Frontiers in Medicine最新文献

Objectives: Undiagnosed osteoporosis before spinal surgery increases severe complication risks. This study develops the machine learning-based CT radiomics model to preoperatively screen lumbar osteoporosis.

Materials and methods: This retrospective study enrolled 166 patients undergoing concurrent dual-energy X-ray absorptiometry (DEXA), spinal CT and MRI. Vertebral data from normal and osteoporotic cases were partitioned into training and validation cohorts (8:2 ratio). A total of 851 radiomics features were extracted from lumbar spine CT scans using the 3D slicer PyRadiomics module. Feature selection employed mRMR (minimum redundancy maximum relevance) for preliminary screening followed by LASSO regression for dimensionality reduction. Four machine learning classifiers were developed: logistic regression (LR), support vector machines (SVM), XGBoost, and random forest (RF). Model performance was assessed through receiver operating characteristic (ROC) analysis with DeLong test comparisons. Clinical utility was quantified via decision curve analysis (DCA).

Results: Nine radiomic features based on spine CT images were constructed to develop the model. The radiomic-XGBoost model with the highest area under the curve (AUC) of 0.89 of the training cohort and 0.91 of the test cohort among the machine learning algorithms. The DeLong test showed that the differences between the radiomic-XGBoost, vertebral bone quality (VBQ) and Hounsfield unit (HU) models were statistically significant (p < 0.05). DCA revealed that the radiomics-based model offers a superior net benefit compared to the other two models.

Conclusion: CT-based machine learning radiomics significantly outperformed VBQ scoring and HU measurements in osteoporosis diagnostic accuracy.

Introduction: Lorazepam has been used off-label for analgosedation in pediatric intensive care units (PICU) as an alternative to midazolam. While its intermediate duration of action makes it suitable for continuous sedation, there is limited evidence to guide dosing in children. This study illustrates how pharmacokinetic modeling and extrapolation principles can be used to (1) identify regimens that maintain the desired analgosedation levels and (2) optimize the design of a prospective protocol in children requiring mechanical ventilation.

Methods: Pharmacokinetic data and COMFORT-B scores from a preliminary pilot study in six mechanically-ventilated pediatric patients (aged 0.8-4.8 years) were available for the purpose of the current investigation. A previously published population pharmacokinetic model was used to characterize the disposition of lorazepam, accounting for developmental growth and metabolic maturation in children. Parameter distributions were used as priors. Clinical trial simulations (CTS) were subsequently performed in a virtual cohort of 100 children (aged 1.0-12 years) to explore optimized dosing regimens, combining intermittent bolus dosing and continuous infusions over a 72-h period. A target concentration of 500 ng/ml was selected considering the available clinical data and literature evidence on the analgosedative effects and safety profile of lorazepam. Simulation scenarios also explored sample size and sampling time requirements for a prospective clinical trial.

Results: The pharmacokinetic model adequately described the concentration vs. time profiles, despite appreciable interindividual variability. Population estimates for clearance and volume of distribution were 0.23 L/h/kg and 2.3 L/kg, respectively. Simulation results showed that intermittent bolus dosing every 4 h, followed by continuous infusion allowed for lorazepam steady state concentrations to fluctuate around 500 ng/ml. An initial dose of 0.2 mg/kg given as bolus every 4 h over the first 24 h, followed by a similar regimen with 0.1 mg/kg over the subsequent 24 h and continuous infusion of 0.03 mg/kg/h until the end mechanical ventilation was identified as the recommended regimen to be evaluated in a prospective clinical trial.

Conclusion: Our study underscores the importance of model-based approaches to identify suitable dosing regimens to be used in children when limited pharmacokinetic and pharmacodynamic data are available. The proposed dosing regimen balances efficacy and safety data, thereby offering the foundation for the repurposing of lorazepam as an alternative, second line option for analgosedation of mechanically ventilated subjects in a pediatric intensive care unit setting.

