Category: Statistics

Eur Radiol. 2025 Aug 4. doi: 10.1007/s00330-025-11887-5. Online ahead of print.

ABSTRACT

OBJECTIVE: Phantoms are essential for minimizing radiation-intensive experiments on humans and enhancing patient safety through model-based teaching and experimentation. While many phantoms are commercially available, their widespread use is limited by manufacturing complexity and high costs. This study aims to demonstrate a dough-based method to create customizable, realistic CT phantoms using affordable, readily available ingredients.

MATERIALS AND METHODS: For this study, various doughs composed of flour, salt, water, and oil were created, scanned, and evaluated to assess their suitability for CT applications. Additionally, the effects of storage conditions, preservation, and temperature variations were analyzed. As an example, a liver was segmented from a 3D CT scan, scaled to 1:2, and a negative mold was 3D printed. The mold was subsequently filled with the most suitable dough composition to replicate the organ’s anatomy and density.

RESULTS: The evaluation of the scanned ingredients and doughs demonstrated that Hounsfield unit (HU) values ranging from below -200 HU to above 1200 HU can be achieved, enabling the simulation of various human tissue densities. Based on the analysis, simple recipes are proposed to replicate radiodensities of different anatomical structures. Additionally, the results from the liver phantom confirm the feasibility of mimicking liver tissue and morphology.

CONCLUSIONS: CT phantoms with specific radiodensities, mimicking human tissues, can be created using simple dough recipes, as demonstrated with our liver CT phantom. Refrigeration or freezing extends usability for a longer time, but temperature effects must be considered to ensure accurate HU values in CT scans.

KEY POINTS: Question CT phantoms are costly and complex, limiting accessibility for customized imaging studies. Can common ingredients from the kitchen be used to create tissue-mimicking phantoms? Findings The dough-based phantoms simulated radiodensities from -200 to 1200+ HU, enabling the simulation of various human tissue densities. Clinical relevance Affordable, anatomically accurate CT phantoms can be made from kitchen ingredients for training and research, as demonstrated with our liver CT phantom.

PMID:40760118 | DOI:10.1007/s00330-025-11887-5

Eur Radiol. 2025 Aug 4. doi: 10.1007/s00330-025-11905-6. Online ahead of print.

ABSTRACT

OBJECTIVES: Magnetic resonance elastography (MRE) provides multiple quantitative biomarkers for assessing chronic liver disease. This study evaluated the prognostic value of MRE-based long-term liver stiffness (LS) changes in chronic hepatitis C virus (HCV) infection and explored short-term viscosity changes (loss modulus, LM) for monitoring responses to direct-acting antiviral treatment (DAA-Tx).

MATERIALS AND METHODS: One hundred fifty-nine adults were included. We retrospectively analyzed 138 HCV participants with at least two 2D MREs, defining progression as an LS increase ΔLS ≥ 19% above baseline (LS_B), predicting liver-related events (LREs) with Kaplan-Meier plots and Cox proportional hazards regression. We also prospectively analyzed 21 participants, 11 of whom underwent DAA-Tx and ten healthy controls. Dual-frequency 3D MRE-derived LS, LM, and damping ratio (DR) were compared at five time points during DAA-Tx.

RESULTS: Among the 138 HCVs, 38 (27.5%) were progressors, with a ΔLS = 0.393 kPa/year; 11 developed LREs, showing a steeper rise of 0.614 kPa/year. LS_B and the combination of LS_B and ΔLS were significantly associated with LREs. Specifically, each 1 kPa increase in LS_B was associated with doubled risk of developing LREs [adjusted hazard ratio (aHR) = 2.05; C-statistic = 0.831, p < 0.001]. The combination of LS_B and ΔLS further improved predictive performance (aHR = 2.92; C-statistic = 0.883, p < 0.001). Progressors had a hazard ratio (HR) of 15.44 for developing cirrhosis (p < 0.001). LS decreased post-DAA-Tx, while LM initially increased (p < 0.05) and normalized by 12 months post-DAA-Tx.

CONCLUSIONS: Two-dimensional MRE-assessed longitudinal change in LS can predict future clinical events in HCV patients, providing better risk stratification than LS alone. Multiparametric dual-frequency 3D MRE has promise for short-term changes in treatment effects.

