Category: Statistics

Adv Clin Exp Med. 2026 Jun 5. doi: 10.17219/acem/220688. Online ahead of print.

ABSTRACT

BACKGROUND: Intracardiac leads commonly produce metal artifacts on computed tomography (CT) images. These artifacts may be reduced using dedicated metal artifact reduction algorithms, such as metal artifact reduction (MAR).

OBJECTIVES: The aim of this study was to develop a method for measuring lead-related artifacts in CT and to assess the suitability of various reconstruction presets for lead visualization.

MATERIAL AND METHODS: Fifty-four patients (mean age: 73.9 ±11.32 years) with implanted cardiac implantable electronic devices (CIEDs) who underwent cardiac CT, chest CT, or pulmonary angio-CT were included in the study. Images were reconstructed using at least 2 kernels (soft tissue and lung) with slice thicknesses of 0.6 mm or 1.0 mm. A tissue density volume >1,000 HU, corresponding to the presumed volume of hyperdense artifacts, was isolated within a manually drawn spherical region of interest (ROI), and the values were recorded. The obtained values for each iterative metal artifact reduction (iMAR) reconstruction preset were compared with native images (without iMAR) to calculate the percentage reduction in hyperdense artifacts.

RESULTS: All tested algorithm variants reduced artifact volume; however, only 2 presets achieved statistically significant reductions: “dental fillings” (p = 0.001) and “neuro coils” (p = 0.000). Pacemaker-dedicated presets reduced metal artifacts in all cases, although the reductions were not statistically significant (p = 0.667), which may limit their reliability in routine clinical practice.

CONCLUSIONS: We proposed a method for evaluating intracardiac leads that enables precise three-dimensional (3D) assessment of hyperdense artifacts. The metal artifact reduction technique demonstrated promising results, particularly for the “dental fillings” and “neuro coils” presets.

PMID:42247616 | DOI:10.17219/acem/220688

ESC Heart Fail. 2026 Jun 5:xvag148. doi: 10.1093/eschf/xvag148. Online ahead of print.

ABSTRACT

BACKGROUND: The renal effects of sacubitril/valsartan (Sac/Val) in heart failure (HF) remain incompletely defined, partly because kidney outcomes in pivotal HF trials have been variably prespecified and heterogeneously defined. We performed an updated systematic review and meta-analysis to assess whether the apparent renal signal of Sac/Val varies according to endpoint definition.

METHODS: This updated systematic review was conducted in accordance with PRISMA 2020. Building on the pre-existing evidence base from prior meta-analyses, we performed an updated search of PubMed and Web of Science. Randomized controlled trials (RCTs) and observational comparative studies in adults with HF comparing Sac/Val with angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, or standard care were included if renal outcome data were extractable. Renal outcomes were analyzed using a hierarchical cross-endpoint approach, including sustained ≥50% estimated glomerular filtration rate (eGFR) decline, end-stage kidney disease (ESKD), composite kidney outcomes, and annualized eGFR decline. Random-effects meta-analysis was performed, with Hartung-Knapp and RCT-only sensitivity analyses.

RESULTS: Overall, 13 study-level reports comprising 34,969 patients were included. Sac/Val was associated with a lower risk of sustained 50% eGFR decline (RR 0.68, 95% CI 0.57-0.82) and composite kidney outcome (RR 0.70, 95% CI 0.58-0.84), with the composite endpoint showing the most robust and consistent signal, including in RCT-only analyses. By contrast, the association for ESKD alone was directionally favorable but not statistically significant (RR 0.80, 95% CI 0.64-1.00). Sac/Val was also associated with a slower annualized eGFR decline (MD 0.52 mL/min/1.73 m2/year, 95% CI 0.35-0.69).

CONCLUSIONS: The renal signal associated with Sac/Val in HF appeared at least partly dependent on endpoint definition. Composite kidney outcomes may best capture its potential nephroprotective effect, with sustained 50% eGFR decline showing a consistent pattern, whereas isolated ESKD remains inconclusive. These findings support a potential nephroprotective role of ARNIs but future research are needed.