The iStent series constitutes a range of trabecular bypass minimally invasive glaucoma surgery (MIGS) devices, which offers intraocular pressure (IOP) reduction with favourable safety profiles, in patients with open-angle glaucoma (OAG). Having undergone significant evolution since its initial US FDA-approval in 2012, successive generations address previous limitations, while enhancing IOP-lowering efficacy through device and delivery system design iterations. Longer-term and real-world iStent data demonstrate the durability of IOP- and medication-lowering outcomes with minimal complications, while preliminary studies across a wider spectrum of glaucoma subtypes and severities provide limited evidence of successful outcomes beyond mild-to-moderate OAG, both with and without concomitant cataract surgery. Aqueous humour outflow assessment and novel intra-operative techniques may further facilitate more accurate and effective iStent positioning. Despite typically higher upfront costs, results from both cost-effectiveness and patient-reported outcome studies are encouraging. Combination MIGS with the iStent, leveraging on the multiple mechanisms of actions of various procedures, may provide greater IOP-lowering efficacy without compromising safety. With expanding clinical data and progressive enhancements, iStent technology is likely to remain a key component of the evolving MIGS landscape.

Background: Inadequate intraoperative cerebral oxygen supply is one of the inciting causes of postoperative cognitive disturbances. Numerous studies have investigated the association between regional cerebral oxygen saturation (rScO₂) monitoring and postoperative cognitive dysfunction. However, results are inconsistent, owing to differences in surgery type, patient population, and monitoring protocols. Therefore, we conducted a meta-analysis to comprehensively evaluate the association between rScO₂ monitoring and the incidence of postoperative neurocognitive disorders.

Methods: A comprehensive literature search was conducted across multiple databases from their inception to June 2025 to identify randomized controlled trials (RCTs) that compared the impact of rScO₂ monitoring versus no monitoring on cognitive function. The primary outcome was the incidence of perioperative neurocognitive disorders (PNDs). Secondary outcomes were the incidences of postoperative cognitive dysfunction (POCD) and postoperative delirium (POD), as well as the economic indicators of the number needed to treat (NNT) and cost-benefit ratio (CBR).

Results: A total of 28 RCTs were included. Overall, we found that intraoperative rScO₂ monitoring significantly reduced the incidence risk of PND (relative risk [RR] = 0.47, 95% confidence interval [CI]: 0.41, 0.54), POCD (RR = 0.47, 95% CI: 0.39, 0.57), and POD (RR = 0.45, 95% CI: 0.35, 0.57). Subgroup analyses based on surgery type (cardiac, orthopedic, abdominal, and others) demonstrated consistent protective effects of monitoring. Sensitivity analyses using leave-one-out analysis, excluding Chinese-language publications, low-quality studies, and studies with a baseline rScO₂ < 80%, confirmed the robustness of results. The economic evaluation showed that rScO₂ monitoring is both clinically beneficial and cost-effective, as reflected in the low NNT values and favorable CBRs, which indicated that the cost of prevention is substantially lower than that of managing complications.

Conclusion: Intraoperative rScO₂ monitoring significantly reduces the incidence of PND, including POCD and POD. Consistent protective effects were observed across a wide range of surgery types, demonstrating its broad clinical applicability. Furthermore, its favorable cost-benefit profile demonstrated that the prevention of neurocognitive complications has a substantially lower cost than the estimated economic burden of managing these complications. Widespread adoption of rScO₂ monitoring is recommended to improve postoperative cognitive outcomes.

Background: Salmonella typically causes gastroenteritis and rarely leads to invasive infections.

Case presentation: A 14-year-old boy, without a definitive history of an unsanitary diet or open wounds, was residing in an area with a high prevalence of tuberculosis. His primary symptoms included fever, cough, lumbar pain, and weight loss. The initial pathogen test was negative. Medical imaging revealed pulmonary nodules, intervertebral space narrowing, vertebral bone destruction, and a psoas muscle abscess. Empirical antibiotic therapy and diagnostic anti-tuberculosis treatment yielded poor results. Ultimately, pathogen testing of the surgically excised lesion identified Salmonella Dublin. Antimicrobial therapy guided by susceptibility testing yielded favorable outcomes.

Conclusion: Empirical therapy is often necessary during the initial phase of treatment. However, clinicians should consider uncommon conditions and employ appropriate approaches to obtain pathogen-specific test results, which can guide targeted therapeutic strategies when the anticipated clinical outcome is suboptimal.