KEY POINTS: Question Accurately predicting clinical outcomes and monitoring direct-acting antiviral (DAA) treatment efficacy in HCV patients remains an unmet need. Findings Longitudinal LS changes predict future clinical events in HCV patients. Multiparametric dual-frequency 3D MRE detects short-term inflammatory and treatment-related changes, enhancing disease monitoring. Clinical relevance This study demonstrates longitudinal MRE as a reliable follow-up tool for chronic liver disease, enabling outcome prediction, risk stratification, and treatment monitoring, advancing personalized, non-invasive management for HCV patients.

PMID:40760117 | DOI:10.1007/s00330-025-11905-6

Eur Radiol. 2025 Aug 4. doi: 10.1007/s00330-025-11854-0. Online ahead of print.

ABSTRACT

OBJECTIVES: This study examined whether CT metrics at the 12th thoracic vertebra (T12) can replace those at the 3rd lumbar vertebra (L3) for diagnosing sarcopenia, and analyzed their impact on outcomes in esophageal squamous cell carcinoma (ESCC).

MATERIALS AND METHODS: We retrospectively reviewed 405 CT scans (person-times) of 206 ESCC patients (stages II-IV) who had received chemoradiotherapy (CRT). Metrics including Skeletal Muscle Index (SMI), Skeletal Muscle Attenuation (SMA), Intramuscular Adipose Content (IMAC), and Visceral-to-Subcutaneous Adipose Tissue Ratio (VSR) were measured at both T12 and L3. These were compared with clinical outcomes.

RESULTS: Significant correlations were found between body composition parameters measured at T12 and L3, with Spearman’s coefficients of 0.62 for SMI, 0.72 for SMA, 0.59 for IMAC, and 0.46 for VSR (all p < 0.01). After adjusting for age, gender, and tumor stage, a low post-CRT T12 SMI was significantly associated with reduced overall survival (adjusted hazard ratio = 1.56, p = 0.04), which corresponded to low post-CRT L3 SMI (adjusted hazard ratio = 1.65, p = 0.04).

CONCLUSIONS: We concluded that measuring skeletal muscle at T12 can effectively diagnose sarcopenia using chest-only CT scans. Post-CRT T12 SMI may serve as a prognostic indicator for ESCC patients.

KEY POINTS: Question Assessing sarcopenia in esophageal cancer patients is crucial for prognosis, yet traditional metrics rely on the 3rd lumbar vertebra, which may not be optimal. Findings Our study demonstrates that the muscle index at the 12th thoracic vertebra correlates well with that of the 3rd lumbar vertebra, effectively predicting patient survival. Clinical relevance Evaluating sarcopenia at the 12th thoracic vertebra via chest CT offers a reliable prognostic tool for esophageal squamous cell carcinoma patients, potentially improving survival predictions after chemoradiotherapy.

PMID:40760116 | DOI:10.1007/s00330-025-11854-0

Eur Radiol. 2025 Aug 4. doi: 10.1007/s00330-025-11833-5. Online ahead of print.

ABSTRACT

OBJECTIVES: To determine whether an AI system can identify breast cancer risk in interval breast cancer (IBC) screening mammograms.

MATERIALS AND METHODS: IBC screening mammograms from a Swiss screening program were retrospectively analyzed by radiologists/an AI system. Radiologists determined whether the IBC mammogram showed human visible signs of breast cancer (potentially missed IBCs) or not (IBCs without retrospective abnormalities). The AI system provided a case score and a prognostic risk category per mammogram.

RESULTS: 119 IBC cases (mean age 57.3 (5.4)) were available with complete retrospective evaluations by radiologists/the AI system. 82 (68.9%) were classified as IBCs without retrospective abnormalities and 37 (31.1%) as potentially missed IBCs. 46.2% of all IBCs received a case score ≥ 25, 25.2% ≥ 50, and 13.4% ≥ 75. Of the 25.2% of the IBCs ≥ 50 (vs. 13.4% of a no breast cancer population), 45.2% had not been discussed during a consensus conference, reflecting 11.4% of all IBC cases. The potentially missed IBCs received significantly higher case scores and risk classifications than IBCs without retrospective abnormalities (case score mean: 54.1 vs. 23.1; high risk: 48.7% vs. 14.7%; p < 0.05). 13.4% of the IBCs without retrospective abnormalities received a case score ≥ 50, of which 62.5% had not been discussed during a consensus conference.