PMID:42247581 | DOI:10.1093/eschf/xvag148

Scand J Clin Lab Invest. 2026 Jun 5:1-11. doi: 10.1080/00365513.2026.2684025. Online ahead of print.

ABSTRACT

Standard operating procedures (SOPs) are foundational to quality assurance in ISO-accredited clinical laboratories, but manual SOP development is labor-intensive and prone to error. With the emergence of large language models (LLMs), such as ChatGPT-5, there is growing interest in their potential to generate compliant, context-specific laboratory documentation. To test a regulatory-aligned, clause-based evaluation framework for assessing whether ChatGPT-5 can generate context-aware, fit-for-purpose SOP drafts within a real ISO-aligned laboratory environment. In a single-site, paired proof-of-concept study, 10 high-priority SOPs were generated using ChatGPT-5 and compared with matched SOPs written by experienced laboratory professionals. AI prompts were enriched with laboratory-specific inputs (e.g. operator manuals, reagent inserts and LIS codes). SOPs were evaluated using a seven-domain ISO/CLSI-aligned rubric, a laboratory-specificity audit, clause-mapping checklists, content validity indexing and usability testing by junior staff. Inter-rater reliability and paired non-parametric statistics were applied. AI-assisted SOPs demonstrated higher median quality scores, more complete ISO clause referencing, improved traceability and stronger lifecycle conformity compared with manual SOPs. Drafting time was reduced by approximately 91%. Expert reviewers showed excellent agreement (ICC = 0.91), and content validity indices exceeded established thresholds. Junior staff rated AI-assisted SOPs as clearer and more independently usable. Context-anchored prompts improved laboratory-specific relevance. This study demonstrates a replicable, clause-based framework for evaluating AI-assisted SOP generation within a regulated laboratory context. While findings support the feasibility of AI as a documentation co-author under expert oversight, external multi-center validation is required before broader regulatory or clinical adoption.

PMID:42247578 | DOI:10.1080/00365513.2026.2684025

JMIR Nurs. 2026 Jun 5;9:e88567. doi: 10.2196/88567.

ABSTRACT

BACKGROUND: For clinical nurses, manually entering information into hospital information systems (HISs) remains time-consuming and prone to omissions. Although speech recognition can reduce the need for manual entry, its use in clinical settings has historically been limited by code-switching, medical terminology, and noisy ward environments. Recent advances in customized automatic speech recognition (ASR) and large language models (LLMs) now make speech-based, structured documentation aligned with nursing frameworks such as DART (data, action, response, and teaching) increasingly feasible.

OBJECTIVE: This study developed and evaluated an integrated ASR and LLM system that transforms spoken nursing input into structured DART notes and evaluated its accuracy, usability, and clinical feasibility within HIS workflows.

METHODS: A code-switching nursing speech corpus from emergency and ward settings was used to fine-tune the Whisper large-v2 model with parameter-efficient adaptation. The LLM generated schema-constrained DART records from ASR transcripts, which were verified by nurses before being uploaded to the corresponding HIS fields. Evaluation included mixed error rate for ASR accuracy, F1-scores, and agreement statistics for DART classification, hallucination assessments based on factual correctness, and analysis of nurse feedback on system use.

RESULTS: The fine-tuned ASR model reduced the mixed error rate from 44.79% to 6.67%. DART generation achieved a macroaveraged F1-score of 0.82 (95% CI 0.80-0.84) and met the noninferiority margin relative to human transcripts (δ=-0.04). The hallucination rate was 2.51%. During deployment, the monthly volume of valid nursing notes generated through voluntary use of the ASR system increased from 32,724 to 65,417, where each note represented a single documentation entry generated per patient care episode. Among 120 participating nurses, 91 (75.8%) reported reduced workload and improved completeness.

CONCLUSIONS: The integrated ASR and LLM system was feasible and showed strong performance, with good acceptance among clinical nurses. It reduced the manual documentation burden, improved record completeness, and demonstrated the value of an ASR- and LLM-supported workflow for nursing documentation.

