Faculty Bibliography

BACKGROUND:Previous studies evaluated large language model (LLM) performance on the American Society of Nuclear Cardiology (ASNC) Board Preparation Exam. Without domain-specific context, the best model (GPT-4o) achieved 63.1%, below the estimated 65% passing threshold and the 78% mean score of human fellows-in-training (FITs). Providing textbook context improved GPT-4o to 73.8% on text-only questions, but still fell short of human trainees. Whether next-generation LLMs with retrieval-augmented generation (RAG) can exceed this gap is unknown. METHODS:Claude Opus 4.7 and GPT-5.5 were administered all 168 questions (141 text-only, 27 image-based) from the 2023 ASNC Board Preparation Exam across 5 iterations each, using RAG with a nuclear cardiology textbook, companion atlas, and ASNC clinical guidelines. Claude used local FAISS-based semantic retrieval; GPT-5.5 used Azure's cloud-hosted vector store. Performance was compared to prior LLM results and 13 human FITs. RESULTS:Across 5 iterations, Claude Opus 4.7 achieved a mean accuracy of 86.3% ± 1.4% (text 88.8%, image 73.3%). GPT-5.5 achieved 86.7% ± 2.2% (text 88.5%, image 77.0%) but refused a mean of 12.2 questions (7.3%) per iteration due to safety filters. Both models surpassed the human FIT mean (78.0%) and the estimated passing threshold. Compared to GPT-4o without context (63.1%), this represents a 23-percentage-point improvement in 18 months. CONCLUSION/CONCLUSIONS:Next-generation LLMs with RAG now surpass average human trainee performance on nuclear cardiology board preparation questions, suggesting significant potential as educational tools and knowledge-reference aids in cardiovascular imaging.

PURPOSE/OBJECTIVE:U.S. ACGME-approved neuroradiology fellowship positions continue to increase in the setting of a national radiologist shortage, and the challenges faced by expanding fellowship programs could vary with program size. The aim of our project was to survey neuroradiology fellowship Program Directors (PDs) at US ACGME-approved diagnostic neuroradiology fellowship programs to better understand the challenges and practices in fellow recruitment, growth of fellowship class size, and other factors. MATERIALS AND METHODS/METHODS:An online 21-question survey was distributed to PDs of US ACGME-approved diagnostic neuroradiology fellowships (n=91). All variables were appropriately summarized and correlations assessed. Programs were dichotomized as large (greater than median number of approved spots) or small (less than or equal to median). Differences between large and small programs were evaluated with appropriate statistical tests. Effects associated with P<0.05 were considered statistically significant. RESULTS:PDs from 60/91(66%) institutions had evaluable responses. For the 2024-2025 academic year, 37/60(62%) programs filled all their fellowship positions. The proportion of large programs (20/25,80%) filling all their positions was higher (P=0.017) than small programs (17/35,49%). Large programs (14/25,56%) increased the number of fellowship positions more than small programs (5/35,14.3%) since 2019-2020 (P=0.002). Small programs had more difficulty filling all fellowship positions (19/34,54%) than large programs (5/25,20%) through national residency match program (P<0.001). Both large and small programs were challenged by other large programs and geographical location. Unique challenges for large and small programs were the high number of fellowship positions and competition from nearby programs, respectively. Length of PD tenure was not statistically correlated with the ability to fill the fellowship program (P=0.75). Small programs had a larger median of ABR-alternative pathway candidates [29%(0-66.7%)] than large programs [9.5%(0-16.7%)], the distribution did not differ significantly (P=0.19). CONCLUSIONS:Large neuroradiology fellowship programs filled all ACGME-approved positions with less difficulty and often increased their available fellowship positions in the last 5 years, compared to small programs. Competition from large programs and geographical location challenge recruitment for neuroradiology fellowship programs irrespective of size. Strategies for expanding and filling U.S. neuroradiology fellowships should vary based on fellowship size with a goal to help alleviate the on-going radiologist shortage.

