clinical judgment. Additionally, AI integration may paradoxically increase workload and contribute to radiologist burnout when poorly implemented. Patient-centered studies consistently indicate a preference for AI-augmented, rather than autonomous, diagnostic models, underscoring the enduring importance of physician oversight and communication. Among trainees, concerns regarding job security persist, though these are mitigated by increased AI literacy and structured educational initiatives. Ultimately, AI is best conceptualized not as a replacement for radiologists, but as a complementary tool. Thoughtful integration, combined with robust training, validation, and human oversight, will be essential to ensure that AI enhances rather than diminishes the quality of radiologic care.

OBJECTIVES/OBJECTIVE:This study evaluates how access to abortion training influenced OBGYN residents who selected residency programs before and after the Dobbs decision. STUDY DESIGN/METHODS:Current OBGYN residents completed a survey in January 2024 as part of their annual in-training exam. Residents rated the importance of abortion training and compared access to abortion training to other factors (e.g., residency culture and happiness, geographic location) when selecting and ranking programs. We computed the percentage of residents reporting abortion training as important and a top factor when ranking programs. We used chi-squared test to compare differences between cohorts. RESULTS:Of the 6,134 residents surveyed, 3,182 consented to participation and had complete responses (52%). Overall, 2,281 (72%) considered abortion training important when choosing programs and 2,179 (69%) when ranking programs; 1,105 (71%) of post-Dobbs and 1,074 (66%) of pre-Dobbs residents considered abortion training important when ranking programs (p<0.01). Access to abortion training was a top three factor for 741 (23%) respondents when ranking programs. More post-Dobbs (392, 25%) than pre-Dobbs residents (349, 21%) considered access to abortion training as a top factor when ranking programs (p=0.01). CONCLUSIONS:Applicants weigh a variety of factors when selecting programs and may match to abortion restricted environments - this includes the 23% who considered abortion training a top factor. Future work should continue to focus on the impact of the Dobbs decision on residency program selection and the ability of all residents to develop abortion-care skills. IMPLICATIONS/CONCLUSIONS:Abortion training is an important consideration in residency program selection and essential for patient care. Without prioritization of abortion education, programs risk losing viable candidates and graduating residents without essential skills.

BACKGROUND AND STUDY AIMS/UNASSIGNED:Cholangiocarcinoma (CCa) is a complex malignancy of the biliary tract, classified as intrahepatic, perihilar, or distal. Digital single-operator cholangioscopy (D-SOC) enhances evaluation of biliary strictures, although it remains limited by suboptimal biopsy yield and technical constraints. Artificial intelligence (AI), particularly convolutional neural networks (CNNs), has emerged as a promising adjunct. However, performance across anatomical subtypes is not well defined. This study evaluated diagnostic performance of an AI-based model in detecting CCa lesions by location. PATIENTS AND METHODS/UNASSIGNED:A YOLOv8-based CNN was trained and validated using 315,993 D-SOC images from 183 patients across six international high-volume centers. Images were labeled as benign or malignant based on expert consensus. Frame-based analysis assessed diagnostic performance using macro-average F1-score, precision, and recall. Subgroup analysis explored anatomical site-specific performance. RESULTS/UNASSIGNED:Among the included patients (mean age 66.1±12.1 years; 64.5% male), 43.2% had perihilar, 37.2% intrahepatic, and 19.7% distal biliary strictures. The model demonstrated high overall performance: F1 score 95.3%, precision 95.5%, and recall 95.1%. Site-specific analysis revealed an F1 score of 89.6% for distal strictures and 91.1% for perihilar strictures. Receiver operating characteristic curves showed areas under the curve of 0.980 for intrahepatic strictures and 0.990 for perihilar and distal strictures. CONCLUSIONS/UNASSIGNED:This is the first study to demonstrate AI-based diagnostic performance across CCa topography using a large, multicenter D-SOC dataset. Although anatomical complexity affects detection, the model's high precision and generalizability suggest potential for clinical utility. These results support application of an AI-enhanced algorithm decision algorithm for cholangioscopy.

