Citations

Abstract Clinical narrative text contains crucial patient information, yet reliable extraction remains challenging due to linguistic variability, documentation habits, and differences across care settings. Large language models (LLMs) have shown strong accuracy on clinical information extraction (IE), but their reproducibility (stability under repeated runs) and robustness (stability under small, natural prompt variations) are less consistently quantified, despite being central to clinical deployment. In this study, we evaluate three open-weight LLMs representing distinct modeling choices: a dense general-purpose model (Llama 3.3), a mixture-of-experts (MoE) general-purpose model (Llama 4), and a domain-tuned medical model (MedGemma). We focus on binary clinical IE aligned with four mobility classes from the International Classification of Functioning, Disability and Health (ICF) framework. Using a controlled experimental design, we quantify (1) intra-prompt reproducibility across repeated sampling and (2) inter-prompt robustness across paraphrased prompts. We jointly report predictive performance (F1-score) and stability (Fleiss’ Kappa). And we test factor effects using three-way ANOVA with post-hoc comparisons. Results show that increasing temperature generally degrades agreement, but the magnitude depends on model and task; furthermore, prompt paraphrasing can substantially reduce stability, with particularly large drops for the MoE model. Finally, we evaluate a practical mitigation, self-consistency via majority voting, which improves κ substantially and often improves or preserves F1-score, at the cost of additional inference. Together, these findings provide a reproducible framework and concrete recommendations for evaluating and improving LLM reliability in clinical IE.

ABSTRACT Purpose To simultaneously measure skeletal muscle energetics and blood flow (BF) before, during, and after dynamic plantar flexion exercise (PFE). Methods Non‐localized pulse‐acquire phosphorus‐31 magnetic resonance spectroscopy ( 31 P MRS) and phase contrast flow magnetic resonance imaging ( 1 H MRI) using golden‐angle rotated spiral readouts were acquired in an interleaved fashion. Data were collected at rest, during PFE to volitional fatigue, and post‐exercise recovery in 10 healthy adults. Interleaved popliteal artery BF was validated against conventional cine phase‐contrast measures at rest and post‐exercise recovery, while calf muscle energetics were validated by comparing to conventional 31 P MRS‐only measures at rest. To minimize motion artifacts during dynamic PFE, flow data were resolved in 3 dimensions (cardiac phase, pedal position, and exercise stage). Results Resting phosphocreatine (PCr) and inorganic phosphate concentrations were similar between both techniques (both p > 0.05). During PFE, PCr declined to 46% ± 10% of rest. PCr recovery time was 43 ± 15 s. Resting BF was 132 ± 43 mL/min and increased up to four‐fold during PFE depending on exercise stage and pedal position. Notably, there was excellent agreement between the interleaved and conventional BF measures ( r 2 = 0.95, slope = 0.977, p < 0.0001) with minimal bias (12.62 mL/min) over a wide range of physiologically relevant BF values (from ∼60 to 600 mL/min). Conclusion The feasibility of simultaneous assessment of skeletal muscle energetics and BF during dynamic exercise by interleaved 31 P MRS/ 1 H flow MRI was demonstrated along with good agreement between interleaved and conventional techniques during rest and recovery. One can now simultaneously quantify critical BF‐metabolism relationships and use this to probe disease‐related adaptations.