INTRODUCTION/BACKGROUND:With expanded candidacy, cochlear implantation candidates increasingly present with residual acoustic hearing. The level and stability of residual hearing is particularly relevant when considering electro-acoustic stimulation (EAS). While adult data suggest relative stability, the natural history of residual hearing in children-who often have distinct and potentially progressive etiologies-remains poorly defined. METHODS:We conducted a retrospective cohort study of pediatric (<18 years) and adult patients who underwent unilateral cochlear implantation. Inclusion required contralateral (non-implanted) low-frequency pure-tone average (LF-PTA; 250-500 Hz) ≤85 dB HL and interaural difference ≤15 dB HL. One-year changes in LF-PTA, mid-frequency PTA (MF-PTA; 750-2000 Hz), and PTA (500-4000 Hz) were compared using Mann-Whitney U tests. The proportion with clinically meaningful decline (≥15 dB HL) was compared using Fisher's exact test. Longitudinal trajectories were assessed descriptively. RESULTS:Mean one-year LF-PTA, MF-PTA, and PTA changes between children and adults were not statistically different. However, clinically meaningful LF-PTA decline (≥15 dB) occurred more frequently in children (5/24; 20.83%) than adults (2/46; 4.34%) (p = 0.04). Longitudinal trajectories demonstrated substantial inter-individual variability without a consistent pattern of progressive loss. Etiology-specific trends showed relative stability in birth-related hearing loss, whereas ototoxic, genetic, and enlarged vestibular aqueduct etiologies demonstrated greater decline. CONCLUSIONS:Residual low-frequency hearing in children is generally stable but more variable, with a higher rate of clinically significant decline than in adults one year post-implant. These findings suggest underlying disease progression contributes to hearing changes and should inform counseling and EAS candidacy.

More than 100 prediction models have been developed to support the diagnosis, prognosis, or treatment of ADHD, yet none has reached routine clinical practice. In this Personal View, we argue that an important reason for this implementation gap is the absence of clearly defined prediction-based decision rules-formalised mappings from a model's output to specific clinical actions. Most published models report discrimination metrics such as the area under the receiver-operating-characteristic curve, but do not specify what a clinician should do differently for a patient classified as at high risk versus low risk. Without this link to action, even an accurate model remains clinically inert. We describe how prediction-based decision rules, combined with target trial emulation of their clinical utility in large observational datasets, offer a feasible path forward. We illustrate the approach with two worked ADHD examples (treatment intensity guided by predicted persistence and medication selection guided by predicted treatment response) and propose four priorities to shift the field from model development towards decision-oriented evaluation and implementation.

Adequate bowel preparation is critical to the effectiveness of colonoscopy. Bowel preparations have evolved over the years and now there are many options available to patients. These options fall into three categories of regular volume, low-volume and ultra-low volume preparations. It is important to select a preparation that maximises patient adherence and improves patient experience.

BACKGROUND:Evidence supporting platelet-rich plasma (PRP) injection for plantar plate injuries remains limited. PURPOSE/OBJECTIVE:To evaluate patient-reported outcomes, return to activity, complications, and failures at midterm follow-up in patients with plantar plate injuries treated with PRP injection. STUDY DESIGN/METHODS:Retrospective case series evaluating outcomes following PRP injection for plantar plate injuries. METHODS:Patients with plantar plate injuries were treated with ultrasound-guided PRP injection following failure of nonoperative management. Clinical outcomes evaluated were Foot and Ankle Outcome Score (FAOS) and Visual Analog Scale (VAS). Complications and failures were also reported. RESULTS:Twenty patients (23 feet) with a mean age of 51.0 ± 14.5 years and mean follow-up time of 58.0 ± 28.4 months were included. Mean FAOS scores improved 32.2 points (p < 0.001), with 20 of 23 (87%) feet achieving the minimal clinically important difference (MCID). Mean VAS scores improved 4.5 points (p < 0.001), with 19 of 23 (83%) feet achieving the MCID. Among 10 athletes, 7 (70%) returned to sport at a mean time of 3.1 ± 1.1 months. Complications occurred in 1 of 23 (4%) feet, and 3 of 23 (13%) feet required subsequent surgical intervention. CONCLUSIONS:At midterm follow-up, PRP injection for plantar plate injuries was associated with significant clinical improvement and a favorable safety profile, suggesting a potential role as an intermediary treatment option between conservative management and surgery. Female sex, smoking history, and higher-grade injuries were associated with inferior outcomes, underscoring the importance of individualized patient counseling and injury severity assessment. LEVEL OF CLINICAL EVIDENCE/METHODS:IV, retrospective case series.