Brain tumors are a life-threatening condition, and their early detection is crucial for effective treatment and improved survival rates. Traditional manual evaluation techniques, such as expert radiologist assessments and visual inspections, are widely used for diagnosing brain tumors. While these methods can be highly reliable, they are often time-consuming, prone to human error, and challenging to scale for large datasets. Consequently, there is a growing demand for Computer-Aided Diagnostic (CAD) systems to overcome these limitations and deliver fast, accurate, and scalable solutions. Despite these promising advancements, the study highlights potential limitations, including susceptibility to overfitting due to the limited availability of labeled data and the need for extensive hyperparameter tuning to generalize across diverse datasets. This study proposes a scalable multi-class brain tumor classification framework optimized for small-form-factor devices. We introduced a novel, lightweight custom convolutional neural network (CNN) that maintains high classification accuracy while significantly reducing computational complexity. We evaluated the model's capacity by training and testing it on five different datasets, and it performed well on all five. We observed a significant improvement in performance with the model on larger datasets, but it struggled with smaller and imbalanced datasets. We achieved significant scores on the datasets, and we had the highest testing accuracy on Dataset-5 (99.67% training accuracy, 98.17% validation accuracy, and 98.30% testing accuracy). What is important to note is that we had the lowest testing accuracy on Dataset-3 (99.99% training accuracy, 74.11% validation accuracy, and 75.63% testing accuracy). The proposed framework leverages state-of-the-art pretrained deep learning models, including EfficientNetb3, ResNet-101, ResNet-50, Xception, AlexNet, DenseNet121, Swin Transformer, and our custom lightweight CNN model. Experimental evaluations demonstrate that EfficientNetb3 achieves the highest accuracy of 99.11%, while the custom lightweight CNN attains 98% accuracy with 4.1 × fewer parameters and reduced training time. These results highlight the effectiveness of computer-aided approaches in achieving near-expert performance, making them suitable for integration into clinical workflows. This research paves the way for deploying efficient and scalable deep learning models in real-world medical applications, thereby expanding accessibility to accurate brain tumor diagnosis.

Purpose: To explore the knowledge, attitudes, and practices (KAP) regarding chemotherapy adverse reactions and care among patients with gastrointestinal lymphoma.

Methods: This cross-sectional study was conducted between March, 2024, and May, 2024, at the Peking University Cancer Hospital Inner Mongolia Hospital, China. The participants included patients with gastrointestinal lymphoma. The KAP scores were collected using a researcher-developed questionnaire. The KAP levels were evaluated based on Bloom's cutoff value, and the associations among KAP were evaluated by logistic regression and structural equation modeling (SEM) analysis.

Results: A total of 422 patients with gastrointestinal lymphoma participated in this study. The mean scores for knowledge, attitude, and practice were 12.93 ± 4.21 (range: 0-22), 27.57 ± 3.72 (range: 7-35), and 33.57 ± 3.19 (range: 8-40), respectively. The regression analysis indicated that knowledge scores significantly influenced attitudes (OR = 1.397, P < 0.001) and practice (OR = 1.235, P < 0.001). SEM analysis revealed that knowledge significantly influences practice behaviors directly (β = 0.161, P < 0.001) and indirectly through attitudes (β = 0.649, P < 0.001).

Conclusion: Patients with gastrointestinal lymphoma demonstrated moderate knowledge, positive attitudes, and good practices regarding chemotherapy adverse reactions and care. Targeted interventions to improve knowledge, especially among rural and lower-income patients, may enhance overall attitudes and practices toward chemotherapy management.

Background: Non-invasive prenatal testing (NIPT) is widely used for screening common fetal aneuploidies such as trisomy 21 (T21), trisomy 18 (T18), and trisomy 13 (T13). However, its utility in detecting trisomy 3 (T3) has been rarely reported. Furthermore, uniparental disomy (UPD) involving chromosome 3 is a rare genetic condition with potential phenotypic consequences.