CONCLUSION: An AI system can identify IBC screening mammograms with a higher risk for breast cancer, particularly in potentially missed IBCs but also in some IBCs without retrospective abnormalities where radiologists did not see anything, indicating its ability to improve mammography screening quality.

KEY POINTS: Question AI presents a promising opportunity to enhance breast cancer screening in general, but evidence is missing regarding its ability to reduce interval breast cancers. Findings The AI system detected a high risk of breast cancer in most interval breast cancer screening mammograms where radiologists retrospectively detected abnormalities. Clinical relevance Utilization of an AI system in mammography screening programs can identify breast cancer risk in many interval breast cancer screening mammograms and thus potentially reduce the number of interval breast cancers.

PMID:40760115 | DOI:10.1007/s00330-025-11833-5

Sci Rep. 2025 Aug 4;15(1):28460. doi: 10.1038/s41598-025-13705-w.

ABSTRACT

Soil microbiome plays a crucial role in ecosystem; however, the responses of the soil microbiome to nonconventional irrigation water sources remain poorly understood. This study employed 16 S rRNA sequencing to investigate microbial community shifts in soil samples collected from four geographically distinct locations affected by different irrigation water sources: saline ground water affected by seawater (SW), a brackish water lake (BW), a wastewater drain (WW), and a freshwater canal that receives inflows from multiple agricultural drains (FW). Our findings revealed distinct microbial signatures shaped by water quality, with Firmicutes dominating WW soils (49.2%) due to metal resistance (DESeq2, p = 3.67 × 10^{– 4}), whereas Chloroflexi and Cyanobacteria thrived in BW environments (LEfSe, LDA > 4, p = 8.23 × 10^{– 6}), reflecting adaptations to chloride-rich conditions. FW soils enriched Acidobacteria and Verrucomicrobia, which are associated with moderate salinity and nutrient cycling, whereas SW samples harbored halotolerant Actinobacteria and Deinococcus-Thermus (DESeq2, p = 1.47x^{– 05}). Statistical analyses revealed key potential biomarkers, including Streptococcus (WW, DESeq2 p = 3.67x^{– 24}), RB41 (BW, LEfSe p = 1.62x^{– 13}), and Candidatus_Udaeobacter (SW, DESeq2 p = 1.47x^{– 05}). Physicochemical drivers such as salinity (R² =0.319, p = 0.00041) and heavy metals (Pb/Mn in WW) strongly influence community structure. Notably, WW irrigation reduced alpha diversity (Shannon index: 4.79-5.41 vs. 6.65-7.43 in FW; Kruskal-Wallis p = 0.0056), highlighting pollutant-induced stress. These findings highlight the balance between water reuse and soil health, offering a foundation for microbiome-driven bioremediation approaches in arid environments. By utilizing native, stress-resilient microbial communities, our research promotes sustainable agricultural practices in water-limited regions.

PMID:40760076 | DOI:10.1038/s41598-025-13705-w

Sci Rep. 2025 Aug 4;15(1):28334. doi: 10.1038/s41598-025-12872-0.

