Search Results (469)

Early detection of clinical deterioration in emergency department (ED) patients is critical for timely interventions. This study evaluated the predictive performance of the National Early Warning Score (NEWS) parameters using machine learning. We conducted a single-center retrospective observational study including 27,238 adult ED patients admitted to Merano Hospital (Italy) between June 2022 and June 2023. NEWS vital signs were collected at triage, and mortality at 48 h, 7 days, and 30 days was obtained from ED database. Decision tree analysis (CHAID algorithm) was used to identify predictors of mortality; 10-fold cross-validation was applied to avoid overfitting. Mortality was 0.4% at 48 h, 1% at 7 days, and 2.45% at 30 days. For 48-h mortality, oxygen supplementation (FiO₂ >21%) and AVPU = “U” were the strongest predictors, with a maximum risk of 31.6%. For 7-day mortality, SpO₂ was the key predictor, with mortality up to 48.1%. At 30 days, patients with AVPU ≠ A, FiO₂ > 21%, and SpO₂ ≤ 94% had a mortality risk of 66.7%. Decision trees revealed different cut-offs compared to the standard NEWS. This study demonstrated that for ED patients, the NEWS may require some adjustments in both the cut-offs for vital parameters and the methods of collecting these parameters. Full article

Background: Comprehensive Geriatric Assessment (CGA) is essential for evaluating older cancer patients, but significant gaps persist in both research and clinical practice. This study aimed (I) to identify the CGA elements that most influence anti-cancer treatment decisions in older patients and (II) to explore the predictive value of CGA components for mortality. Methods: This observational study included older patients with newly diagnosed, histologically confirmed solid or hematological cancers, recruited consecutively from 2003 to 2023. Participants were followed for four years. The data collected included CGA measures of functional (Activities of Daily Living-ADL), cognitive (Mini-Mental State Examination-MMSE), and emotional (Geriatric Depression Scale-GDS) domains. Patients were categorized into frail, vulnerable, or fit groups based on Balducci’s criteria. Statistical analyses included decision tree modeling and Cox regression to identify predictors of mortality. Results: A total of 7022 patients (3222 females) were included, with a mean age of 78.3 ± 12.9 years. The key CGA factors influencing treatment decisions were ADL (first step), cohabitation status (second step), and age (last step). After four years, 21.9% patients had died. Higher GDS scores (OR 1.04, 95% CI 1.01–1.07, p = 0.04) were independently associated with survival in men and living with family members (OR 1.67, 95% CI 1.35–2.07, p < 0.001) in women. Younger patients (<77 years) showed both MMSE and GDS as significant risk factors for mortality. Conclusions: Functional capacity, cohabitation status, and GDS scores are crucial for guiding treatment decisions and predicting mortality in older cancer patients, emphasizing the need for a multidimensional geriatric assessment. Full article

Background/Objectives: This study aims to present a structured clinical workflow for offline adaptive Biology-guided Radiotherapy (BgRT) using the RefleXion X1 PET-linac system, addressing challenges introduced by inter-treatment anatomical and biological changes. Methods: We propose a decision tree offline adaptation framework based on real-time assessments of Activity Concentration (AC), Normalized Target Signal (NTS), and bounded dose-volume histogram (bDVH%) metrics. Three offline strategies were developed: (1) preemptive adaptation for minor changes, (2) partial re-simulation for moderate changes, and (3) full re-simulation for major anatomical or metabolic alterations. Two clinical cases demonstrating strategies 1 and 2 are presented. Results: The preemptive adaptation strategy was applied in a case with early tumor shrinkage, maintaining delivery parameters within acceptable limits while updating contours and dose distribution. In the partial re-Simulation case, significant changes in PET signal necessitated a same-day PET functional modeling session and plan re-optimization, effectively restoring safe deliverability. Both cases showed reduced target volumes and improved OAR sparing without additional patient visits or tracer injections. Conclusions: Offline adaptive workflows for BgRT provide practical solutions to address inter-fractional changes in tumor structure and function. These strategies can help maintain the safety and accuracy of BgRT delivery and support clinical adoption of PET-guided radiotherapy, paving the way for future online adaptive capabilities. Full article

