213476 research outputs found
Sort by
SWIN-DS: A DEEPLY SUPERVISED TRANSFORMER WITH GEOMETRIC GUIDANCE FOR ROBUST LACUNE DETECTION
No full textAutomated detection of lacunar infarction—a key marker of cerebral small vessel disease—is challenging due to their small size and morphological similarity to mimics. To enhance detection accuracy while minimizing false positives, we propose Swin-DS, a multitask framework that simultaneously performs segmentation and distance map regression. Our approach incorporates deep supervision across decoder resolutions to enhance multi-scale feature learning, with task losses dynamically balanced through homoscedastic uncertainty weighting. On the Public VALDO challenge dataset, SwinDS achieves competitive performance, with an F1-score of 0.513 and the lowest reported false positives per subject (FPs/subject) of 3.37 among published results. The framework’s effective balance between sensitivity and specificity results in improved robustness and clinical relevance, addressing a key limitation of existing methods
Unlocking the Therapeutic Potential of Integrin-Linked Kinase Inhibitors in Bioengineered 3D Breast Tumor Stroma Models
No full textThe tumor microenvironment (TME) plays a pivotal role in breast cancer progression and metastasis, and the efficacy of targeted therapies is influenced by the heterogeneous nature of the TME. Interactions between breast cancer cells and their surrounding stromal cells modulate proliferation, invasion, and survival pathways, often via integrin-mediated mechanotransduction and growth factor signaling. Integrin-linked kinase (ILK) is a serine/threonine protein kinase that has been widely established as a critical driver of breast cancer progression, metastasis, and therapeutic resistance. Its expression is frequently upregulated in breast cancer tumors and correlates with poor prognosis. Given that ILK activity is highly dependent on cell-matrix interactions that are only recapitulated in 3D culture, we investigated the effect of an ILK inhibitor in 3D bioengineered compartmentalized breast tumoroid models to better mimic in vivo conditions. Two tumor cell masses (MDA-MB-231 or MCF-7) were cultured within a primary breast tissue stromal compartment representative of breast tissue or a metastatic representative of lung tissue. In highly invasive and highly hypoxic MDA-MB-231 3D tumoroid models, ILKI treatment was 2.2 fold more effective in 3D models representative of breast tissue (p-value < 0.0001) compared to those with the metastatic lung compartment (p-value = 0.03). However, ILKI treatment was slightly more effective (1.4 fold) in the less invasive and less hypoxic MCF-7 3D tumoroid models with the metastatic lung compartment compared to those with the primary breast compartment. Non-invasive imaging of oxygen gradients in the 3D models shows alleviation of hypoxia following treatment and correlation with enhanced treatment efficacy. These results emphasize the necessity of modeling both the tumor and the stroma since this interaction can directly influence drug efficacy. Moreover, ILK inhibitor treatment holds promise for breast cancer therapy particularly in chemotherapeutic resistant cases
Treatment modifiers of interpersonal functioning in psychotherapy for people with borderline personality disorder: Systematic review with meta-analyses of individual participant data
No full textBackground:
Borderline personality disorder (BPD) is often accompanied by interpersonal dysfunction. Psychotherapy can improve interpersonal functioning, but individual characteristics may moderate outcomes. This systematic review used individual participant data meta-analysis (IPD-MA) to examine such moderators.
Method:
A literature search up to 26 November 2025 across 10 databases (including PubMed, Medline, Embase, PsychINFO, CINAHL, Web of Science, and Cochrane CENTRAL) identified randomised clinical trials (RCTs) investigating the effects of psychotherapy on interpersonal functioning in individuals with BPD compared to treatment as usual (TAU) or clinical management control interventions (CM). Authors of included trials were contacted to retrieve IPD. IPD-MAs employed a one-stage random-effects approach to estimate treatment effects on interpersonal functioning and potential moderators in bivariate linear mixed-effects models. The study was registered with PROSPERO (CRD42021210688).
