53417 research outputs found
Sort by
Matrix structure and microenvironment dynamics correlate with chemotherapy response in ovarian cancer
No full textMacronutrient excess in critical illness: too much of a good thing?
No full textPURPOSE OF REVIEW: Critical care nutrition has traditionally focused on preventing underfeeding, but recent evidence highlights the detrimental effects of early macronutrient excess exceeding metabolic capacity. Here we review recent advances elucidating mechanisms of macronutrient toxicity, clinical and biochemical markers of intolerance, and guideline-recommended strategies that tailor nutrient delivery to the patient's phase of illness. RECENT FINDINGS: Recent randomized controlled trials have suggested lack of benefit and even potential harm from early supplemental macronutrient supplementation in the ICU. Excess provision of energy, protein, carbohydrates, or lipids may exacerbate mitochondrial dysfunction, hyperglycaemia, hepatic steatosis, and immune impairment. These insights emphasize the need for developing biomarkers to guide precision nutrition in critically ill adults, aiming to optimize recovery and minimize metabolic complications. SUMMARY: Current evidence supports a phased, tolerance-based approach that avoids overfeeding during acute catabolism while supporting increased demands during recovery
Computational hermeneutics: evaluating generative AI as a cultural technology
Get PDFGenerative AI (GenAI) systems are increasingly recognized as cultural technologies, yet current evaluation frameworks often treat culture as a variable to be measured rather than fundamental to the system's operation. Drawing on hermeneutic theory from the humanities, we argue that GenAI systems function as "context machines" that must inherently address three interpretive challenges: situatedness (meaning only emerges in context), plurality (multiple valid interpretations coexist), and ambiguity (interpretations naturally conflict). We present computational hermeneutics as an emerging framework offering an interpretive account of what GenAI systems do, and how they might do it better. We offer three principles for hermeneutic evaluation—that benchmarks should be iterative, not one-off; include people, not just machines; and measure cultural context, not just model output. This perspective offers a nascent paradigm for designing and evaluating contemporary AI systems: shifting from standardized questions about accuracy to contextual ones about meaning.</jats:p
Quality of Cancer-Related Clinical Coding in Primary Care in North Central London: Mixed Methods Quality Improvement Project.
No full textBACKGROUND: The North Central London (NCL) Cancer Alliance carried out a quality improvement (QI) project to fill a distinct knowledge gap regarding the quality of clinical coded data in a primary care electronic health care record system across the whole cancer pathway. OBJECTIVE: This study aims to establish the quality of cancer-related clinical coding in NCL primary care, encompassing both quantitative measures (eg, coding completeness and diversity) and qualitative dimensions such as clinical relevance and workflow alignment. METHODS: This was a mixed methods QI project in which we combined an observational dataset review and qualitative data from stakeholder interviews, workshops, and discussions. In the dataset review, we evaluated completeness, diversity, validation, and granularity in cancer clinical coding along the patient cancer pathway, which was split into three domains: (1) patient characteristics and risk factors, (2) cancer screening attendance, and (3) living with cancer. It was conducted in NCL primary care electronic health record systems, covering a population of over 1.4 million adults across 5 boroughs. RESULTS: Cancer-related clinical coding in NCL primary care revealed significant gaps despite high completeness for ethnicity (912,679/1,055,083, 86.5%) and language (898,023/1,307,601, 68.7%). Employment status (29,848/1,229,644, 2.4%) and family history of cancer (183,424/1,236,580, 14.8%) were underrecorded, with wide variation in coding practices. Screening data showed good alignment with national datasets for cervical and bowel screening but fragmented and inconsistent breast screening data due to a lack of standardized codes. Cancer diagnosis coding was incomplete (4604/5260, 87.5% recorded), and treatment and staging data were almost entirely absent, limiting proactive management of long-term consequences. Stakeholder input highlighted inconsistent template use, limited data updates, and insufficient incentives as key barriers to better coding. CONCLUSIONS: The QI project has provided a detailed insight into the many dimensions of cancer coding and sheds light on many factors that underpin variation and coding preference. We offer a number of recommendations. The prioritized ones include the need for a cancer clinical coding data framework for primary care supported by appropriate funding and incentivization; improvements in the breast screening pathway and its interface with primary care; improvements in the quality of secondary care information that is sent to primary care; and dissemination of the importance of coding of cancer activity in primary care