ABSTRACT

Malaria remains a significant global health concern, contributing to substantial morbidity and mortality worldwide. To inform efforts aimed at alleviating the global malaria burden, this study utilized spatial analysis, advanced machine learning (ML), and explainable AI (XAI) to identify high-risk areas, uncover key determinants, predict disease outcomes, and establish causal relationships. This study analyzed data from 106 countries between 2000 and 2022, sourced from the World Health Organization, World Bank and UNICEF. A high-performance ML classifier, XGBoost, combined with XAI and causal AI (CAI) techniques was employed to evaluate malaria incidence and mortality. Spatial autocorrelation analyses, such as Getis-Ord Gi* and Moran’s I, were utilized to detect significant geographical clusters and hotspots of malaria. In 2022, malaria cases reached 251.75 million, while the peak of malaria-related fatalities occurred in 2020, totaling 99,554. Nigeria recorded the highest malaria incidence (1,332.99 million), followed by the Democratic Republic of the Congo (623.16 million) and India (319.83 million). South Sudan (149,753), Zambia (143,546), and the Central African Republic (124,801) exhibited the highest malaria mortality rates. High-incidence clusters were observed in Benin, Burkina Faso, and Ghana, with substantial mortality clusters in Benin, the Central African Republic, and Liberia. The XGBoost model demonstrated the best predictive performance for malaria incidence and mortality (RMSE = 0.63, r² = 0.93, adjusted r² = 0.92, and MAE = 0.46). The XAI and CAI methodologies identified key determinants of malaria, such as access to basic sanitation, electricity availability, population growth, and the under-5 mortality rate. Our integrated framework, driven by machine learning and artificial intelligence, offers actionable insights for identifying determinants and hotspots of malaria through spatial analysis. The study advocates for the incorporation of AI-driven spatial models into national malaria surveillance systems to facilitate evidence-based and targeted interventions.

PMID:40760068 | DOI:10.1038/s41598-025-12872-0

Sci Rep. 2025 Aug 4;15(1):28374. doi: 10.1038/s41598-024-82294-x.

ABSTRACT

Mid-Upper Arm Circumference (MUAC), Mid-Upper Arm Circumference-for-Age Z-Score (MUACZ), or Weight-for-Length Z-Score (WHZ) are used to screen for acute malnutrition in children. Studies conducted in various countries, including Ethiopia, have indicated variability in the agreement between these assessments at the World Health Organization recommended cutoffs across different ethnic groups with varying body frames. The low sensitivity of MUAC at standard cutoffs has important implications for program effectiveness. Therefore, this study aimed to validate the diagnostic performance of MUAC and MUACZ in screening for acute malnutrition among children aged 6-23 months in Ethiopia. A community-based cross-sectional study was conducted on 457 randomly selected children aged 6-23 months in the Amhara Region, Ethiopia, from February 2-18, 2023. The Spearman’s rank correlation test, Cohen’s kappa statistics, and Receiver Operating Curve analysis were conducted. The optimal cutoff points for MUAC and MUACZ were determined by selecting the points that maximized the Youden index. Statistical significance was determined with a p-value < 0.05, using a 95% confidence interval. MUAC, MUACZ, and WHZ results revealed that 11.0%, 8.6%, and 13.2% of children were wasted, respectively, and the percentage of misclassification in screening acute malnutrition was approximately 16%. MUAC and MUACZ had low sensitivity but high specificity in screening for acute malnutrition. MUAC and MUACZ showed poor correlation with WHZ when screening subjects for acute malnutrition using the World Health Organization standard cutoffs. In the Receiver Operating Characteristics curve analysis, significant predictive ability was only observed with MUAC when screening global acute malnutrition cases, and it showed a poor predictive ability (AUC = 0.61, 95% CI: 0.53, 0.70) (p < 0.001). The Youden index statistics revealed that the optimal cutoff for MUAC and MUACZ to define global acute malnutrition at WHZ < -2 SD was 13.6 cm and -0.43 SD, respectively. In addition, the optimal cutoffs for diagnosing severe acute malnutrition in children with WHZ of < -3 SD were found to be 13.1 cm and -1.91 SD for MUAC and MUACZ, respectively. The optimal cutoff values also vary in sex and age categories. Both MUAC and MUACZ had poor performance in screening acute malnutrition, and a significant proportion of children were missed despite they were wasted as compared to WHZ at the standard cutoffs. The optimal cutoff levels differ for different age and sex categories. This may affect admission and discharge rates and have a significant impact on children’s health. Modifications in the standard cutoffs are needed to ensure the quality of acute malnutrition screening and treatment services.

PMID:40760063 | DOI:10.1038/s41598-024-82294-x

Support Care Cancer. 2025 Aug 4;33(8):754. doi: 10.1007/s00520-025-09801-2.

ABSTRACT

PURPOSE: Although renal cell carcinoma (RCC) presents unique nutritional challenges due to the disease itself and treatment side effects, little is known about the prevalence of nutritional issues among RCC patients in a real-world setting. This study aimed to investigate the patient-reported prevalence of nutritional issues and the response of healthcare teams to these challenges.

