Publications by Year: 2026

Artificial intelligence (AI) is increasingly being integrated into everyday tasks and work environments. However, its adoption in medical image analysis has progressed more slowly due to high clinical stakes, limited availability of labeled data, and substantial variability in imaging protocols and population. These challenges are further pronounced in the field of fetal, infant, and toddler (FIT) neuroimaging, where datasets are especially scarce and subject to large amounts of anatomical variability. However, deep learning (DL), a specific method within machine learning, which is itself a subfield of AI, has emerged as a powerful framework to adapt to the challenges of medical image analysis. This review is written for the broad FIT research community, including clinicians, neuroscientists, and develop mental scientists who may not have formal training in AI. To make the material accessible, we provide a concise overview of DL concepts before reviewing a selected, and non-exhaustive, list of applications of DL in FIT neuroimaging, including structural image analysis, enhancement of data acquisition, modeling of cognitive and perceptual processes, and automated video tagging. In closing, we discuss best practices for data curation, ongoing challenges, and opportunities for future research.

Deformable templates, or atlases, are images that represent a prototypical anatomy for a population, and are often enhanced with probabilistic anatomical label maps. They are commonly used in medical image analysis for population studies and computational anatomy tasks such as registration and segmentation. Because developing a template is a computationally expensive process, relatively few templates are available. As a result, analysis is often conducted with sub-optimal templates that are not truly representative of the study population, especially when there are large variations within this population. We propose a machine learning framework that uses convolutional registration neural networks to efficiently learn a function that outputs templates conditioned on subject-specific attributes, such as age and sex. We also leverage segmentations, when available, to produce anatomical segmentation maps for the resulting templates. The learned network can also be used to register subject images to the templates. We demonstrate our method on a compilation of 3D brain MRI datasets, and show that it can learn high-quality templates that are representative of populations. We find that annotated conditional templates enable better registration than their unlabeled unconditional counterparts, and outperform other templates construction methods.

BACKGROUND: Head and neck squamous-cell carcinoma (HNSCC) accounts for ∼5.3% of cancer-related mortality worldwide, with an estimated 890 000 new diagnoses and 450 000 deaths annually. Despite curative-intent therapy, 10% to 50% of patients experience recurrence. Prognosis for recurrent or metastatic disease is poor, with limited treatment options, underscoring the need for accurate prognostic models to guide treatment escalation or de-escalation and avoid over-treatment.

METHODS: We conducted a multicenter prognostic study of patients undergoing curative-intent surgery at Samsung Medical Center and Massachusetts Eye and Ear Infirmary/Massachusetts General Hospital from 2008 to 2024. Baseline clinicopathologic variables were integrated with longitudinal laboratory measurements from surveillance. A random 80/20 split defined development and internal-validation cohorts. Using XGBoost, we trained two models to predict recurrence-free survival (RFS) and overall survival (OS) at 1, 2, 3, 4, and 5 years from each visit.

RESULTS: A total of 975 patients with HNSCC (oral cavity, oropharyngeal, hypopharyngeal, and laryngeal subsites) were included. The areas under the curve (AUCs) for predicting 1-, 2-, 3-, 4-, and 5-year RFS from the surveillance time point were 0.785 (sensitivity, 72.8%; specificity, 71.5%), 0.831 (79.7%; 73.7%), 0.788 (74.0%; 73.3%), 0.769 (72.6%; 70.5%), and 0.795 (72.1%; 74.7%), respectively. For OS prediction, AUCs were 0.788 (72.1%; 73.6%), 0.797 (75.7%; 71.8%), 0.796 (81.0%; 68.4%), 0.820 (77.5%; 76.5%), and 0.815 (75.8%; 75.8%), respectively. In subgroup analysis, the model showed strong OS prediction in human papilloma virus (HPV)-positive oropharyngeal cancer, with AUCs of 0.943, 0.736, 0.699, 0.835, and 0.765 at 1-, 2-, 3-, 4-, and 5-years, respectively. In non-HPV-positive HNSCC, OS AUCs ranged from 0.780 to 0.813 and RFS AUCs from 0.774 to 0.830 across the same time points.

CONCLUSIONS AND RELEVANCE: In this multicenter study, an artificial intelligence (AI)-powered model using multimodal and longitudinal data accurately predicted RFS and OS at multiple time points following curative-intent surgery for HNSCC.

