Citations

Abstract Glioblastoma (GBM), isocitrate dehydrogenase wild-type (IDHwt) and central nervous system diffuse large B-cell lymphoma (CNS-DLBCL) are aggressive brain tumors with overlapping MRI features, yet distinct treatment approaches. Noninvasive tools are needed to aid in differential diagnosis. Deep learning on T1 postcontrast and T2-weighted MRI sequences were used to differentiate GBM and CNS-DLBCL. A three-stage temporal study design was utilized. Model development was performed on 146 patients with CNS-DLBCL and 146 age-matched, sex-matched, and MRI year–matched patients with GBM diagnosed at Mayo Clinic between 1998 and 2019. Models were tested on independent temporal test cohorts. Initial testing included 240 independent GBM diagnosed at Mayo Clinic between 1998 and 2019. The prospective test cohort included 37 patients with CNS-DLBCL and 256 patients with GBM diagnosed at Mayo Clinic after January 1, 2020, and 36 patients with CNS-DLBCL diagnosed at an external institution. Of the patients diagnosed at Mayo Clinic, 47% had MRIs generated from non-Mayo institutions. Two different model approaches were compared: (i) ensemble approach using area under the receiver operating characteristic curve (AUC) and cross-validation for model selection and (ii) loss approach minimizing cross-entropy loss and cross-validation to evaluate prediction performance. The AUCs on the prospective test cohort were 0.84 [95% confidence interval (CI), 0.78–0.90] and 0.83 (95% CI, 0.77–0.88) for the ensemble and loss approaches, respectively. Stability of ensemble prediction improved with the increasing number of models. Stratified AUC analysis demonstrated consistent performance across sex and age. We utilized a robust temporal study design and applied 2 different analytic approaches to develop a classification model. The findings confirm the feasibility of using MRI-based deep learning models to differentiate GBM from CNS-DLBCL. Significance: GBM, IDHwt and CNS-DLBCL are aggressive brain tumors with overlapping MRI features, yet distinct treatment approaches. Noninvasive tools are needed to aid in differential diagnosis. We developed MRI-based deep learning models to differentiate GBM, IDHwt from CNS-DLBCL using a rigorous three-stage temporal design that included prospective validation. The model AUC on a prospective cohort was 0.84.

Abstract Background Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) have demonstrated cardiovascular benefits in type 2 diabetes mellitus (T2DM); however, their association with post-stroke outcomes following acute ischemic stroke (AIS) remains uncertain. Methods We conducted a retrospective propensity score-matched (PSM) cohort study using the TriNetX Global Collaborative Network. Adults with T2DM and AIS between January 2020 and January 2025 were included. Pre-stroke GLP-1 RA users were compared with non-users. Primary outcomes were all-cause mortality, intracerebral hemorrhage (ICH), ICD-10-coded stroke severity (NIHSS categories), and transient ischemic attack (TIA). Results After PSM, GLP-1 RA use was associated with significantly lower all-cause mortality at all time points (RR 0.44–0.52; HR 0.43–0.45; all p<0.001). ICH risk was reduced within 1–3 days (RR 0.683; 95% CI 0.496–0.940; p=0.019) and within 7 days (RR 0.575; 95% CI 0.516–0.641; p<0.001). A severity-dependent gradient was observed across NIHSS categories, with risk reductions ranging from 37% for minor strokes (NIHSS 0–5; RR 0.626) to 67% for very severe strokes (NIHSS >20; RR 0.327; all p<0.001). TIA risk was 29% lower (RR 0.712; 95% CI 0.668–0.759; p<0.001). E-value analysis demonstrated that an unmeasured confounder would need to be associated with both GLP-1 RA use and 30-day mortality by a risk ratio of at least 3.95 to fully explain the observed association, and by a risk ratio of at least 3.53 to shift the confidence interval to include the null. Conclusions In this large real-world cohort, pre-stroke GLP-1 RA use was associated with lower post-stroke mortality, reduced ICH, and a severity-dependent reduction in ICD-10-coded stroke severity among patients with T2DM. Residual confounding cannot be excluded, and these findings warrant confirmation in prospective randomized trials.

