Category: Statistics

Radiol Cardiothorac Imaging. 2026 Aug;8(4):e250464. doi: 10.1148/ryct.250464.

ABSTRACT

Purpose To identify a stable radiomic feature (RF) subset for pulmonary nodules by analyzing test-retest and acquisition parameter variability using phantoms and to evaluate its clinical utility. Materials and Methods CT scanners at multiple centers were used to scan a chest phantom embedded with artificial pulmonary nodules under varied doses and scanning and reconstruction parameters. An artificial intelligence-based, fully automated three-dimensional segmentation method was used. Three sources of variability were examined: test-retest, inter-CT scanner, and intra-CT scanner. RF stability was assessed using the concordance correlation coefficient (CCC), dynamic range, and intraclass correlation coefficient (ICC). Statistical tests were used to analyze the impact of these factors, with stable RFs identified through UpSet plots. Multiple radiomic models were constructed and validated using multicenter clinical ground-glass nodule data. Results In terms of test-retest variability, repeatable RFs represented 1715 of 2264 features (76%, CCC > 0.9). The reproducibility rate for inter-CT variability was 64% (1439 of 2264 features, ICC > 0.9). Variance component analysis identified section thickness/interval (24.9%), low-dose scanning (24.3%), inter-CT scanner variation (17.5%), and field of view (FOV, 14.9%) as the largest contributors to radiomic feature variability, with consistent influence patterns across all acquisition parameters. A stable subset of 466 RFs was used to construct six radiomics models, which, compared with prescreening RF models, demonstrated greater clinical efficacy and predictive performance on the validation sets (AUC range, 0.86 [95% CI: 0.81, 0.91] to 0.90 [95% CI: 0.86, 0.94)]. Conclusion Variations in section thickness, FOV, interscanner differences, and low-dose settings significantly affected RF stability. A stable RF subset for pulmonary nodules was identified and demonstrated clinical utility. Keywords: Phantom Studies, Applications-CT, Clinical Testing, Machine Learning, Model Training, Model Validation, Radiomics, Pulmonary, Lung, Pulmonary Nodule, Computed Tomography, Repeatability, Reproducibility Supplemental material is available for this article. © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license.

PMID:42658069 | DOI:10.1148/ryct.250464

Chaos. 2026 Aug 1;36(8):082101. doi: 10.1063/5.0337419.

ABSTRACT

Complex systems are difficult to study not only because they are nonlinear, multiscale, and nonstationary, but because their scientifically relevant organization is often distributed across components, relations, and interaction orders. Topology provides a mathematical language for describing that organization through connectedness, recurrence, branching, closure, cavities, and persistence across scale. This review synthesizes persistent homology, Mapper, simplicial complexes, hypergraphs, and relation-level operator methods through a unified workflow from empirical data to representation, topological construction, output, and scientific interpretation. Across nonlinear dynamics, finance, neuroscience, biology, ecology, materials, and engineered systems, topological and topology-inspired methods make state-space organization, collective constraints, and structural reorganization available as observables that can be integrated with statistics, dynamics, mechanistic models, and machine learning. The review distinguishes the claim that a representation makes structure visible from the stronger claim that it improves detection or prediction, and it summarizes comparative evidence where such benchmarks exist. Prospective early-warning evidence remains uneven, but several studies demonstrate useful structural diagnostics, data-efficient classification, anomaly detection, and reductions in false alarms. The central conclusion is that topology is most valuable when representation is treated as a scientific hypothesis and topological descriptions are connected to domain-matched inference and mechanism.

PMID:42658065 | DOI:10.1063/5.0337419

Pediatr Int. 2026 Jan-Dec;68(1):e70534. doi: 10.1111/ped.70534.

ABSTRACT

BACKGROUND: Pediatric home mechanical ventilation (HMV) use has risen globally, driven by advances in pediatric critical care and a shift toward community-based management. However, long-term outcome data from Japanese regional core hospitals remain limited.

