Nevin Manimala

Ann Lab Med. 2026 Jun 25. doi: 10.3343/alm.2025.0716. Online ahead of print.

ABSTRACT

BACKGROUND: Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality worldwide. To enhance early detection, the ASAP (age, sex, alpha-fetoprotein [AFP], protein induced by vitamin K absence or antagonist-II [PIVKA-II]) and GAAD (gender, age, AFP, des-gamma-carboxyprothrombin [DCP]/PIVKA-II) models were developed by integrating demographic data with serum biomarkers. We compared their performance in a Korean chronic liver disease cohort.

METHODS: We retrospectively analyzed data from 524 patients, including 132 with and 392 without HCC. AFP and PIVKA-II levels were measured using Abbott (ASAP) and Roche (GAAD) analyzers. Performance was assessed based on area under the ROC curve (AUROC) and optimal cutoff values for the overall cohort, etiologic subgroups (hepatitis B virus [HBV], hepatitis C virus [HCV], alcohol-related), and early-stage HCC (modified Union for International Cancer Control stage I or II).

RESULTS: In the overall cohort, both models demonstrated high, comparable performance (P =0.482). The ASAP model achieved an AUROC of 0.945 (sensitivity 81.8%, specificity 93.4%; cutoff 0.404), whereas the GAAD model yielded an AUROC of 0.950 (sensitivity 85.6%, specificity 93.6%; cutoff 1.34). No statistically significant differences were observed in etiologic subgroups or early-stage HCC (P =0.702), with AUROCs remaining high (0.911 for ASAP and 0.916 for GAAD).

CONCLUSIONS: The ASAP and GAAD algorithms provide excellent and comparable diagnostic performance for detecting HCC, including in early-stage cases, regardless of etiology. Given Korea’s high HBV prevalence and platform variability, these models serve as robust, non-invasive complementary tools for surveillance. This validation supports their clinical utility in the Korean population.

PMID:42343147 | DOI:10.3343/alm.2025.0716

Nature. 2026 Jun 24. doi: 10.1038/s41586-026-10702-5. Online ahead of print.

ABSTRACT

Galaxy redshift surveys map the cosmic web and provide a key observational test of whether the Universe becomes statistically homogeneous and isotropic on sufficiently large scales, as assumed by the cosmological principle underpinning the standard cosmological model¹. In this framework, beyond the nonlinear regime of structure formation, inhomogeneous and anisotropic features are expected to fade rapidly, reflecting the near-isotropic primordial density field and its subsequent gravitational evolution. Although supported by the small amplitude of cosmic microwave background anisotropies², this view is increasingly challenged by the complex network of large-scale structures and voids in the galaxy distribution^3-6, as well as by independent probes reporting possible large-scale deviations from statistical homogeneity⁷ and isotropy^8,9. Here we show that the galaxy distribution exhibits persistent anisotropic structures extending to scales on the order of one gigaparsec. Using the Angular Distribution of Pairwise Distances (ADPD)¹⁰, a parameter-free statistic that measures directional correlations, we detect anisotropy signals exceeding those in isotropic controls and geometry-matched ΛCDM mock catalogues with conservative significance greater than 3σ. These results provide direct evidence that directional coherence persists to larger scales than predicted in the standard framework, challenging the assumption of large-scale isotropy. They call for a reassessment of how homogeneity and isotropy are realized in the observed Universe and motivate new tests of cosmological models based on directional statistics.

PMID:42343127 | DOI:10.1038/s41586-026-10702-5

Nat Med. 2026 Jun 24. doi: 10.1038/s41591-026-04508-1. Online ahead of print.

NO ABSTRACT

PMID:42343117 | DOI:10.1038/s41591-026-04508-1

Sci Rep. 2026 Jun 25. doi: 10.1038/s41598-026-59459-x. Online ahead of print.

ABSTRACT

This study examines how wind characteristics can associate with dust events in Northeastern Iran’s Khorasan Razavi province, an area frequently affected by dust storms. Ten synoptic stations provided daily wind and dust data from 2014 to 2023, which were analyzed to explore seasonal and spatial patterns. All wind characteristics and dust event data were extracted from diurnal SYNOP codes and contributed to the GIS-based database. Then, data were interpolated to produce regionalization maps in GIS and to estimate temporal statistics and correlation coefficients in SPSS. Findings reveal that prevailing north and northwest winds are closely linked to dust storms across all stations and seasons, with the Sarakhs region identified as the main dust hotspot, especially during spring and summer when wind speeds peak. In Mashhad, dust events are also linked to southerly winds during autumn and winter. The study highlights that understanding wind direction and speed enhances the estimation of dust storms and supports strategies to mitigate their negative impacts. Practical recommendations include early warning systems in high-risk areas like Sarakhs based on wind-speed thresholds, land restoration in dust-prone wind sectors, and adjusting air quality alerts in Mashhad according to seasonal wind trends. These measures aim to protect public health and regional ecosystems from the adverse effects of dust storms.