RELA-related disease is a rare, autosomal-dominant condition presenting with immune dysregulation and immunodeficiency. RELA encodes the p65 subunit, a component of the canonical NFκB pathway. The underlying pathogenesis, at least in part, involves impaired NFκB activation, increased susceptibility to TNFα-induced apoptosis, and hyperinflammation. To date, only a modest number of cases have been documented worldwide, and the full clinical and functional spectrum of disease remains incompletely defined. This study represents a large global cohort identified with RELA variants, comprising 40 patients from 30 institutions, expanding our previous report of a single large kindred (1). The clinical spectrum includes arthritis, mucocutaneous ulcers, uveitis, dermatitis, irritable bowel disease, recurrent infections, lymphopenia, hypogammaglobulinemia, recurrent fevers, and susceptibility to neoplasia. The patients represent 36 kindreds with 28 unique RELA variants, of which 15 are missense, 7 nonsense, and 6 frameshift variants. Fifteen are in the Rel homology domain, 7 in the transcriptional activation domain, and 6 in the C-terminal region proximal to TAD (Fig. 1). Functional studies assessed the canonical NFκB pathway in 27% of the cohort (n = 11; healthy controls = 40). Patient peripheral blood mononuclear cells (PBMCs) were used to quantify total and phosphorylated p65 (ph-p65, Ser529) by flow cytometry after stimulation with PMA (100 ng/ml) and ionomycin (1 uM). In 8 patients, we assessed the kinetics of IκBα degradation, required for p65 phosphorylation and nuclear translocation of p65/p50. All patients tested had decreased total p65, while 64% had decreased ph-p65 with altered phosphorylation kinetics. Nearly 45% had impaired IκBα degradation. One patient (P4) demonstrated increased IκBα degradation and ph-p65, while a related proband (P1) had decreased ph-p65. Additionally, P4 did not have the same mucocutaneous clinical phenotype as P1 despite sharing the same RELA variant. This could reflect the effect of other genetic modifiers, epigenetic, environmental, or stochastic variations, which could lead to variable expressivity and/or incomplete penetrance. All patients appeared to have haploinsufficiency, as none had normal levels of p65. This study expands the clinical and genetic spectrum of RELA-related disease and provides mechanistic insights into the impact of these variants on p65 phosphorylation and downstream function, resulting in variable expressivity. Figure 1. RelA protein structure. Represents unique variant locations and categories, key domains, and phosphorylation sites. Source: UniProt. Adapted from (2). BioRender.

Introduction Asymptomatic individuals with negative Interferon-γ release assay (IGRA) and tuberculin skin test (TST) results are often labeled as tuberculosis-unexposed individuals. We hypothesize that flow cytometric (FC) detection of antigen-stimulated T-cells in peripheral blood can differentiate IGRA(-)/TST(-) individuals with and without prior potential M. tuberculosis (Mtb) exposure. Methods We used FC multiparameter profiling of T-cell activation markers in TST/IGRA-negative individuals at risk for Mtb exposure and in unexposed controls. Results We studied 54 IGRA(-)/TST(-) subjects, including 27 individuals at risk for prior Mtb exposure. While quantitative IGRA results were negative in all patients, FC profiling of CD3/CD4 + IFN-γ + HLA-DR + T-cells after stimulation with a pool of Mtb-specific region of difference-1 (RD1) peptides, mycobacterial (PPD and MTB300), or Candida antigens, statistically differentiated between unexposed and potentially exposed individuals (P&lt;0.05). ROC analysis showed PPD and RD1 peptides responses had AUCs &gt;0.7 (sensitivity of 55.5–74% and specificity of 66.6–85.1%). The presence of a polyfunctional RD1-reactive CD3/CD8 + IFN-γ + TNF-α + T-cell subset also significantly differentiated these groups (AUC 0.66). Other T-cell markers demonstrated no significant differences. Conclusions FC detection of antigen-specific IFN-γ + HLA-DR + CD4+T-cells and IFN-γ + TNF-α + CD8 + T-cells can differentiate IGRA(-)/TST(-) individuals at risk of Mtb exposure from unexposed controls. Further validation is needed to determine whether this immunoprofiling approach can improve detection of Mtb infections and identification of unexposed individuals for future biomarker studies and vaccine trails.

ABSTRACT Purpose To validate a custom smartphone application for at‐home visual acuity (VA) measurement in children. Methods A total of 452 children aged 3–17.5 years participated. Certified examiners measured in‐office test–retest VA (logMAR) using gold‐standard Amblyopia Treatment Study HOTV (3‐to‐6‐year‐olds, younger cohort) or electronic Early Treatment of Diabetic Retinopathy Study (7‐to‐17.5‐year‐olds, older cohort) protocols at 3–4.5 m and app‐based VA at 1.5 m. Caregivers measured at‐home app‐based VA at 1.5 m. Results Comparing at‐home app‐based with gold‐standard VA, in eyes 20/40 or better, 95% (143/151) and 93% (91/98) of the younger and older cohorts were within 2 lines, respectively (mean differences: younger = −0.03, older = −0.04; 95% limits‐of‐agreement half‐width (LOA): younger = ±0.26, older = ±0.22). In eyes 20/50 or worse, 66% (42/64) and 75% (76/101) of the younger and older cohorts were within 2 lines, respectively (mean differences: younger = 0.11, older = 0.13, LOA: younger = ±0.50, older = ±0.51). Comparing in‐office app‐based VA with gold‐standard VA, in eyes 20/40 or better, 98% (160/164) and 94% (99/105) of the younger and older cohorts were within 2 lines, respectively (mean differences: younger = −0.03, older = −0.03; LOA: younger = ±0.22; older = ±0.24). In eyes 20/50 or worse, 85% (60/71) and 91% (101/111) of the younger and older cohorts were within 2 lines, respectively (mean differences: younger = 0.04; older = 0.04; LOA: younger = ±0.39; older = ±0.24). For gold‐standard test–retest, in eyes 20/40 or better, 99% (163/164) and 99% (104/105) of the younger and older cohorts had retest within 2 lines, respectively (mean differences: younger = 0.00; older = 0.01; LOA: younger = ±0.17; older = ±0.11). For 20/50 or worse, 92% (66/72) and 100% (111/111) in the younger and older cohorts were within 2 lines, respectively (mean differences: younger = 0.01; older = 0.02; LOA: younger = ±0.35; older = ±0.15). Conclusions Our app demonstrated good concordance with the gold standard at home and in the office for eyes with VA of 20/40 or better. However, concordance decreased considerably for eyes with VA 20/50 or worse, particularly at home.