METHODS: We conducted a single-center retrospective cohort study of all patients between 1995 and 2024 aged ≤ 20 years for whom there was initiation of HMV (n = 37). HMV included invasive mechanical ventilation (tracheostomy positive-pressure ventilation) and/or noninvasive ventilation. Cumulative survival after HMV initiation was estimated using the Kaplan-Meier method and compared using log-rank tests according to age at initiation (≤ 7 vs. > 7 years), ventilation modality (invasive mechanical ventilation vs. noninvasive ventilation), predominant pathophysiological indication (respiratory pump insufficiency [Pump], airway disease [Airway], or restrictive lung mechanics [Restrictive]), and presence of severe motor and intellectual disabilities (SMID).

RESULTS: The Kaplan-Meier-estimated cumulative survival was 69.1% (maximum follow-up, 29.1 years). In our cohort, survival did not significantly differ by age at initiation or ventilation modality. In contrast, it was significantly lower in patients with restrictive lung mechanics and lower in those with SMID. Among the eight deaths, cardiac disease was common, and death was seen to be clustered in the winter months; it often followed respiratory infections, and it was associated with notable growth impairment.

CONCLUSIONS: Our results suggest that children receiving HMV can achieve long-term survival. Nevertheless, high-risk groups, particularly those with restrictive lung mechanics or SMID, warrant risk-stratified care, proactive prevention of winter infection, and timely advance care planning.

PMID:42658053 | DOI:10.1111/ped.70534

Am J Biol Anthropol. 2026 Aug;190(4):e70338. doi: 10.1002/ajpa.70338.

ABSTRACT

OBJECTIVES: This paper uses deciduous dental morphology to evaluate the spatial structure of the Archaic period site of Phaleron (Athens, Greece) with the goal of identifying possible biological kin- or community structuring of the burial ground.

MATERIALS AND METHODS: Deciduous morphological data (77 total root-trait combinations) were collected by the author using published standards for 231 preadult individuals from Phaleron. After pre-analysis data treatments evaluating the effects of age, sex, and inter-trait correlation, remaining variables were compared among burial subsectors using Mean Measure of Divergence statistics, with statistical significance assessed following both standard protocols and with bootstrapping methods.

RESULTS: Results are consistent with, but not identical to, those presented using permanent dental morphology, with deciduous data presenting stronger evidence of spatial structuring. This suggests that the same principles governing grave placement applied to both adults and preadults, which I argue reflects biological kinship as a structuring principle of Archaic period burial grounds.

DISCUSSION: This paper makes three contributions: (1) enhances understanding of ancient Greek burial practices, particularly of the non-elite individuals who lived during the critical centuries of Athenian political development, (2) demonstrates the complementary value of using both deciduous and permanent dentition in analyses of biological distance, and (3) presents the first large sample of deciduous trait frequencies from ancient Greece, thus informing future comparative studies.

PMID:42658037 | DOI:10.1002/ajpa.70338

Pediatr Int. 2026 Jan-Dec;68(1):e70527. doi: 10.1111/ped.70527.

ABSTRACT

BACKGROUND: Kawasaki disease is an acute systemic vasculitis, and the clinical significance of serum thymus and activation-regulated chemokine (TARC) remains incompletely understood. We examined whether pre-treatment age-adjusted serum TARC was associated with inflammation-based intravenous immunoglobulin (IVIG) resistance risk stratification.

METHODS: We retrospectively studied 109 children with Kawasaki disease after excluding documented allergic comorbidities and age younger than 6 months. All included patients had serum TARC measured before initial treatment, and 18 of them also had paired post-treatment measurements. Serum values were expressed as the percentage of the age-specific upper limit of normal. The primary analysis compared pre-treatment values between high- and low-risk groups according to the Kobayashi score. Paired pre- and post-treatment changes and exploratory analyses of other IVIG resistance risk scores and treatment non-response were also assessed.

RESULTS: Twenty children were classified as high risk by the Kobayashi score. Pre-treatment age-adjusted values were lower in the high-risk group than in the low-risk group, but logistic regression analysis did not support a statistically robust association. In 18 children with paired samples, values decreased significantly after treatment. Exploratory analyses showed no significant associations with other IVIG resistance risk scores, whereas analyses of treatment non-response and coronary artery lesions were limited by the small number of events.

