Tag: pubmed

Health Care Manag Sci. 2025 Oct 13. doi: 10.1007/s10729-025-09724-8. Online ahead of print.

ABSTRACT

In the context of diagnosis of COVID-19, this paper shows how to convert a Causal Network to a Large Language Model (LLM). The Causal Network was converted to the language model using prompts and completions. Prompts were composed from the full-factorial combination of the text associated with statistically significant variables in the Causal Network. Completions were based on the evaluation of the probability of COVID-19 using the Causal Network. The accuracy of the Causal Network and LLM was tested using two databases. The first database was based on a survey of 822 patients, collecting 12 direct (parents on the Markov blanket of COVID-19 diagnosis node), 7 indirect (associated with COVID-19 but not direct cause) symptoms of COVID-19. The second set was based on 80 patients reporting their symptoms in open-ended questions, often reporting some of the direct predictors and rarely reporting any indirect predictors of COVID-19. The accuracy of Causal Network and Markov blanket was tested using Area under the Receiver Operating Curve (AUROC). When indirect information was available, the Causal Network model (AUROC = 0.91) was significantly more accurate than the LLM (AUROC = 0.88), even though LLM model was trained to duplicate predictions of the Causal Network. Where the indirect information was not available, both models had lower accuracy (AUROC of 0.75 and 0.76). The accuracy of LLM depends not only on patterns among direct predictors of the outcome but also data not reported to the LLM. Conversational LLMs need to go beyond information the patient supplies and proactively ask about missing, typically indirect, information.

PMID:41082130 | DOI:10.1007/s10729-025-09724-8

CVIR Endovasc. 2025 Oct 13;8(1):81. doi: 10.1186/s42155-025-00609-z.

ABSTRACT

PURPOSE: This study evaluates four large language models’ (LLMs) ability to answer common patient questions preceding transarterial periarticular embolization (TAPE), computed tomography (CT)-guided high-dose-rate (HDR) brachytherapy, and bleomycin electrosclerotherapy (BEST). The goal is to evaluate their potential to enhance clinical workflows and patient comprehension, while also assessing associated risks.

MATERIALS AND METHODS: Thirty-five TAPE, 34 CT-HDR brachytherapy, and 36 BEST related questions were presented to ChatGPT-4o, DeepSeek-V3, OpenBioLLM-8b, and BioMistral-7b. The LLM-generated responses were independently assessed by two board-certified radiologists. Accuracy was rated on a 5-point Likert scale. Statistics compared LLM performance across question categories for patient-education suitability.

RESULTS: DeepSeek-V3 attained the highest mean scores for BEST [4.49 (± 0.77)] and CT-HDR [4.24 (± 0.81)] and demonstrated comparable performance to ChatGPT-4o for TAPE-related questions (DeepSeek-V3 [4.20 (± 0.77)] vs. ChatGPT-4o [4.17 (± 0.64)]; p = 1.000). In contrast, OpenBioLLM-8b (BEST 3.51 (± 1.15), CT-HDR 3.32 (± 1.13), TAPE 3.34 (± 1.16)) and BioMistral-7b (BEST 2.92 (± 1.35), CT-HDR 3.03 (± 1.06), TAPE 3.33 (± 1.28)) performed significantly worse than DeepSeek-V3 and ChatGPT-4o across all procedures. Preparation/Planning was the only category without statistically significant differences across all three procedures.

CONCLUSION: DeepSeek-V3 and ChatGPT-4o excelled on TAPE, BEST, and CT-HDR brachytherapy questions, indicating potential to enhance patient education in interventional radiology, where complex but minimally invasive procedures often are explained in brief consultations. However, OpenBioLLM-8b and BioMistral-7b exhibited more frequent inaccuracies, suggesting that LLMs cannot replace comprehensive clinical consultations yet. Patient feedback and clinical workflow implementation should validate these findings.

PMID:41082087 | DOI:10.1186/s42155-025-00609-z

Biochem Genet. 2025 Oct 13. doi: 10.1007/s10528-025-11255-4. Online ahead of print.