PMID:42247576 | DOI:10.2196/88567

Biomol Biomed. 2026 Jun 5. doi: 10.17305/bb.2026.14308. Online ahead of print.

ABSTRACT

Early hospital arrival is essential for timely reperfusion therapy in acute stroke. This study aimed to identify clinical, symptom-related, and prehospital factors associated with delayed hospital arrival among patients with first-ever acute stroke, with particular focus on the 4.5-hour therapeutic window. This prospective single-center observational study was conducted at a comprehensive stroke center in Istanbul, Türkiye, between December 2023 and October 2024. Among 362 patients screened for suspected acute cerebrovascular events, 205 adults with first-ever acute ischemic stroke, intracerebral hemorrhage, or transient ischemic attack managed through the acute stroke pathway were included. Data were collected using a structured clinical form and a questionnaire on barriers to accessing acute stroke treatment, administered through face-to-face interviews with patients’ relatives. Prehospital delay was defined as the interval from last known well (LKW) to hospital arrival. Patients were analyzed according to arrival within 1 hour, 2 hours, 3 hours, and 4.5 hours, using appropriate comparative statistical tests. The mean LKW-to-arrival time was 338.66 ± 345.67 minutes, with a median (IQR) of 240 (90-720) minutes. Overall, 29.3% of patients arrived within 1 hour, 41.0% within 2 hours, 48.3% within 3 hours, and 58.5% within 4.5 hours. Facial droop was consistently associated with earlier hospital arrival across multiple time thresholds (p ≤ 0.004), and syncope was more frequent among early presenters (p = 0.001). Conversely, visual symptoms were associated with delayed presentation, including vision loss after 3 hours and diplopia beyond 4.5 hours (p = 0.042 for both). Diabetes mellitus was associated with delayed arrival at the 1-hour and 2-hour thresholds, while hypertension was more common among patients arriving after 4.5 hours. Prehospital delay remains a substantial barrier to timely acute stroke care. Recognizable symptoms such as facial droop may facilitate earlier presentation, whereas less typical symptoms, particularly visual disturbances, may contribute to delayed arrival. These findings support locally tailored public awareness strategies and optimized prehospital stroke pathways that emphasize both typical and atypical stroke symptoms.

PMID:42247575 | DOI:10.17305/bb.2026.14308

JMIR AI. 2026 Jun 5;5:e77351. doi: 10.2196/77351.

ABSTRACT

BACKGROUND: Medical documentation imposes a significant administrative burden on physicians and reduces time for direct patient care. Artificial intelligence (AI)-assisted tools such as automatic speech recognition and large language models (LLMs) promise to reduce this burden, but their performance in multilingual environments has not been explored. Switzerland is highly multilingual, and non-native German-speaking physicians may find documentation particularly challenging.

OBJECTIVE: This study aimed to compare the efficiency and documentation quality of four clinical documentation workflows-including both AI-assisted and traditional methods-in a Swiss tertiary hospital setting characterized by linguistic diversity.

METHODS: In this proof-of-concept study at a Swiss tertiary hospital (Department of Plastic and Hand Surgery, Cantonal Hospital Aarau), two physicians-a native Swiss German speaker and a non-native German speaker-documented encounters with simulated patients having common hand disorders. Four documentation workflows were tested: (1) traditional dictation with transcription by a secretary; (2) real-time dictation using speech recognition software for voice to text transcription; (3) postencounter dictation transcribed by an AI (Whisper) and processed by a GPT-based agent; and (4) AI-assisted ambient dictation of entire appointments using audio recording and automatic transcription. Documentation efficiency was measured by recorded physician time, and note quality was assessed using a modified Physician Documentation Quality Instrument (PDQI-9) scored by three different LLMs. To protect patient privacy, only synthetic (simulated) patient data were used.