ABSTRACT Objective To describe the practice patterns of point‐of‐care ultrasound (POCUS) and compare its usage across subgroups (adult vs. paediatric practitioners, physicians vs. nurse practitioners [NPs], high income [HI] vs. low and middle ‐income [LMI] countries). Methods Survey distributed via the Society of Critical Care Medicine in May–June 2024. Respondents reported the percentage of eligible patients in the past year for whom POCUS was used for vascular access and diagnostic purposes. Responses (0%, < 25%, 25%–50%, 50%–75%, > 75%) were converted into binary categories (> 50% use) and compared across subgroups using logistic regression. Results The survey was completed by 503 respondents from 39 countries (6% response rate): 77.7% from HI countries, 20.9% paediatric practitioners and 18.3% Nurse Practitioner (NPs)/Physician assistants (PAs). POCUS was commonly used for vascular access, and similar usage was found among NP/PAs, physicians, and adult and paediatric practitioners. Practitioners from LMI countries were less likely to use POCUS for vascular access. Basic diagnostic applications, such as assessing pleural effusion (72.5%) and left ventricular function (69.2%), were more commonly used than advanced applications like pulmonary artery pressure measurement (22%). Physicians were more likely than NPs (for some indications), and adult practitioners were more likely than paediatric practitioners (for most indications) to use POCUS for diagnostic purposes. Practitioners in LMI countries reported higher use of diagnostic applications than practitioners in HI countries. Conclusions POCUS is widely adopted in critical care but is less accepted for diagnostic use among paediatric practitioners and NP/PAs. Diagnostic utilisation is higher in LMI countries, while procedural use is lower. Impact Statement Point‐of‐care ultrasound (POCUS) is increasingly used in intensive care units to guide diagnosis and procedures. Despite its importance, little is known about how consistently it is used by different types of clinicians around the world. This international survey of critical care practitioners from 39 countries shows that POCUS is widely used, particularly for procedures, but diagnostic use remains uneven. Paediatric clinicians and advanced practice providers report lower diagnostic use, while clinicians in low‐ and middle‐income countries report higher diagnostic adoption. These findings highlight opportunities to improve access to training and support broader, more equitable use of POCUS in critical care.

Dynamic fluid and solute transportation across the arachnoid barrier in the meninges of healthy participants was measured by tracking intravenous gadolinium-based contrast agent entering cerebrospinal fluid from dural blood vessels using dynamic dual-spin-echo perfusion MRI.

Abstract Recent advances in long-read transcriptome sequencing enable high-throughput profiling of full-length RNA isoforms in bulk, single-cell, and single-nucleus samples. However, long-read datasets typically contain a mixture of complete and partial transcripts, leading to pervasive ambiguity in read-to-isoform assignment and complicating accurate isoform identification and quantification, particularly in the absence of reliable reference annotations. These challenges are further amplified in single-cell and single-nucleus samples, where coverage is sparse and transcriptional heterogeneity is high. Here, we present the Long Read Alignment Assembler (LRAA), a unified and versatile computational framework for isoform identification and quantification from long-read RNA sequencing data across bulk, single-cell, and single-nucleus transcriptomic samples. LRAA combines splice-graph based structural modeling with expectation maximization based optimization to probabilistically resolve ambiguous read assignments and improve isoform abundance estimation. The framework supports quantification-only, reference-guided, and fully reference-free (de novo) modes of analysis within a single methodological paradigm. We benchmarked LRAA using both simulated and genuine long-read datasets spanning sequencing standards and whole transcriptomes. Central to this evaluation is a novel benchmarking strategy based on Multiplexed Overexpression of Regulatory Factors (MORFs), which provides biologically expressed, barcoded isoforms with unambiguous read-level ground truth. Across all benchmarks, including MORFs, synthetic spike-ins, and whole-transcriptome datasets, LRAA consistently outperformed state-of-the-art methods in isoform identification accuracy, sensitivity, and expression quantification. Finally, we demonstrate the biological utility of LRAA by resolving cell-type-specific isoform usage across peripheral blood immune cell populations and by detecting a pathogenic cryptic isoform of STMN2 with associated transcriptional changes in single-nucleus RNA-seq data from frontal cortex tissue of an individual with frontotemporal dementia (FTD). Together, these results establish LRAA as a robust and general solution for resolving transcript diversity in complex biological systems, from development to disease.

GGGGCC (G 4 C 2 ) repeat expansion in C9ORF72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). Toxicity is thought to result from the accumulation of either repeat RNAs and/or dipeptide repeat proteins (DPRs) translated from repeat-containing transcripts through repeat-associated non-AUG (RAN) translation. To disentangle RNA from DPR toxicity, we mutated a CUG codon predominantly used to initiate DPR translation from all three reading frames. This mutation disrupted DPR synthesis while preserving the expression of repeat-containing RNAs. Despite the accumulation of RNA foci, behavioral deficits and pathological abnormalities, including p-TDP-43 inclusions, STING activation, motor neuron loss, neuroinflammation, and increased plasma neurofilament concentration, were alleviated in C9ORF72 mice. Base editing of the CUG codon also improved molecular phenotypes and survival in patient induced pluripotent stem cell–derived neurons, which highlights the potential of therapeutically targeting DPR production rather than repeat RNAs.