Prostatic artery embolization (PAE) is a safe and effective minimally invasive treatment for lower urinary tract symptoms (LUTS) attributed to benign prostatic hyperplasia (BPH), including in patients with large prostates, who are poor surgical candidates or wish to avoid the morbidity of conventional surgery or transurethral interventions. As PAE becomes more widely adopted for BPH-related LUTS, objective measures of assessing treatment response are needed beyond symptom-based tools such as the International Prostate Symptom Score (IPSS) and Quality-of-Life (QoL) index. Imaging biomarkers complement these subjective endpoints by quantifying structural, functional, ischemic, and microstructural changes within the prostate before and after embolization. This review summarizes the principal imaging modalities used for biomarker assessment in PAE, with emphasis on ultrasound, magnetic resonance imaging (MRI), and computed tomography (CT) scans, and examines the current literature on imaging biomarkers of treatment response. Volumetric biomarkers, including whole prostate volume, central gland volume, and zonal volumetry index, provide quantitative assessment of gland reduction and may help predict clinical improvement, with zonal parameters showing stronger associations with symptomatic outcomes than whole-gland measurements alone. Functional biomarkers such as postvoid residual volume offer objective evaluation of bladder emptying alongside volumetric assessment, whereas ischemia-based biomarkers derived from contrast-enhanced MRI directly reflect embolization-induced devascularization. Diffusion-weighted imaging and apparent diffusion coefficient mapping provide additional information regarding post-embolization tissue remodeling, with b-value selection influencing early infarct detection. Emerging tools, including ultrasound elastography, radiomics, and artificial intelligence-assisted image analysis, may further refine patient selection, standardize response assessment, and improve predictive modeling, though prospective PAE-specific validation remains needed. Imaging biomarkers provide a more objective framework for evaluating PAE treatment response, with the potential to standardize outcome reporting, facilitate earlier identification of treatment success or failure, and guide biomarker-driven clinical decision-making. Prospective validation in larger PAE-specific cohorts remains essential before broader clinical adoption.

INTRODUCTION/UNASSIGNED:There is uncertainty about the benefits of cerclage in patients with transvaginal cervical length (TVCL) ≤10 mm and no prior spontaneous preterm birth. Our aim was to assess whether cervical cerclage in these patients was associated with prolonged pregnancy latency. METHODS/UNASSIGNED:This was an observational study of asymptomatic singleton pregnancies without a history of spontaneous preterm birth with extremely short cervix (TVCL ≤10 mm) in the second trimester. Exposure and outcome data were extracted manually from medical charts by obstetric providers, with all outcomes occurring prior to study initiation. All patients were prescribed vaginal progesterone and those with a cervical dilation >1.5 cm were excluded. The primary outcome was time interval from diagnosis to delivery (or 37 weeks' gestation, whichever occurred first). Secondary outcomes included gestational age at delivery, mode of delivery, and neonatal outcomes. We adjusted for parity and used log-binomial regression to estimate the relative risk for categorical variables and linear regression to estimate mean differences for continuous variables. RESULTS/UNASSIGNED:There were 247 patients with TVCL ≤10 mm during anatomy scan. After exclusions, 87 remained, of which 55 (63.2%) received cervical cerclage. At the time of diagnosis, the mean cervical length was 5.8 mm (cerclage group) versus 7.4mm (no cerclage group, difference: -1.6, 95% CI -0.6, -2.6) and the gestational age was 21.0 weeks (cerclage group) versus 22.1 weeks (no cerclage group, -1.1, 95% CI -1.8, -0.3). Mean pregnancy latency was longer in the cerclage group compared to the no cerclage group (13.4 vs 11.1 weeks, 2.2, 95% CI -0.3, 4.9), though there was a high level of uncertainty in the estimate. Term delivery occurred 30% more often in the cerclage group compared to the no cerclage group (RR 1.31, 95% CI 0.96, 1.79). CONCLUSION/UNASSIGNED:In our cohort, patients with extremely short cervix who received a cerclage had longer pregnancy latency than those who did not receive cerclage and more patients achieved term gestation, suggesting a potential benefit of cerclage in this population. However, given the high level of uncertainty of our estimates, additional research is needed to investigate these findings.

Although previously B cells had been underestimated in comparison to T cells in their role in autoimmunity, now, their impact is well established. Via secretion of autoantibodies, presentation of autoantigens, regulation of antigen processing and presentation, and release of inflammatory cytokines, B cells can mediate cytotoxicity and lead to organ damage. B cell depletion via CD20 targeting effectively eliminates B cells in the blood and primary lymph organs and has found an effective role in the treatment of both rheumatological and neurological diseases. The neonatal Fc receptor (FcRn) is a key component of immune regulation that prevents IgG (produced by B cells) from degradation by lysosomes, sending it back into the extracellular compartment, thereby extending its half-life. This abundance of pathogenic IgG can lead to the development of autoimmune disease. The interplay between these two mechanisms of autoimmunity provides the great potential for combination therapy to reduce existent pathogenic IgG as well as prevent the production of new autoantibodies, though further investigation is needed to determine the risks, particularly of infection. This paper will explore existing B cell depleting treatments and FcRn inhibitors, and consider the potential impact for autoimmune disease as well as for the treatment of atherosclerosis.