RESULTS: AI-assisted workflows-particularly workflow 4 (AI-assisted ambient dictation)-produced the shortest physician documentation times per report. In post-hoc comparisons, workflow 4 was significantly faster than solely the speech recognition software workflow (workflow 2) for both physicians (adjusted P<.001). For the non-native speaker, workflow 4 was not significantly faster than traditional dictation (workflow 1) after adjustment (P=.08). LLM evaluators assigned high absolute scores (median PDQI-9 >47/50); however, inter-rater reliability was poor (Krippendorff’s alpha=-.433, 95% CI: -0.444 to -0.416), indicating systematic disagreement that precludes definitive conclusions about documentation quality from these scores alone.

CONCLUSIONS: AI-assisted documentation demonstrated significant time savings for the native speaker, though the reduction for the non-native speaker did not reach statistical significance in this pilot (P=.08). Such tools show potential to alleviate the linguistic challenges faced by non-native speakers, reduce administrative burdens, and enable physicians to spend more time with patients. However, the inconsistency of AI-based quality scoring suggests that LLMs cannot yet reliably replace human evaluation. Future studies should evaluate these workflows in real-world clinical implementation, address data privacy and security issues, and include human evaluators to validate the benefits observed in this study.

PMID:42247573 | DOI:10.2196/77351

Proc Natl Acad Sci U S A. 2026 Jun 9;123(23):e2528457123. doi: 10.1073/pnas.2528457123. Epub 2026 Jun 5.

ABSTRACT

Dynamical heterogeneity, in which transitory local fluctuations occur in the conformation and dynamics of constituent particles, is widely hypothesized to be essential to the evolution of supercooled liquids into the structural glass state. Yet its microscopic spatiotemporal phenomenology is challenging to detect directly in molecular glass forming liquids. Because recent theoretical advances predict that corresponding dynamical heterogeneity could occur in supercooled magnetic monopole fluids (Proc. Nat. Acad. Sci. 112, 8549 (2015)), we searched for such phenomena in Dy₂Ti₂O₇. By measuring its microsecond-resolved spontaneous magnetization fluctuations [Formula: see text] we detected a sharp bifurcation in monopole noise characteristics below [Formula: see text], with the appearance of powerful spontaneous monopole current bursts. This intense dynamics emerges upon entering the supercooled monopole fluid regime, reaches maximum strength near [Formula: see text] and then collapses along with coincident loss of ergodicity approaching [Formula: see text]. Moreover, when the four-point dynamical susceptibility [Formula: see text] is determined directly from temperature dependence of correlations in [Formula: see text], it evolves as predicted when dynamical heterogeneity is present, revealing its simultaneously and rapidly escalating length and time scales, [Formula: see text] and [Formula: see text]. This overall phenomenology greatly expands our empirical knowledge of supercooled monopole fluids and, more generally, demonstrates techniques for detection of the time sequence, magnitude, statistics, and correlations of dynamical heterogeneity, access to which may greatly accelerate fundamental vitrification studies.

PMID:42247567 | DOI:10.1073/pnas.2528457123

Proc Natl Acad Sci U S A. 2026 Jun 9;123(23):e2522504123. doi: 10.1073/pnas.2522504123. Epub 2026 Jun 5.

ABSTRACT

We prove that kernel covariance embeddings lead to information-theoretically perfect separation of distinct continuous probability distributions. In statistical terms, we establish that testing for the equality of two nonatomic (Borel) probability measures on a locally compact uncountable Polish space is equivalent to testing for the singularity between two centered Gaussian measures on a reproducing kernel Hilbert space. The corresponding Gaussians are defined via the notion of kernel covariance embedding of a probability measure, and the Hilbert space is that generated by the embedding kernel. Distinguishing singular Gaussians is structurally simpler from an information-theoretic perspective than nonparametric two-sample testing, particularly in complex or high-dimensional domains. This is because singular Gaussians are supported on essentially separate and affine subspaces. Our proof leverages the classical Feldman-Hájek dichotomy, and shows that even a small perturbation of a continuous distribution will be maximally magnified through its Gaussian embedding. This “separation of measure phenomenon” appears to be a blessing of infinite dimensionality, by means of embedding, with the potential to inform the design of efficient inference tools in considerable generality. The elicitation of this phenomenon also appears to crystallize, in a precise and simple mathematical statement, a core mechanism underpinning the empirical effectiveness of kernel methods.