ABSTRACT

Schizophrenia (SCZ) is a deleterious neuropsychological disorder with a worldwide incidence of 1% and unknown etiology. Understanding the role of genetic variants in disease development would enable us to explain the disorder’s molecular mechanism and find a more effective prognostic approach. Several studies in various European and East Asian populations have displayed the association of schizophrenia with functional polymorphisms such as rs16944 and rs1143634 that lie within inflammatory pathway genes. This study aimed to evaluate the association of Interleukin-1 beta (IL1B) variants (rs16944, rs1143634) with schizophrenia in the Iranian population for the first time. 565 individuals (240 cases and 325 controls) were recruited. Genotyping was conducted for IL1B single nucleotide polymorphisms (SNPs) (rs16944 and rs1143634) using polymerase chain reaction-Restriction fragment length polymorphism (PCR-RFLP). In addition, the haplotype analysis was conducted. All statistical analysis was performed using SPSS version 26. Regarding rs1143634 (C > T), T carrier genotypes (CT, TT) compared to CC homozygous genotypes showed a significantly more protective effect (p-value < 0.001). Similarly, concerning the co-dominant model, CT heterozygous genotypes in comparison with homozygous genotypes (CC, TT) illustrated a protective impact regarding schizophrenia (p-value < 0.001). Findings showed a significant difference between cases and healthy controls regarding the rs1143634 (C > T) allele frequency (p-value = 0.025), whereas it determined no considerable difference given rs16944 (p-value = 0.41). Furthermore, in the case of rs16944 (T > C), we found no significant association between case and control groups (p-value = 0.69). Haplotype analysis demonstrated that the C-C (rs1143634-rs16944) haplotype was significantly associated with the risk of schizophrenia (p-value = 0.013). The findings suggest that IL1B rs1143634 (C > T) is significantly associated with SCZ in the Iranian population.

PMID:41082027 | DOI:10.1007/s10528-025-11255-4

EJNMMI Res. 2025 Oct 13;15(1):130. doi: 10.1186/s13550-025-01326-3.

ABSTRACT

BACKGROUND: Terbium-161 (¹⁶¹Tb) is a promising β⁻-emitter for theragnostics. However, its complex photon emission pattern-including intense X-rays and low-yield, high-energy γ-emissions-may complicate image-based quantification. This study aimed to assess the feasibility of accurate SPECT/CT-based ¹⁶¹Tb dosimetry through a series of phantom measurements using a GE Discovery NM/CT 670 Pro system. Three collimators were evaluated: extended low-energy general-purpose (ELEGP), low-energy high-resolution (LEHR), and medium-energy general-purpose (MEGP), using two separate energy windows: around the 75 keV γ-peak (± 10%), and around the 49 keV γ-peak and nearby X-rays (40.7-62.9 keV). A clinical OSEM reconstruction algorithm was employed.

RESULTS: On average, the SPECT calibration factors (CFs) were 2-fold higher with ELEGP compared to MEGP and LEHR, and 3-fold higher at 49 keV compared to 75 keV. For each collimator, derived CFs varied substantially depending on measurement and volume-of-interest geometry-more so at 49 keV, compared to 75 keV. Measurements of two 3D-printed kidney inserts revealed superior visual image quality with LEHR compared to ELEGP and MEGP. Across all collimators, the 75 keV window provided better spatial resolution and contrast than the 49 keV window. An anthropomorphic phantom study, including a LungSpine phantom with 8 spherical inserts and 3 different background activity levels, demonstrated a greater quantitative accuracy for MEGP compared to LEHR and ELEGP, with statistical significance for both energy windows (p ≤ 0.001). Errors were generally larger at 49 keV compared to 75 keV. For the low-energy collimators, considerable septal penetration (e.g., at 292 and 475 keV) was observed, along with systematic underestimation at high activity levels.

CONCLUSIONS: This study demonstrates that highly accurate SPECT/CT-based ¹⁶¹Tb quantification is feasible, further cementing ¹⁶¹Tb as a viable theragnostic alternative. A MEGP collimator, a 75 keV window, and a CF derived from a homogeneous cylinder measurement appears preferable. The 49 keV window could be useful at late imaging time points, given its high sensitivity, if further optimized. Degradation from penetration and subsequent downscatter may be mitigated with a more refined reconstruction. Further investigations into dead-time effects are encouraged.

PMID:41082018 | DOI:10.1186/s13550-025-01326-3

Biomech Model Mechanobiol. 2025 Oct 13. doi: 10.1007/s10237-025-02016-8. Online ahead of print.