PMID:42343108 | DOI:10.1038/s41598-026-59459-x

Eur Radiol. 2026 Jun 24. doi: 10.1007/s00330-026-12702-5. Online ahead of print.

ABSTRACT

OBJECTIVES: To evaluate three commercial AI software tools for pulmonary nodule detection and segmentation and to assess their impact on guideline-based management recommendations.

MATERIALS AND METHODS: A total of 740 CT and PET-CT studies from clinical routine were analyzed using three software tools (S1, S2, S3). We compared the total number of detected nodules and “actionable” nodules (per British Thoracic Society (BTS) definition). We further evaluated how measurement variations between tools affected hypothetical management according to Fleischner Society and BTS guidelines for incidental nodules.

RESULTS: The tools differed significantly in the total number of detections (S1: 1336; S2: 1060; S3: 1536; p < 0.001) and wrong findings (S1: 965; S2: 720; S3: 1169; p < 0.001). However, the detection of actionable nodules was comparable across all tools (S1: 375; S2: 341; S3: 373; p = 0.73). While no statistically significant differences were found in mean diameter or volume measurements, small absolute variations led to significant differences in management. Specifically, S2 triggered significantly more 1-year follow-up recommendations than S3 under BTS guidelines (p < 0.001). No significant management differences were observed when applying Fleischner Society guidelines.

CONCLUSION: While the three included AI tools show comparable performance in detecting actionable nodules, minor measurement variations significantly impact downstream management when using guidelines with narrow thresholds, such as the BTS criteria. Fleischner Society guidelines appear more robust to these inter-software variations.

KEY POINTS: Question How do commercial software tools for pulmonary nodule detection perform in real-world settings and impact hypothetical management under BTS and Fleischner guidelines? Findings Detection of actionable nodules was comparable across all tools, but small absolute measurement variations triggered significantly more 1-year follow-up recommendations under BTS guidelines. Clinical relevance AI software can cause inconsistent BTS-based management due to narrow thresholds, while Fleischner criteria appear more stable. Frequent detection of benign lesions potentially poses a risk of overdiagnosis and overtreatment in standalone AI-based reporting.

PMID:42343062 | DOI:10.1007/s00330-026-12702-5

Arch Sex Behav. 2026 Jun 24. doi: 10.1007/s10508-025-03375-0. Online ahead of print.

ABSTRACT

Artificial intelligence (AI) companions are increasingly used for social, emotional, and sometimes romantic fulfillment, raising questions about how people perceive and engage with these technologies. This study explored attitudes toward AI companionship and whether interviewer type (AI, human, or unmoderated) affects disclosure or engagement. A mixed-methods design was employed with 135 adult participants, who completed structured interviews including Likert scale items and open-ended questions about their views on AI companions. Most participants (55.6%) expressed hesitancy or resistance toward emotionally significant relationships with AI, although several reported openness to non-intimate or functional companionship such as friendship or mentoring. Motivations for engaging with AI included its non-judgment and trustworthiness, as well as its ability to provide emotional support and optimize daily life. Key barriers involved its lack of physicality, lack of humanity, incapacity to form an emotional connection, and privacy concerns. Participants in the unmoderated condition rated AI as a friend significantly higher than those in the human interviewer condition, and rated AI as equivalent to a human companion and as a potential romantic partner significantly higher than those in both the chatbot and human interviewer conditions. No statistically significant differences emerged across conditions in self-reported honesty. Participants’ engagement metrics were generally higher in the human and unmoderated interviews compared to AI. These findings offer methodological insights for research on sensitive topics and highlight the complex, context-dependent ways people relate to AI companionship. Implications for policy, clinical practice, and the design of future AI companion systems are discussed.

PMID:42343026 | DOI:10.1007/s10508-025-03375-0

J Imaging Inform Med. 2026 Jun 24. doi: 10.1007/s10278-026-02073-0. Online ahead of print.

ABSTRACT

Deep learning models achieve strong diagnostic performance in medical imaging, yet often exhibit systematic performance disparities across demographic subgroups. Although prior work has shown that attributes such as age, sex and race are encoded within internal representations, it remains unclear how the structure of these representations contributes to subgroup-level differences in prediction behaviour. This study aims to examine how demographic information is embedded in chest X-ray classifiers and how latent-space structure relates to observed sensitivity disparities. We analysed two large-scale chest X-ray datasets, CheXpert and MIMIC-CXR, using DenseNet-121 models trained for multi-label disease classification. In addition to standard output-level evaluation, we conducted representation-level analyses using linear probes, embedding statistics and geometric measures to characterise subgroup differences in activation strength, latent-space proximity and model confidence. Disparities were assessed across age, race and sex by jointly examining feature encodings, logits, energy scores and true-positive rates. Demographic attributes showed limited direct association with disease labels and low standalone predictive utility, yet were strongly encoded within internal features. Younger and Black/African American patients consistently exhibited higher feature norms, greater separation in latent space and lower joint logit energy, despite comparable overall discrimination performance. These representational patterns persisted after accounting for label configuration and were associated with larger sensitivity gaps, consistent with structural suppression in which certain subgroups occupy sparser, lower-activation regions of the representation space. Sex-based differences were comparatively modest across representational and performance metrics. Subgroup disparities in chest X-ray classification are closely linked to how demographic groups are positioned and activated within latent space, rather than to directional misalignment alone. Representation-level diagnostics based on activation magnitude, density and energy provide mechanistic insight into model behaviour and highlight limitations of mitigation strategies that focus solely on feature removal or post hoc thresholding. These findings support the use of representation-level analysis as a principled component of fairness evaluation and mitigation design in clinical AI systems.