Genome-wide analysis of cardiac ventricular phenotypes reveals novel loci and therapeutic targets for heart failure
Get PDFAbstract Left and right ventricular imaging measures are essential for heart failure diagnosis and prognostication, yet their genetic architecture remains underexplored. We conduct genome-wide association analyses of twenty left and right cardiovascular magnetic resonance phenotypes in 56,509 UK Biobank participants, including conventional measurements (e.g., volumes/ejection fraction) and novel parameters (left ventricular global function index and myocardial contraction fraction). We identify 200 loci associated with at least one phenotype ( P < 5×10 -8 ); 58 being novel. A polygenic risk score for left ventricular global function index negative associates with heart failure in phenome-wide scan. Rare variant analysis reveals enrichment of deleterious variants across 13 genes ( P < 2.5×10 -6 ). Colocalisation with heart failure implicates 23 shared loci and bioinformatic analysis prioritises genes including HSPB7, CAMK2D, ALDH2, ENG , and YWHAE . Druggability analysis highlights PDE3A , informing divergent effects of non-selective PDE3 inhibition. In this work, we expand our knowledge of cardiac ventricular genetics, suggesting potential heart failure therapeutic targets. </jats:p
Abstract PS3-09-14: Caught in the Middle: Discrepancies Between Immunohistochemistry and Oncotype DX in Hormone Receptor Assessment Leading to Triple Negative Reclassification—Tumour Heterogeneity or Just Low ER Expression?
No full textMaximum persistent Betti numbers of Čech complexes
Get PDFAbstract This note proves that only a linear number of holes in a Čech complex of n points in R d can persist over an interval of constant length. Specifically, for any fixed dimension p<d p < d and fixed \varepsilon >0 ε > 0 , the number of p -dimensional holes in the Čech complex at radius 1 that persist to radius 1 + ε is bounded above by a constant times n , where n is the number of points. The proof uses a packing argument supported by relating the Čech complexes with corresponding snap complexes over the cells in a partition of space. The argument is self-contained and elementary, relying on geometric and combinatorial constructions rather than on the existing theory of sparse approximations or interleavings. The bound also applies to Alpha complexes and Vietoris–Rips complexes. While our result can be inferred from prior work on sparse filtrations, to our knowledge, no explicit statement or direct proof of this bound appears in the literature. </jats:p
PointExplainer: Towards transparent Parkinson’s disease diagnosis
No full textDeep neural networks have shown potential in analyzing digitized hand-drawn signals for early diagnosis of Parkinson’s disease. However, the lack of clear interpretability in existing diagnostic methods presents a challenge to clinical trust. In this paper, we propose PointExplainer, an explainable diagnostic framework to identify hand-drawn regions that drive model diagnosis, helping clinicians understand the model’s diagnostic logic. Specifically, PointExplainer assigns discrete attribution values to hand-drawn segments, explicitly quantifying their relative contributions to the model’s decision. Its key components include: (i) a diagnosis module, which fuses hand-drawn signals into 3D point clouds to represent hand-drawn trajectories, and (ii) an explanation module, which trains an interpretable surrogate model to approximate the local behavior of the black-box diagnostic model. We also introduce consistency measures to further address the issue of faithfulness in explanations. Extensive experiments on two benchmark datasets and a newly constructed dataset show that PointExplainer can provide intuitive explanations with no diagnostic performance degradation. The source code is available at https://github.com/xc-lab/PointExplainer