Methods: NIPT indicated a high risk for fetal T3. This finding was further investigated using copy number variation (CNV) analysis via trio-based chromosomal microarray analysis (trio-CMA). Subsequent trio-based whole-genome sequencing (trio-WGS) identified a homozygous variant in PLXNA1 associated with a putative autosomal recessive disorder in the fetus. The detected variant was validated by Sanger sequencing in the parents.

Results: NIPT revealed a fetal Z-score (27.22) for T3. Trio-CMA ruled out T3 but confirmed mixed maternal UPD3. Trio-WGS identified a homozygous PLXNA1 variant (NM_032242.3:c.2497G>C, p.Ala833Pro) in the fetus, inherited from the heterozygous mother. The observed severe fetal phenotype was partial consistent with the molecular findings of mixed UPD3 and the homozygous PLXNA1 variant, indicating that this variant may represent a potential pathogenic cause.

Conclusions: While NIPT can signal a high risk for rare aneuploidies, definitive diagnosis requires invasive prenatal testing. Discrepancies between NIPT and fetal tissue analyses may arise from confined placental mosaicism (CPM). We propose a model in which nondisjunction of chromosome 3 during germ cell formation led to trisomy, followed by a postzygotic self-correction event, resulting in mixed maternal UPD3 and increased risk of autosomal recessive disorders.

Background: Cutaneous squamous cell carcinoma (cSCC) is a common non-melanoma skin cancer with potential for local invasion and metastasis. Accurate preoperative assessment is essential for optimal treatment planning.

Materials and methods: We report a case of an 83-year-old female patient who presented with a progressively enlarging scalp mass over 3 months. HR-MRI revealed a mixed-signal lesion (22 × 15 × 26 mm) in the right scalp. On T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI), the lesion exhibited heterogeneous signals with an irregular, crater-like surface. Post-contrast imaging demonstrated significant heterogeneous enhancement. The tumor was located within the epidermis, dermis, and subcutaneous fat, closely adhering to the galea aponeurotica with well-defined margins. Surgical resection and histopathological examination confirmed a (exophytic type) moderately to well-differentiated cSCC (2.8 × 2.3 × 2.0 cm) infiltrating the subcutaneous tissue but without perineural invasion or deeper tissue involvement.

Results: HR-MRI provided clear visualization of tumor morphology, infiltration depth, and relationship with surrounding structures. Compared to conventional MRI, HR-MRI improved the accuracy of tumor boundary delineation, offering valuable information for preoperative planning.

Conclusion: HR-MRI plays a significant role in the evaluation of cSCC, particularly in assessing tumor infiltration depth and differentiating it from other cutaneous malignancies. Its high-resolution imaging facilitates early detection, precise surgical planning, and improved patient outcomes.

Anesthesiology and intensive care medicine are among the most data-rich fields of medicine, where accurate and timely outcome prediction or risk stratification is important. During patient care, heterogeneous data streams, including structured electronic health records, free-text documentation, and high-frequency physiologic time series are recorded. This provides a fertile ground for machine learning (ML) models to make individualized risk predictions. Yet, secondary use of routine data remains difficult due to heterogeneity, missingness, variable granularity, ambiguously defined outcomes, or poor representation of clinical concepts in routine data. Reproducibility and transparency are difficult to achieve with hospital-specific complex data pipelines. New complexities arise when combining different data modalities. This perspective article discusses three common modalities-tabular data, clinical text, and time series-and outlines data modality-specific challenges, data preprocessing strategies, and ML modeling approaches. We examine multimodal fusion strategies through the common taxonomy of early, intermediate, and late fusion. In early fusion, generated features are aggregated into a unified tabular representation, offering simplicity and often serve as first baseline prediction models. Intermediate fusion uses modality-specific encoders with shared layers to learn cross-modal dependencies. This strategy yields the most complex and powerful models. Late decision-level fusion combines outputs from modality-optimized models, providing modularity and robustness to missing modalities, leading to advantages for real-time deployment where data arrive asynchronously. The growth of multi-centric datasets and federated infrastructures may enable intermediate-fusion architectures and multimodal foundation models to better capture patient trajectories, supporting risk stratification and personalized therapy in perioperative and intensive care settings.