Vasa previa is a rare but potentially catastrophic obstetric condition characterized by unprotected fetal blood vessels over or adjacent to the internal cervical os. Although advances in prenatal ultrasound have dramatically improved neonatal survival through planned cesarean delivery, many aspects of vasa previa remain poorly understood. Existing studies are limited by small sample sizes, single-center designs, heterogeneous diagnostic criteria, and inconsistent reporting of clinical outcomes, highlighting the need for large, systematically collected multicenter datasets. Here we describe, to our knowledge, the first United States multicenter registry of pregnancies complicated by vasa previa without concurrent placenta previa, the U.S. Vasa Previa Registry (US-VPR). This report outlines the rationale, design, and implementation of a large, multicenter, retrospective registry developed to characterize patient demographics, risk factors, placental pathology, natural history, antenatal management, maternal outcomes, neonatal outcomes, and healthcare utilization associated with vasa previa. The US-VPR includes contributions from 15 tertiary referral centers and represents the largest multicenter cohort of pregnancies complicated by vasa previa without concurrent placenta previa reported to date. By leveraging multicenter collaboration and standardized data collection, the registry provides a unique opportunity to characterize variation in clinical practice and outcomes across referral centers, strengthen the evidence base for this rare condition, and inform future prospective studies and evidence-based patient counseling.

Primary human myeloid cells hold promise for immunotherapies, yet efficient, scalable technologies for engineering and screening in these cells remain limited. Here we present a virus-like particle (VLP)-based toolkit that delivers diverse CRISPR editing modalities to human monocytes, macrophages and dendritic cells with high efficiency while preserving viability and innate immune responsiveness. VLP-mediated delivery of ribonucleoproteins supports gene knockout, base editing and epigenetic silencing. Combined with adeno-associated virus-mediated donor delivery, this approach enables site-specific integration of large DNA sequences by homology-directed repair. We developed SLICeVLP, which pairs sgRNA delivery by VPX-lentivirus with Cas9 protein delivery by engineered VLPs, and used it for pooled loss-of-function and Perturb-seq screens in human macrophages. We uncovered regulators of tumor necrosis factor (TNF) and CD80 expression, converging on TNFAIP3 as a central regulator of inflammatory polarization. TNFAIP3 ablation drove a proinflammatory state resistant to suppressive repolarization and enhanced cytotoxicity in chimeric antigen receptor macrophages. This system enables unbiased functional genomics in primary human myeloid cells, with implications for myeloid cell therapy design.

PURPOSE/OBJECTIVE:Neuro-oncology generates complex clinical, imaging, and molecular data, yet datasets remain relatively small and fragmented across modalities and institutions. While "big data" is traditionally defined by large sample size, neuro-oncology datasets are often characterized instead by high dimensionality. This study aims to provide an overview of the landscape of major U.S. neuro-oncology data resources and evaluate how these datasets are used in contemporary research. METHODS:A selection of neuro-oncology datasets was evaluated, including population registries, clinical data networks, federal and consortium research cohorts, institutional datasets, specialized resources, and artificial intelligence benchmarking resources. Analytical use was assessed through a large language model-assisted review of PubMed-indexed studies published over the past ten years referencing these datasets. Titles and abstracts were screened using a predefined classification schema, and structured data extraction identified study characteristics, analytical tasks, outcomes, modalities, validation strategies, and longitudinal modeling approaches. RESULTS:Of 11,651 screened publications, 3,608 met inclusion criteria. Analytical use was concentrated in a small number of datasets, particularly TCGA (~ 65%), SEER (~ 14%), and BraTS (~ 13%). Most studies modeled survival or tumor characteristics, whereas fewer than 1% examined functional or quality-of-life outcomes. Approximately 90% relied on a single dataset, and external validation and longitudinal modeling were rare. CONCLUSION/CONCLUSIONS:Big data in neuro-oncology is characterized by rich diversity. Expanding multimodal data capture, improving coverage of underrepresented populations and tumor types, strengthening longitudinal data collection, and enabling cross-dataset integration will be essential for translating high-dimensional datasets into clinically actionable insights.