Magnetic resonance imaging (MRI) plays a crucial role in clinical diagnosis, yet traditional MR image acquisition often requires a prolonged duration, potentially causing patient discomfort and image artifacts. Faster and more accurate image reconstruction may alleviate patient discomfort during MRI examinations and enhance diagnostic accuracy and efficiency. In recent years, significant advancements in deep learning technology offer promise for improving MR image quality and accelerating acquisition. Addressing the demand for cardiac cine MRI reconstruction, we propose KGMgT, a novel MRI reconstruction network based on knowledge-guided approaches. The KGMgT model leverages adaptive spatiotemporal attention mechanisms to infer motion trajectories of adjacent cardiac frames, thereby better extracting complementary information. Additionally, we employ Transformer-driven dynamic feature aggregation to establish long-range dependencies, facilitating global information integration. Research findings demonstrate that the KGMgT model achieves state-of-the-art performance on multiple benchmark datasets, offering an efficient solution for cardiac cine MRI reconstruction. This collaborative approach, combining artificial intelligence technology to assist medical professionals in clinical decision-making, holds promise for significantly improving diagnostic efficiency, optimizing treatment plans, and enhancing the patient treatment experience. The code and trained models are available at https://github.com/MICV-Lab/KGMgT.

BACKGROUND: Dual anti-human epidermal growth factor receptor 2 (HER2) therapy plus chemotherapy followed by maintenance treatment with HER2-targeted and endocrine therapies is standard first-line treatment for hormone-receptor-positive, HER2-positive metastatic breast cancer. On the basis of preclinical and clinical data, the addition of palbociclib (a selective inhibitor of cyclin-dependent kinases 4 and 6) may overcome resistance to both endocrine and HER2-directed therapies.

METHODS: In this phase 3, open-label, randomized trial, we enrolled patients with hormone-receptor-positive, HER2-positive metastatic breast cancer who did not have disease progression after four to eight cycles of chemotherapy plus HER2-targeted therapy. Patients were randomly assigned in a 1:1 ratio to receive maintenance HER2-targeted and endocrine therapies with or without palbociclib. The primary end point was investigator-assessed progression-free survival. Secondary end points included the objective response, clinical benefit, safety, and overall survival.

RESULTS: A total of 518 patients underwent randomization: 261 were assigned to receive palbociclib and 257 to receive standard therapy. At a median follow-up of 53.5 months, patients in the palbociclib group had significantly longer progression-free survival than those in the standard-therapy group (median duration, 44.3 months vs. 29.1 months; hazard ratio for disease progression or death, 0.75; 95% confidence interval, 0.59 to 0.96; two-sided P = 0.02). Grade 3 and 4 adverse events, predominantly from neutropenia, occurred in 79.7% and 10.0% of the patients, respectively, in the palbociclib group, as compared with 30.6% and 3.6% of the patients, respectively, in the standard-therapy group.

CONCLUSIONS: The addition of palbociclib to maintenance anti-HER2 and endocrine therapies led to a significant improvement in progression-free survival over standard therapy, with increased toxic effects, mainly neutropenia. (Funded by Pfizer and others; PATINA ClinicalTrials.gov number, NCT02947685.).

BACKGROUND: Every year, approximately 20,000 youths lose a parent to firearm injury in the United States. Many more youths have parents who sustain nonfatal firearm injuries. The effect of parents' firearm injuries on their children's health and health care is poorly understood.

METHODS: Using U.S. commercial health insurance claims data from the 2007-2022 period, we identified youths, 1 to 19 years of age, whose parents had received treatment for firearm injury (exposure). Each youth with exposure was matched with up to five control youths on the basis of year, month, youth sex, metropolitan statistical area, state, insurance plan type, and prescription drug coverage; mean values of age and a risk score predicting future health care use (to provide a proxy for health status) were balanced. The primary outcome was a diagnosis of psychiatric disorder among youths, assessed as a rate, which was defined as the number of youths with at least one related insurance claim in a given month, divided by the total number of youths. Secondary outcomes included substance use disorder diagnosis, health care use, and medical spending. After matching, we estimated the difference in differences in outcomes between the exposure group and the control group 12 months before the parental injury through 12 months after the injury, using a least-squares regression model with adjustment for age and risk score.

RESULTS: We examined 3790 youths with exposure and 18,535 matched controls. The mean age of the youths was 10.7 years, and 51.5% were male. Parental firearm injury was associated with 8.4 additional psychiatric diagnoses (95% confidence interval [CI], 4.8 to 12.0) per 1000 youths and 23.1 additional mental health visits (95% CI, 8.2 to 38.1) per 1000 youths as compared with control, averaged over the year. This associated increase in the exposure group was largest for diagnoses of trauma-related disorders, including post-traumatic stress disorder, with an additional 8.5 diagnoses (95% CI, 6.0 to 10.9) per 1000 youths as compared with control, averaged over the year. No apparent changes relative to control were observed in rates of other diagnoses, medical encounters, procedures, and services or in medical spending.