CONCLUSIONS: Age-adjusted serum TARC showed limited clinical associations during the acute phase of Kawasaki disease and was not robustly associated with inflammation-based high-risk classification. Its role in treatment response and coronary artery outcomes remains to be clarified.

PMID:42658035 | DOI:10.1111/ped.70527

Indian J Cancer. 2026 Aug 26. doi: 10.4103/ijc.ijc_172_24. Online ahead of print.

ABSTRACT

BACKGROUND: We aimed to explore a method to measure the preoperative repair area of the thyroid gland flap (TGF) in patients with neck cancer.

METHODS: A total of 82 patients with nonthyroid diseases were included in this study. The digital imaging and communications in medicine (DICOM) data of enhanced computed tomography (CT) of the neck were collected. Three-dimensional (3D) reconstruction was performed using Mimics software, and data on the thyroid gland was collected. The projected area of the unilateral glandular lobe was measured. The demographic data of the patients, such as sex, age, height, and body mass index, were collected. The influencing factors for the repair area of TGF were analyzed.

RESULTS: As a pedicled flap, the left thyroid lobe could provide a repair area of 7.27 ± 2.00 cm 2 , and the right thyroid lobe could provide 8.09 ± 2.06 cm 2 . A statistically positive correlation was found between the aforementioned area and patients’ height ( P < 0.05).

CONCLUSION: The preoperative repair area of TGF could be measured using enhanced CT and 3D reconstruction for patients with neck tumors. 3D reconstruction can be used to perform preoperative evaluation of TGF application.

PMID:42658033 | DOI:10.4103/ijc.ijc_172_24

Bioinformatics. 2026 Aug 27:btag635. doi: 10.1093/bioinformatics/btag635. Online ahead of print.

ABSTRACT

MOTIVATION: Prioritizing therapeutic targets from high-dimensional transcriptomic profiles is hindered by the underdetermined nature of the p ≫ n setting. While miRNA signatures can inform target prioritization, conventional accuracy-driven methods may yield unstable predictive signatures, reducing downstream network reliability and topology-guided candidate ranking.

RESULTS: We propose TargetPrior, a stability-aware evolutionary learning framework in which EL-CAML derives reproducible miRNA anchors from relapse-associated transcriptomic variation for candidate target prioritization. In childhood acute myeloid leukemia (CAML), EL-CAML identifies a parsimonious 18-miRNA continuous relapse-risk signature and 10 complementary stability-supported biomarkers, yielding 28 miRNAs for literature-curated miRNA-gene network construction. Repeated perturbation analysis supported the stability of high-frequency miRNAs, while analysis of the independent GSE196886 cell-sorted small RNA-seq dataset identified cell-population-specific expression differences. Benchmarking against an expanded set of clinically and biologically supported AML target references showed stronger early-rank retrieval than network-only and statistical approaches. TargetPrior is presented as a computational proof-of-concept for generating prioritized therapeutic hypotheses, rather than as a universal target-discovery solution.

AVAILABILITY: Code is available at: https://github.com/NYCU-ICLAB/TargetPrior and archived on Zenodo (DOI: 10.5281/zenodo.20394263).

SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

PMID:42658031 | DOI:10.1093/bioinformatics/btag635

Stat Med. 2026 Sep;45(20-22):e70720. doi: 10.1002/sim.70720.

ABSTRACT

While balancing covariates between groups is central for observational causal inference, selecting which features to balance remains a challenging problem. Kernel balancing is a promising approach that first estimates a kernel that captures similarity across units and then balances a (possibly low-dimensional) summary of that kernel, indirectly learning important features to balance. In this paper, we propose forest kernel balancing, which leverages the underappreciated fact that tree-based machine learning models, namely random forests and Bayesian additive regression trees (BART), implicitly estimate a kernel based on the co-occurrence of observations in the same terminal leaf node. Thus, even though the resulting kernel is solely a function of baseline features, the selected nonlinearities and other interactions are important for predicting the outcome-and therefore are important for addressing confounding. Through simulations and applied illustrations, we show that forest kernel balancing leads to meaningful computational and statistical improvement relative to standard kernel methods, which do not incorporate outcome information when learning features.