METHODS: A survey among RCC patients in Germany was developed in collaboration with patient organizations and included 46 questions covering demographics, nutritional issues, and cancer care experiences. It was distributed online from April to July 2022. Responses from 94 German RCC patients were analyzed using descriptive and inferential statistics.

RESULTS: Nutritional concerns were reported by 60.6% of participants, with diarrhea (23.4%), loss of appetite (21.3%), and nausea (20.2%) being the most common issues. Unintentional weight loss was reported by 49.4% of patients, but only 13.9% were referred to nutrition specialists. More than two-thirds reported a negative or extremely negative impact due to these problems on their physical condition and quality of life. Additionally, 67% of patients felt that their nutritional needs were not taken seriously by their healthcare teams. Most patients (84%) think that nutritional care should be part of routine cancer care.

CONCLUSION: The findings reveal significant gaps in the nutritional care of RCC patients. Screenings and proactive assessments do not appear to be performed as suggested by nutritional guidelines. Thus, nutritional counseling and support are obviously still not integrated into real-world comprehensive oncological care.

PMID:40760061 | DOI:10.1007/s00520-025-09801-2

Eur J Orthop Surg Traumatol. 2025 Aug 4;35(1):339. doi: 10.1007/s00590-025-04418-x.

ABSTRACT

PURPOSE: Osteoporosis is a well-recognized risk factor for complications after total knee arthroplasty (TKA). However, the effect of pre-TKA osteoporosis on healthcare utilization and patient-reported outcomes is poorly understood. Here, we characterize the association between pre-TKA osteoporosis and (1) healthcare utilization and patient-reported pain and function outcome measures; and (2) dual X-ray absorptiometry (DEXA) scan T-scores and the aforementioned outcomes.

METHODS: A prospective cohort of primary elective TKA patients between July 2015 and January 2020 was obtained (n = 6318), of which 4922 (77.9%) completed 1-year follow-up. Outcomes included healthcare utilization (prolonged length of stay (LOS) ≥ 3D, discharge disposition (DD), 90-day readmission, and 1-year reoperation) as well as Knee Injury and Osteoarthritis Outcome Score (KOOS) Pain, KOOS-function (PS) and satisfaction.

RESULTS: The prevalence of pre-TKA osteoporosis was 66.8%, of which 28.7% had a DEXA scan and 66.3% were on osteoporosis medications. Medicated osteoporotic patients were independently associated with higher odds of prolonged LOS (Odds Ratio (OR): 1.21 (95% Confidence Interval (CI) 1.02-1.43)) and non-home DD (OR:1.56 (95%CI 1.25-1.95)). Medicated and non-medicated osteoporosis patients were associated with higher odds of 90-day readmission. The odds of failing to achieve MCID or satisfaction were not associated with preoperative OP diagnosis.

CONCLUSION: Two-thirds of primary TKA recipients had osteoporosis. Among these patients, two-thirds were on medication and one-third had a DEXA scan. Osteoporotic patients are at a higher risk of 90-day hospital readmission, longer hospital stays and non-home discharge. Interestingly, osteoporosis status was not associated with failure to achieve clinically significant improvements or satisfaction at 1 year following TKA.

PMID:40760057 | DOI:10.1007/s00590-025-04418-x

Nat Hum Behav. 2025 Aug 4. doi: 10.1038/s41562-025-02273-8. Online ahead of print.

ABSTRACT

Scientific publishing is the primary means of disseminating research findings. There has been speculation about how extensively large language models (LLMs) are being used in academic writing. Here we conduct a systematic analysis across 1,121,912 preprints and published papers from January 2020 to September 2024 on arXiv, bioRxiv and Nature portfolio journals, using a population-level framework based on word frequency shifts to estimate the prevalence of LLM-modified content over time. Our findings suggest a steady increase in LLM usage, with the largest and fastest growth estimated for computer science papers (up to 22%). By comparison, mathematics papers and the Nature portfolio showed lower evidence of LLM modification (up to 9%). LLM modification estimates were higher among papers from first authors who post preprints more frequently, papers in more crowded research areas and papers of shorter lengths. Our findings suggest that LLMs are being broadly used in scientific writing.

PMID:40760036 | DOI:10.1038/s41562-025-02273-8