Retrieval-augmented generation (RAG) is rapidly emerging as a transformative paradigm for large language models (LLMs), especially in high-stakes domains like oncology that demand precision, factual grounding, and up-to-date knowledge. By pairing LLMs with external knowledge repositories, RAG systems explicitly ground model outputs in relevant retrieved documents, helping to reduce hallucinations and ensure responses reflect current evidence. In oncology, where clinical knowledge evolves continually with new research and drug approvals, RAG offers a way to integrate the latest data (e.g., trial results, guidelines, genomic databases) into decision-making. This review synthesizes the technical foundations of RAG, including its architecture and key components, and examines current applications in oncology such as clinical decision support, patient education, radiology reporting, pathology analysis, and genomics-driven precision medicine. We highlight recent studies that demonstrate RAG’s potential—for instance, improving treatment recommendations by incorporating genetic profiles and literature, and enhancing diagnostic accuracy by integrating guidelines. We also discuss emerging developments like multimodal RAG (combining text with imaging or other data), ensemble model approaches, and new explainability tools that trace model outputs to sources. Finally, we critically analyze the limitations and challenges of deploying RAG in healthcare, including computational costs, retrieval errors, noise or conflicts in retrieved information, and ethical and regulatory considerations. While RAG-based systems show promise in augmenting oncologists’ expertise with timely knowledge, careful implementation, high-quality curation of knowledge bases, and human oversight will be crucial for safe and effective adoption in clinical practice.

INTRODUCTION: Nosocomial infections (NIs) in cirrhosis are associated with high mortality but could be preventable. Logistic regression (LR) models have failed to identify high-risk patients. We aimed to develop machine learning (ML) models to predict NI. METHODS: The CLEARED consortium consists of prospectively enrolled cirrhosis inpatients from >120 centers. Using day-of-admission clinical data, 3 ML approaches (random forest [RF], extreme gradient boosting, and neural networks [NNs]) were used to predict NI. Data were split 80:20 for training and testing stratified by the outcome. Models were compared using area under the receiver operating characteristic curve (AUC). RESULTS: In total, 8,263 patients (55.90 ± 13.34 years; 64.1% men) from 127 centers in 37 countries were included. NI developed in 869 (10.5%), a median of 6 (4–11) days of postadmission. Major NIs were respiratory (29.6%) and urinary tract infection (15.7%), and spontaneous bacterial peritonitis (13.5%). NIs occurred more frequently in patients from low/low-middle income countries and those with severe liver disease, alcohol etiology, and admission infections. NIs were associated with inpatient mortality (31.9% vs 8.1%, P < 0.001) and liver transplantation (4.8% vs 1.9%, P < 0.001). Although the RF model (AUC 0.69) showed good calibration (Brier score 0.09), outperforming extreme gradient boosting, neural network, and LR models (AUC 0.66 for all; LR comparison P = 0.043), no model achieved AUC ≥0.80 for clinical utility. At 10% predicted probability threshold, the RF model demonstrated only 75.4% sensitivity, 52.9% specificity, and 15.9% positive predictive value (PPV). DISCUSSION: NIs cannot be accurately predicted from day-of-admission data using ML models, even in a large, prospective, global cirrhosis cohort. Every hospitalized patient with cirrhosis should receive protocolized infection control measures.

ABSTRACT In 2018, the International Niemann‐Pick Disease Alliance (INPDA) and the International Niemann‐Pick Disease Registry (INPDR) developed and published comprehensive clinical management guidelines to support inclusive and standardized care pathways in Niemann‐Pick disease type C (NPC)—an ultra‐rare, autosomal recessive, neurovisceral lysosomal disorder. Since then, advances in diagnostics, care, and the approval of two novel disease‐modifying agents have underscored the need to revise these guidelines to ensure safe, consistent, and high‐quality care for those affected by NPC. In response, the INPDA and INPDR convened a multidisciplinary Guidelines Development Group (GDG) comprising individuals with NPC expertise from 14 countries across five continents, representing a broad range of specialties, as well as patients and families involved in NPC care. Informed by a comprehensive literature review and two meetings, the GDG systematically reviewed, revised, and updated the 2018 guideline statements, re‐evaluating the level of evidence, strength of recommendations, and expert agreement for each. The resulting 2025 consensus clinical management guidelines constitute a timely, up‐to‐date, and internationally applicable resource for the diagnosis, treatment, and holistic management of individuals with NPC. These guidelines serve as a critical resource for specialist centers, hospital‐based medical teams, staff involved in NPC patient care, family physicians and other primary caregivers, and, importantly, patients and their families.