ABSTRACT

Computed tomography (CT)-based finite element (FE) models can non-invasively assess bone mechanical properties, but their clinical application in paediatrics is limited due to fewer datasets and models. Statistical Shape-Density Model (SSDM)-based FE models using statistically inferred shape and density have application to predict bone stress and strains; however, their accuracy in children remains unexplored. This study assessed the accuracy of stress-strain distributions estimated from SSDM-based FE models of paediatric femora and tibiae. CT-based FE models used geometry and densities derived from 330 CT scans from children aged 4-18 years. Paediatric SSDMs of the femur and tibia were used to predict bone geometries and densities from participants’ demographics and linear bone measurements. Forces during single-leg standing were estimated and applied to each bone. Stress and strain distributions were compared between the SSDM-based FE models and CT-based FE models, which served as the gold standard. The average normalized root-mean-square error (NRMSE) for Von Mises stress was 6% for the femur and 8% for the tibia across all cases. Principal strains NRMSE ranged from 1.2% to 5.5%. High correlations between the SSDM-based and CT-based FE models were observed, with determination coefficients ranging from 0.80 to 0.96. These results illustrate the potential of SSDM-based FE models for paediatric application, such as personalized implant design and surgical planning.

PMID:41082014 | DOI:10.1007/s10237-025-02016-8

Bull Math Biol. 2025 Oct 13;87(11):158. doi: 10.1007/s11538-025-01541-y.

ABSTRACT

The coexistence of species in predator-prey systems is a critical ecological issue due to the intricate interactions among multiple influencing factors. In this study, we develop a predator-prey model that incorporates prey herd behavior, cooperative hunting strategies among predators, and the establishment of a reserved area for prey protection. We establish conditions for the positivity and boundedness of the system to ensure long-term biological feasibility. The existence and stability of equilibrium points, along with the conditions for Hopf and saddle-node bifurcations, are rigorously analyzed. Numerical simulations are performed to validate the analytical findings. Global sensitivity analysis reveals that key parameters, including the size of the reserved area, predator cooperation, and migration rates, significantly affect system dynamics and species coexistence. Our numerical results suggest that expanding the reserved area promotes prey recovery, with predator populations initially growing but eventually declining towards extinction. Increased hunting cooperation among predators initially boosts predator populations but ultimately accelerates prey depletion, leading to predator collapse due to overhunting.

PMID:41082013 | DOI:10.1007/s11538-025-01541-y

Eur J Clin Microbiol Infect Dis. 2025 Oct 13. doi: 10.1007/s10096-025-05317-z. Online ahead of print.

ABSTRACT

PURPOSE: Artificial Intelligence (AI) and Machine Learning (ML) are innovative technologies, gathering particular interest in all laboratory settings. The aim of this study was to assess the current state of AI and ML adoption in Italian laboratories of clinical microbiology.

METHODS: A structured 63-question survey was developed by the AI/ML in Microbiology Study Group and distributed to all members of the Italian Association of Clinical Microbiologists between December 2024 and March 2025. Responses were collected anonymously and analyzed using descriptive statistics analysis.

RESULTS: A total of 163 professionals completed the survey. While 25.4% reported current AI/ML usage-primarily in bacteriology and virology-the majority had limited experience with AI technologies. Only 13.6% of respondents had a good understanding of AI/ML concepts, and 2.5% reported having trained data scientists on staff. Major barriers included lack of trained personnel and insufficient infrastructure. Most participants (99.0%) expressed interest in targeted AI training, and 57.5% showed willingness to collaborate on AI-related initiatives. Large language models (LLMs) were seen as promising, especially for data interpretation, despite low adoption rates.

CONCLUSION: The survey might provide valuable insights to identify priority areas for intervention, guide future training initiatives and develop targeted strategies to promote the adoption of these technologies through a fruitful dialogue with companies and IT professionals.

PMID:41081991 | DOI:10.1007/s10096-025-05317-z

Eur J Clin Microbiol Infect Dis. 2025 Oct 13. doi: 10.1007/s10096-025-05278-3. Online ahead of print.

ABSTRACT

BACKGROUND: Recurrent Clostridioides difficile infection (CDI) is a persistent clinical challenge due to the high rate of relapse following treatment with standard antibiotics. Fecal microbiota transplantation (FMT) has emerged as a promising alternative, with comparable results. Aiming to restore intestinal microbial balance and reduce recurrence. Comparative evidence on the efficacy and safety of FMT versus antibiotics remains variable across studies, warranting a comprehensive synthesis to guide clinical decision-making.

AIM: This systematic review and meta-analysis aims to present an updated comparison of the effectiveness and safety of FMT versus Vancomycin/ fidaxomicin in patients with CDI.

METHODS: A comprehensive search of PubMed, Embase, and the Cochrane Library was conducted to identify randomized controlled trials comparing FMT with standard antibiotic therapy for recurrent CDI. Primary outcomes included resolution of infection, recurrence, mortality, and adverse events. A random-effects model was used to calculate risk ratios with 95% confidence intervals. Statistical heterogeneity was assessed using the I-squared statistic. The quality of the included studies was evaluated using the Cochrane Risk of Bias version 2 and ROBINS-1 tools.