Results:
Out of 23,735 identified records, 32 RCTs (2762 participants) met inclusion criteria. Individual participant data (IPD) were available for 17 trials (1431 participants). All trials were rated as having either high risk of bias or some concerns. Missing data were common, with 321 out of 1431 participants (23%) lost to follow-up. Meta-analyses of both aggregate data and IPD yielded comparable effect estimates, though statistical significance differed (IPD-MA: β = −0.21, CI: −0.45 to −0.02, SE = 0.12, p = .0778; 17 trials, 1071 participants). In unadjusted analyses, the presence of co-occurring anxiety disorder(s) (β = −0.40, 95% CI: −0.73 to −0.08) and a higher number of co-occurring disorders (β = −0.08, 95% CI: −0.15 to −0.01) were associated with larger treatment effects (not significant after alpha correction).
Conclusion:
Psychotherapy appears to be effective for individuals with BPD. Although moderator effects did not remain statistically significant after alpha correction, unadjusted analyses suggested larger treatment effects in individuals with co-occurring anxiety and greater clinical complexity. Importantly, these findings indicate that such comorbidities may not be a contraindication for psychotherapy for BPD
Agnostic Biomolecular Binding Affinity Prediction via Frame Averaging Graph Transformer
No full textPredicting binding affinity between biomolecules is
a critical task in drug discovery, where deep learning methods have achieved significant progress. However, many existing approaches employ modality-specific network architectures, limiting their direct applicability across diverse biomolecular interaction types. In this work, we propose F3Affinity, a novel structure-based graph transformer that is agnostic to biomolecular interaction type in its architectural design. Leveraging a frame
averaging technique, our model flexibly learns SE(3)-invariant representations of input structures. We further demonstrate that F3Affinity performs competitively on multiple binding affinity prediction tasks, including protein–ligand, protein–protein, and protein–nucleic acid interactions, without requiring any pre-trained embeddings. These results validate its broad applicability
and effectiveness across multiple biomolecular interaction types
Methodological review of the level of statistical support declared in radiological research articles
No full textOBJECTIVES: We assessed if there was disparity between qualified statisticians and other researchers regarding the level of statistical assistance deemed necessary to support radiological research. METHODS: We categorised 50 consecutive, eligible original research articles published in an indexed imaging journal (European Radiology) 2024, according to authors' statements regarding statistical support, declared in the "statistics and biometry" section. Two reviewers extracted data related to study design, statistical methods, and analysis. Two medical statisticians categorised each study as presenting "complex" statistical methods or not and then compared this with authors' own assessment of statistical complexity, stated in the published article. We performed descriptive analyses. RESULTS: Most studies were observational (49, 98%) and retrospective (38, 76%). 35 (70%) studies were diagnostic, 7 (14%) prognostic, and 6 (12%) mixed. Malignancy was the most frequent topic (29 studies, 58%), and MRI the most frequent modality (35 studies, 70%). We deemed most studies (33, 66%) presented complex statistical methods. Of these, 13 studies (26% overall) declared that "no complex statistical methods were necessary for this paper". However, 10 of these employed hypothesis testing, frequently using multiple methods; 9 employed agreement and/or reliability analyses; all presented accuracy measures; 11 (85%) presented a regression model. CONCLUSION: We found that approximately one quarter of original research articles published in our sample stated that "no complex statistical methods were necessary", but then presented complex analyses. ADVANCES IN KNOWLEDGE: Some radiological researchers may underestimate the complexities of statistical analysis and requirement for specialist statistical support, which risks inappropriate analyses and misleading results
Brain aging in bipolar disorder using a neuroimaging and machine learning-derived metric: Findings from the ENIGMA BD Working Group