ABSTRACT Advanced chronic diseases, or multicomplexity in older adults presents unique challenges. Transthyretin cardiac amyloidosis (ATTR‐CA) is one such scenario where heart failure is a common presentation, and management remains challenging. We describe the challenges involved in the diagnosis of ATTR‐CA in older adults, which has implications on treatment options. We discuss the application of comprehensive geriatric assessment (CGA) and shared decision making (SDM) in the context of these challenges. Based on geriatric medicine principles, an innovative framework for applying SDM in ATTR‐CA patients is proposed, including consideration of the patient priorities care approach. The proposed framework emphasizes assessing functionality, frailty, and life expectancy to help categorize risk. This framework can be applied in various advanced chronic diseases or multicomplexity. Based on risk categorization, treatment burden, and alignment with values and preferences, management pathways are suggested for each risk category. Incorporating CGA and SDM, the proposed framework supports patient‐centered care, ensuring that clinical recommendations are tailored to each older adult's unique needs and goals. In ATTR‐CA, a collaboration between cardiology and geriatric medicine provides significant value in managing older adults. The need to prevent age‐related bias in clinical decision‐making exists across all health conditions, and the proposed framework allows for a thorough evaluation of multimorbidity, frailty, disability, and patient preferences. While ATTR‐CA is used as a prototype, this integrated approach can be applied across all health conditions and is essential for delivering holistic care, improving communication, and aligning treatment plans with patient values.

Abstract Purpose Acetabular overcoverage (pincer morphology) has been suggested to be both a source of pain and dysfunction as well as potentially protective against osteoarthritic change. The purpose of this study was to compare minimum 5-year clinical outcomes of patients undergoing primary hip arthroscopic labral repair with a lateral center edge angle (LCEA) of ≥ 40° as compared to non-dysplastic controls with an LCEA of 25-40°. Methods We identified all patients that underwent primary hip arthroscopy with labral repair between 2009-2019 at a single academic institution. Patients with LCEA ≥ 40° (pincer) were propensity matched to non-dysplastic controls with LCEA 25-40° on a 1:2 basis by sex, age, body mass index (BMI), and year of surgery. Patient-reported outcome measures (PROMs) including Visual Analog Scale (VAS) for pain, Tegner activity score, modified Harris Hip Score (mHHS), International Hip Outcome Tool (iHOT-12), and Hip Outcome Score (HOS) were analyzed, as were reoperations and conversions to total hip arthroplasty (THA). Results Fifty-five patients with pincer morphology (mean preoperative LCEA: 41.9°, Tonnis angle: -7.3°, alpha angle: 58.8°) were matched to 110 non-dysplastic controls (LCEA: 31.2°, Tonnis angle: 2.2°, alpha angle: 58.9°), resulting in a total of 165 hips in 163 patients (67% females, mean age: 37.7 years, BMI: 26.8 kg/m2). Pincer patients were corrected to a mean postoperative LCEA of 38.7° (p&lt;0.0001), with 22 pincer patients having residual overcoverage, namely a postoperative LCEA ≥ 40° (range: 40.0-45.8°). At mean 7.8 years follow-up (range: 5.1-13.2), there was no difference in any postoperative PROMs between pincers and controls (p&gt;0.15) and no difference in PROMs between patients with residual overcoverage and pincer patients with a postoperatively normalized LCEA (p&gt;0.25). At final follow-up, 4 patients in the pincer cohort (7%) and 21 controls (19%) had converted to THA (p=0.046). Conclusion Patients with lateral acetabular overcoverage demonstrated similar and satisfactory postoperative PROMs at a minimum of 5-year follow-up compared to propensity-matched non-dysplastic controls, regardless of if the LCEA was corrected to &lt;40°. Of note, non-dysplastic controls demonstrated an increased rate of conversion to THA, supporting the potentially chondroprotective role of acetabular overcoverage.