RESULTS: A total of 9 clinical trials involving 759 patients were included. FMT was significantly more effective in resolving CDI compared to antibiotic therapy, with a risk ratio (RR) of 1.51 (95% CI: 1.29 to 1.78). Recurrence rates were significantly lower in the FMT group, with a RR of 0.38 (95% CI: 0.29 to 0.50). Mortality did not differ significantly between groups (RR = 0.95). Adverse events (AEs) were comparable between FMT and antibiotics, and no serious AEs directly related to FMT were reported. In the subgroup analysis, the lower GI route adminstration showed significant results (p = 0.02) for both recurrence and resolution of CDI.

CONCLUSION: FMT is more effective than standard antibiotic therapy for achieving resolution and reducing recurrence in patients with recurrent CDI.

PMID:41081988 | DOI:10.1007/s10096-025-05278-3

Sleep Breath. 2025 Oct 13;29(6):317. doi: 10.1007/s11325-025-03490-1.

ABSTRACT

OBJECTIVE: The study aims to assess the effect of a continuous positive airway pressure (CPAP) trial before titration on acceptability, CPAP pressures and sleep efficiency, and to predict long-term adherence to CPAP titration. The study assesses the treatment response and determines the optimum pressures.

METHODS: Prospective, randomised, single-centre, Cohort Study conducted in a tertiary care hospital in India. Patients of Indian origin with moderate to severe OSA diagnosed by level 1 polysomnography in our hospital were included in the study. Patients with haemodynamic instability and psychiatric disorders, and those who refused to consent to the study, were excluded from the study. Patients randomised into two groups were assessed for acceptability to CPAP following the titration study, and other parameters like sleep efficiency and mean CPAP pressures were noted. Anxiety is assessed using the Beck Anxiety Inventory at baseline and at three months. Compliance with CPAP is evaluated at three months using a proforma with the data obtained from the CPAP apparatus.

RESULTS: The acceptability of CPAP was better in the interval group. The results showed a statistically significant negative correlation between acceptability and average use of CPAP (r=-0.382, p-value < 0.001) and with the number of nights CPAP was used last month for > 4 h (r=-0.514, p-value < 0.001). The sleep efficiency at titration was better in the interval group.

CONCLUSION: A trial of CPAP before a titration study could improve acceptability, lower the CPAP pressures and improve long-term adherence in patients with moderate to severe OSA.

PMID:41081982 | DOI:10.1007/s11325-025-03490-1

J Robot Surg. 2025 Oct 13;19(1):684. doi: 10.1007/s11701-025-02878-1.

ABSTRACT

Enhanced recovery after surgery (ERAS) protocols emphasize multimodal interventions to accelerate recovery. Minimally invasive elective robotic-assisted colorectal surgery is an ideal candidate for ERAS protocols. This study evaluates the role of intrathecal analgesia with diamorphine in optimizing postoperative outcomes, focusing on pain control, length of stay (LOS), and complications. A retrospective cohort study was conducted on 246 patients undergoing elective robotic-assisted colorectal surgeries between July 2021 and November 2024 at two hospitals in a single trust (Hospitals A and B). The patients were grouped based on analgesia type: intrathecal analgesia with diamorphine (IA) [Intervention] and non-intrathecal analgesia (nIA) [control] groups. The primary outcomes included LOS, postoperative pain scores, and opioid consumption. The secondary outcomes were postoperative ileus (POI) and time to first bowel movement. Intrathecal analgesia with diamorphine was used in 61.8% (Hospital A) and 0% (Hospital B) of cases. Patients receiving IA in Hospital A compared to nIA in Hospital A and Hospital B, respectively, had significantly shorter LOS (5.3 days vs. 6.1 days and 7.0 days, p < 0.001), reduced opioid requirements (88% reduction, p < 0.00001), and a lower incidence of POI (3.6% vs. 13.2% and 14.7%, p = 0.02). Time to bowel function recovery was also faster in the IA group (2.6 days vs. 3.0 days in other groups, p = 0.03). Intrathecal analgesia with diamorphine significantly enhances recovery in elective robotic-assisted colorectal surgery, aligning with ERAS objectives. Its integration into routine practice could optimize patient outcomes, reduce complications, and improve resource utilization. Further prospective studies are needed to confirm these findings.

PMID:41081975 | DOI:10.1007/s11701-025-02878-1