PMID:42343005 | DOI:10.1007/s10278-026-02073-0

Nat Aging. 2026 Jun 24. doi: 10.1038/s43587-026-01158-3. Online ahead of print.

ABSTRACT

Geroscience has advanced rapidly, yet its clinical translation remains limited. A central barrier is the lack of trial outcomes that capture the multidimensional effects of geroprotective interventions while meeting clinical and regulatory standards. Mortality is objective and regulatorily salient but often impractical. By contrast, surrogate measures of healthspan improve feasibility and may better reflect the quality of extended life, but they are generally considered soft endpoints that require further validation. Here, we propose hierarchical composite endpoints using time-to-worst-event analysis as a pragmatic and scientifically sound compromise. Participant pairs are compared using win statistics according to a prespecified clinical hierarchy, in which more severe and objective clinical events are prioritized, while health surrogates and biomarkers contribute information at lower tiers. When outcome selection, ordering and tie rules are clinically and mechanistically justified and agreed with regulators, this approach may improve geromedicine trial efficiency and allow overall treatment effects to be captured without compromising clinical priorities.

PMID:42342910 | DOI:10.1038/s43587-026-01158-3

Sci Rep. 2026 Jun 25. doi: 10.1038/s41598-026-58662-0. Online ahead of print.

ABSTRACT

Concrete blocks made from waste aggregates have become a promising way to reduce waste and conserve natural resources while still offering good mechanical performance in both solid and hollow concrete blocks. This is especially relevant as more sustainable construction projects increasingly use recycled and alternative materials. This research develops a series of novel hybrid machine learning (ML) models to accurately predict compressive strength, using a dataset of 544 concrete samples from various sources. The six novel hybrid ML frameworks are designed as Hybrid Stacked Ensemble (HSE), Hybrid Residual Learning (HRL), Hybrid Weighted Ensemble (HWE), Hybrid Meta-Learning (HML), Hybrid Bayesian Stacking (HBS), and Hybrid Feature Fusion (HFF). Results show that novel Hybrid Bayesian Stacking (HBS) algorithms deliver excellent predictive accuracy across all evaluation metrics, with (R² = 0.998, RMSE = 0.665) during training and (R2 = 0.987, RMSE = 1.836) during testing. Furthermore, Individual Conditional Expectation (ICE) and SHapley Additive exPlanations (SHAP) analyses identified important input features and their effects on compressive strength. A graphical user interface (GUI) was developed to make predictive models accessible for practical engineering.

PMID:42342892 | DOI:10.1038/s41598-026-58662-0

Lasers Med Sci. 2026 Jun 25;41(1):131. doi: 10.1007/s10103-026-04893-6.

ABSTRACT

PURPOSE: Hidradenitis suppurativa (HS) is a chronic, debilitating skin disease often requiring multimodal therapy. Laser hair removal (LHR) is an emerging treatment option, yet patient-centered data is limited. This study aimed to assess patient perspectives on the effectiveness, safety, motivations, and barriers associated with LHR for HS.

METHODS: An anonymous cross-sectional online survey was administered via REDCap (July-December 2024) to adults with HS living in the United States. Respondents reported prior treatments, LHR parameters, outcomes, adverse effects, and barriers. Descriptive statistics were used.

RESULTS: Of 110 participants who completed the survey (110/132, 83%), 24 (22%) had used LHR and comprised the analytic cohort, with a median of 8 LHR sessions (IQR 6-12). Leading motivations included reducing inflammation (92%), relieving pain (75%), and seeking durable treatment (71%). Highest median improvements (score 4, IQR 4-5) were in lumps/abscesses, swelling, and flare frequency. Other symptoms, including pain, odor, and quality of life, also showed moderate improvement. Half reported benefits lasting over 12 months. While biologics were perceived as most effective (median 4.5, IQR 3.5-5), LHR received one of the highest median scores among non-biologic options (3.5, IQR 3-5). Barriers included cost, insurance limitations, and low awareness; 63% paid over $1,000, and 38% discontinued early. Common adverse effects included discomfort (71%) and transient erythema (46%).

CONCLUSION: Most patients perceived LHR as beneficial for HS, but affordability and awareness remain barriers. Findings highlight the need for payer advocacy and additional trials defining LHR’s role in HS management.

PMID:42342886 | DOI:10.1007/s10103-026-04893-6