PMID:42247566 | DOI:10.1073/pnas.2522504123

IEEE Trans Neural Syst Rehabil Eng. 2026 Jun 5;PP. doi: 10.1109/TNSRE.2026.3700284. Online ahead of print.

ABSTRACT

Cochlear implants (CIs) restore hearing in individuals with severe sensorineural hearing loss. In recent years, electrically evoked auditory steady-state responses (EASSRs) to amplitude modulated (AM) signals have been studied as an objective measure. EASSRs can be objectively detected in electroencephalography (EEG) recordings at the modulation frequency using statistical tests. However, the presence of electrical stimulation artifacts from the CI itself hinders the EASSR detection. Whereas previous research has focused on an experimental characterization of these artifacts, this study presents a theoretical analysis of the stimulation signal together with an experimental analysis of the resulting artifacts to characterize their properties, origins and the effects of system nonlinearities.

METHODS: A stimulation signal model is presented and analyzed. The effects of pulse asymmetry and nonlinearity are examined. The theoretical statements are experimentally validated using an experimental setup containing a head phantom.

RESULTS: The analysis shows that the stimulation artifact at the modulation frequency is inherent to the stimulation signal, even in the absence of system nonlinearities. Moreover, when the pulse asymmetry is taken into account, second and higher order polynomial nonlinearities are found to contribute negligibly to the spectral component at the modulation frequency.

SIGNIFICANCE: The experimental analyses indicate the proposed signal model is a more accurate model for the stimulation signal and the resulting stimulation artifact at the modulation frequency. The model may form an important step in determining artifact contamination in EEG recordings of EASSRs and other envelope-following responses in CI recipients, enabling improved response detection.

PMID:42247548 | DOI:10.1109/TNSRE.2026.3700284

IEEE Trans Biomed Eng. 2026 Jun 5;PP. doi: 10.1109/TBME.2026.3700464. Online ahead of print.

ABSTRACT

OBJECTIVE: Patient-specific instrumentation (PSI) and robotic systems are being developed to improve Total Knee Arthroplasty outcomes in severe knee osteoarthritis. In image-based solutions, preoperative CT scans provide accurate bone geometry for planning, but not cartilage, which can affect intraoperative registration, thus surgical accuracy, if not addressed properly. Solutions exist, such as bone probing through cartilage, but they require additional surgical steps. As an alternative, Statistical Shape Models (SSMs) can automatically predict cartilage from bone shape. The present study evaluates whether SSMs can capture pathological variability, which features complex patterns of cartilage loss and bone deformation.

METHODS: Segmentations from the OAI-ZIB database were classified into healthy and pathological groups based on joint space narrowing. Coupled bone-cartilage SSMs were trained separately on these groups, for femur and tibia. The performance of the SSM was assessed by comparing bone fitting and cartilage prediction accuracy.

RESULTS: Bone fitting errors (median 0.27-0.32 mm) and cartilage prediction errors (median 0.41-0.49 mm, RMSE 0.66-0.79 mm) were comparable to the literature, with cartilage errors close to the inter-observer MRI manual segmentation variability reported for similar datasets. Predictive performance was similar for healthy and pathological cases, suggesting that SSMs can capture pathological variability. However, osteophytes were not fully captured, locally affecting prediction accuracy.

CONCLUSION: The four coupled SSMs accurately reconstructed bone and predicted cartilage in both healthy and arthritic knees, suggesting robustness to pathological variability and suitability for clinical integration.

SIGNIFICANCE: The proposed method could be integrated into CT-based PSI or robotic workflows without requiring MRI or additional steps.

PMID:42247542 | DOI:10.1109/TBME.2026.3700464