No full textBACKGROUND: Bipolar disorder (BD) is associated with clinical and biological markers of premature aging. In this largest study of brain age in BD to date, with 2919 participants, we compared brain-predicted age difference (brain-PAD) in individuals with BD and healthy comparison (HC) participants. Brain-PAD is a machine learning-estimated metric that quantifies the difference between an individual's predicted brain age and their chronological age, a potential clinical bio-signature of premature brain aging. Within individuals with BD, we also examined how medication and clinical characteristics were related to brain-PAD. METHODS: Age was predicted from 77 MRI measures of regional subcortical and lateral ventricle volumes, cortical thickness, and surface area for 1342 BD and 1577 HC adult participants, aged 18-75 yrs. old (μ = 37.2; SD = 12.3), from the curated ENIGMA Bipolar Disorder working group (ENIGMA-BD) and leveraging an ENIGMA machine learning model previously trained and validated using independent samples. Chronological age was subtracted from predicted age to produce an individual-level estimate known as brain-PAD. Linear mixed models (adjusting for sex and age as fixed effects and site as a random effect) were used to examine group differences and clinical associations. RESULTS: BD was associated with higher brain-PAD, compared to HC, primarily among older patients, as demonstrated by a significant age by diagnosis interaction (+0.05 [SE: 0.02] years). Individuals with BD on antiepileptic (AED) medications only (+3.20 [SE: 0.78] years) or on both AED and second-generation antipsychotics (SGA) (+3.74 [SE: 0.89] years) demonstrated greater brain-PAD compared to individuals who were not on any of the examined medications. Those taking lithium, whether alone or with AED and SGA independently, showed no difference in brain-PAD compared to individuals not taking any of the examined medications. However, individuals who were taking lithium showed lower brain-PAD compared to those on AED (-4.48 [SE: 0.84] years) or AED and SGA (-5.01 [SE:0.92] years). Individuals with a BD I subtype diagnosis had a higher brain-PAD (+1.50 [SE:0.55] years) compared to those with BDII or subtypes that are not otherwise specified (NOS). CONCLUSIONS: Results from this study suggest compounding effects of BD diagnosis and older age on brain-PAD, an ML-derived summary metric of structural alterations. Within BD, brain-PAD was differentially related to medication use, consistent with prior findings from ENIGMA-BD. Notably, AED use was generally related to more advanced brain age. Lithium use, alone or in combination with other medications, was not associated with advanced brain age, suggesting a possible neuroprotective effect of lithium. Brain-PAD as an ML-derived summary metric of structural alterations of the brain may provide clinical utility in assessing long-term holistic brain health to monitor the effectiveness of lifestyle modifications or treatments over time. LIMITATIONS: The cross-sectional nature of the study design and the limited granularity of the clinical data limit interpretation. Longitudinal studies with detailed chronicity data, medications and clinical measures overtime will improve brain-PAD modeling in BD
Enhancing Primary and Early Years Teaching: Close-to-Practice Research Insights
No full textThe Helen Hamlyn Centre for Pedagogy (0 to 11 Years) (HHCP) and Primary Postgraduate Certificate in Education (PGCE) team at the UCL Institute of Education (IOE) co-hosted the conference “Enhancing Primary and Early Years Teaching: Researching ‘Close to Practice’” on 11th June 2025. Teachers from partnership schools and colleagues from IOE came together in-person to explore research aimed at strengthening professional knowledge and improving teaching.
The conference focused on fostering collaboration between classroom practitioners and university-based educators and researchers. Through keynote talks and a range of presentations participants discussed research and teaching.
All respondents who completed the evaluation rated the conference as excellent, and qualitative feedback was overwhelmingly positive. Comments included: “It has been so, so good — I’m taking so much back with me,” and “So informative, inspiring and interesting.”
What follows in this report summarises insights across five themes: teacher development, children’s agency, curriculum and assessment, wellbeing, and social justice
Reflections on ‘Abundance’: UK Housing, Planning Regulation and Speculative Production
No full textThe Abundance agenda has drawn attention to the problems of regulatory systems prioritizing processes over outcomes and potentially failing to support the development of essential infrastructure, including new housing. Whilst this problem has been seen in the UK, with some development projects struggling to navigate the politics of local planning, this commentary argues that the major impediments to building the housing that the UK actually needs are, firstly, an over‐reliance on speculative building, and a corresponding lack of diversity in the UK's housing production model, and secondly, an under‐resourcing of public planning. The UK urgently needs a more mixed ecosystem of housing delivery, and greater support for non‐profit public and social providers who can bridge the gap between market demand and broader housing need
Songmakers and Texts in Early Greek Poetry Τραγουδοποιοί και κείμενα στην πρώιμη αρχαιοελληνική ποίηση
No full textThe article examines the early history of ancient Greek poetic texts in the light of what we know about their simultaneous oral and written tradition, suggesting that these facts should be allowed to influence how we conceive of fundamental concepts like text and work in our interpretation of the poetry