CONCLUSIONS: Parents' firearm injuries were associated with increases in rates of psychiatric disorders and mental health visits among their children. (Funded by the National Institute for Health Care Management and the National Institute of Mental Health.).

PROBLEM: One often overlooked cost of the medical school application process is that associated with access to quality premedical advising, which enables applicants to hone their essays and interview skills. To help address this socioeconomic inequity, Giving a Boost (GAB) was created at the University of Pittsburgh School of Medicine to provide free near-peer application advising to medical school applicants.

APPROACH: Launched in 2020, GAB pairs applicants with medical school mentors who have recently completed the application process (based on shared backgrounds and goals), helping to demystify the admission process and boost applicant confidence. All participation in the program is voluntary; thus, it operates at no cost. Mentors review and edit primary and secondary essays, conduct mock interviews, and advise on letters of interest or intent. Additionally, some mentors facilitate specialized, supplemental workshops focusing on interviewing, letter writing, and reapplication strategies.

OUTCOMES: Over the 2020-2021 to 2022-2023 application cycles, GAB matched 231 applicants with 145 volunteer medical student mentors. In post-cycle surveys, the 67 responding applicants perceived GAB as more helpful (average helpfulness rating of 8.4/10) than other premedical advising resources (eg, friends [7.3/10] or paid consulting services [5.0/10]). The acceptance rate among survey respondents was 72.4% (21/29) in 2020-2021, 90.9% (20/22) in 2021-2022, and 100% (16/16) in 2022-2023-significantly higher than the reported national acceptance rates. Taken together, these findings demonstrate the strong performance of GAB applicants and the high perceived value of the GAB program.

NEXT STEPS: The next steps are to evaluate the efficacy of GAB chapters at other institutions across the country and to focus on recruiting more nontraditional applicants. Additionally, more data should be collected to control for potential confounding variables, such as Medical College Admission Test scores and grade point averages, and to evaluate the potential impact on underrepresented applicants.

Through a multidisciplinary quality improvement initiative, the Pre-Immunosuppression (Pre-IS) Clinic was created at a tertiary referral institution to ensure appropriate vaccination and infectious disease screening for patients on immunosuppressive medications. Consensus guidelines on immunisation and infectious disease screening for immunosuppressed patients were created through a multidisciplinary committee. The guidelines included three sections: (1) screening recommendations for chronic/latent infections prior to immunosuppression, (2) immunisation recommendations for immunosuppressed patients and (3) recommendations for household contacts of immunosuppressed patients. The workflow to the Pre-IS Clinic was optimised. We present the vaccination guidelines and workflow as an effective example of a multidisciplinary qualitive improvement initiative.

BACKGROUND: NPs (natriuretic peptides) are bioactive hormones crucial for regulating blood pressure, glucose homeostasis, and lipid metabolism. Despite the high heritability of circulating NP levels, the genetic determinants of NP regulation, particularly across ancestries and sexes, remain poorly understood. The objective of the current study was to identify genetic variants associated with NT-proBNP (N-terminal pro-B-type NP) levels in a multiancestry study population.

METHODS: Whole genome sequencing and array-based data from 81 213 individuals without heart failure were analyzed from the Trans-Omics for Precision Medicine cohorts, UK Biobank, All of Us Research Program, and REGARDS (Reasons for Geographic and Racial Differences in Stroke) study to identify common, rare, and structural variants associated with NT-proBNP levels. The main outcome of the study was rank-based inverse normal and standardized NT-proBNP levels. Genetic associations with NT-proBNP were examined, followed by gene prioritization, transcriptome-wide association studies, colocalization, and rare variant analyses.

RESULTS: Nine novel loci and 3 previously reported loci were identified to be associated with NT-proBNP levels. Novel structural variants were detected across 12 loci. Similar effect sizes were observed for both common and rare variants. Key genes such as BAG3 (10q26.11) and SLC39A8 (4q24) were identified through gene prioritization, with prior animal models supporting their therapeutic relevance. Rare variant analysis identified 6 masks with significant associations, specifically non-coding masks, suggesting regulatory modulation of NT-proBNP.

CONCLUSIONS: This study identifies novel common, rare, and structural variants associated with NT-proBNP levels, highlighting the contribution of both coding and regulatory non-coding variation. These findings advance our understanding of the genetic architecture of NT-proBNP and may inform future cardiometabolic therapeutic strategies.