PMID:42658030 | DOI:10.1002/sim.70720

Bioinformatics. 2026 Aug 27:btag640. doi: 10.1093/bioinformatics/btag640. Online ahead of print.

ABSTRACT

BACKGROUND: Timeseries clinical transcriptomic datasets offer the opportunity to gain insights into the dynamics of disease mechanisms/treatment responses. However, their utility in uncovering temporal patterns is often limited by high noise levels and small sample sizes. Leveraging foundational gene embeddings and incorporating interaction information can help address these challenges, improve gene network analysis, and enable the detection of subtle changes that drive disease progression or drug response.

RESULTS: We finetuned gene embeddings from foundation models using healthy tissue gene expression data and used them in temporal GNNs to model gene expression of responder and non-responders to treatment in 4 disease datasets – ulcerative colitis, Crohn’s disease, Alopecia Areata, and psoriasis. Application of our method to these datasets confirmed known mechanisms associated with drug action, and also identified key differences between activated and repressed pathways for responders and non-responders, including B-Cell activation and mitochondria-related activity in ulcerative colitis patients.

CONCLUSION: Finetuning gene embeddings from foundation models provides a richer context to model gene expression data compared to using them in their naive state. Even with smaller sample sizes, results from GNN-based temporal models outperform traditional methods by detecting known mechanisms of response and unraveling role of genes and mechanisms not known to be associated with response and non-response.

CODE AVAILABILITY: Code and data are available in a public GitHub repository – https://github.com/Sanofi-Public/GNN-Timeseries. DOI: 10.5281/zenodo.20494035.

PMID:42658019 | DOI:10.1093/bioinformatics/btag640

Transl Vis Sci Technol. 2026 Aug 3;15(8):28. doi: 10.1167/tvst.15.8.28.

ABSTRACT

PURPOSE: To evaluate how reliably single Goldmann applanation tonometry (GAT) measurements reflect the preceding month’s intraocular pressure (IOP) profile in operated open-angle glaucoma.

METHODS: This longitudinal study included 24 patients with a suprachoroidal telemetric IOP sensor analyzing 156 in-hospital GAT measurements, 154 simultaneous telemetric readings, and 13,734 at-home measurements across follow-up visits. Telemetric data from the 30 days prior to each visit were stratified by time of day and matched to corresponding GAT measurements, which, in turn, were ranked relative to the interquartile range (IQR) of the preceding at-home measurements. Peak-and-trough analyses were performed using hourly aggregated data.

RESULTS: The GAT and telemetric IOP measurements showed good agreement (mean ΔIOP = -0.1 ± 3.0 mm Hg). However, 59.7% of the GAT-measured IOP values exceeded the IQR, and 14.3% fell below the IQR of the corresponding telemetric distributions; only 26.0% were within the IQR. Deviations from the telemetric median were 4.2 ± 2.4 mm Hg (above IQR), -2.3 ± 1.5 mm Hg (below IQR), and 0.5 ± 1.2 mm Hg (within IQR). Of the 99 statistically significant extrema (45 peaks and 54 troughs), most peaks occurred within 3 hours before GAT measurements. Troughs were observed both before and after GAT, but with larger nyctohemeral misalignment to GAT.

CONCLUSIONS: Single GAT measurements frequently failed to represent the underlying 30-day IOP profile and often missed clinically relevant peaks and troughs, despite good pointwise agreement with telemetric measurements.

TRANSLATIONAL RELEVANCE: High-frequency telemetric monitoring provides a more comprehensive assessment of IOP dynamics and may improve individualized glaucoma management compared with infrequent GAT snapshot measurements.

PMID:42658016 | DOI:10.1167/tvst.15.8.28