Background/Objectives: This study shows a lightweight hybrid framework based on a feedforward neural network using a single frontopolar electroencephalography channel, which is a practical configuration for wearable systems combining memory and a sleep stage transition probability matrix. Methods: Motivated by autocorrelation analysis, revealing strong temporal dependencies across sleep stages, we incorporate prior epoch information as additional features. To capture temporal context without requiring long input sequences, we introduce a transition-aware feature derived from the softmax output of the previous epoch, weighted by a learned stage transition matrix. The model combines predictions from memory-based and no-memory networks using a confidence-driven fallback strategy. Results: The proposed model achieves up to 85.4% accuracy and 0.79 Cohen’s kappa, despite using only a single 30 s epoch per prediction. Compared to other models that use a single frontopolar channel, our method outperforms convolutional neural networks, recurrent neural networks, and decision tree approaches. Additionally, confidence-based rejection of low-certainty predictions enhances reliability, since most of the epochs with low confidence in the sleep stage classification contain transitions between sleep stages. Conclusions: These results demonstrate that the proposed method balances performance, interpretability, and computational efficiency, making it well-suited for real-time clinical and wearable sleep staging applications using battery-powered computing devices. Full article

Background: Staphylococcus aureus is a significant pathogen causing both local and systemic infections in children, with deep tissue involvement leading to severe complications. This study aimed to assess clinical manifestations and identify risk factors for deep tissue involvement in pediatric S. aureus infections. Methods: All children between 1 month and 18 years who had S. aureus growth in blood, pus, or joint fluid culture were included. Results: A total of 61 patients (median age 55 months) were included, with 22.9% having deep tissue infections. Osteoarticular infections, pyomyositis, and pulmonary involvement were common. Deep-seated infections were significantly associated with community-acquired infections and positive hemocultures after 72 h (p < 0.01). Laboratory results showed significantly higher levels of C-reactive protein, sedimentation rate, D-dimer, and fibrinogen in the group with deep-seated infections (p = 0.02, p = 0.018, p = 0.01, and p = 0.015, respectively). The decision tree model showed that the first indicator of deep-seated infection was a D-dimer level above 1.15 mg/L, followed by a fibrinogen level above 334 mg/dL. Conclusions: Deep-seated S. aureus infections are more frequently associated with community-acquired cases, persistent hemoculture positivity, and methicillin-susceptible Staphylococcus aureus (MSSA) strains. Additionally, elevated D-dimer and fibrinogen levels may serve as valuable markers for identifying deep-seated infections in pediatric patients. Full article

Background: Sarcopenia is a progressive muscle disease that compromises mobility and quality of life in older adults. Although dual-energy X-ray absorptiometry (DXA) is the standard for assessing Appendicular Lean Mass Index (ALMI), it is costly and often inaccessible. This study aims to develop machine learning models using anthropometric measurements to predict low ALMI for the diagnosis of sarcopenia. Methods: A cross-sectional study was conducted on 183 Mexican adults (67.2% women and 32.8% men, ≥60 years old). ALMI was measured using DXA, and anthropometric data were collected following the International Society for the Advancement of Kinanthropometry (ISAK) protocols. Predictive models were developed using Logistic Regression (LR), Decision Trees (DTs), Random Forests (RFs), Artificial Neural Networks (ANNs), and LASSO regression. The dataset was split into training (70%) and testing (30%) sets. Model performance was evaluated using classification performance metrics and the area under the ROC curve (AUC). Results: ALMI indicated strong correlations with BMI, corrected calf girth, and arm relaxed girth. Among models, DT achieved the best performance in females (AUC = 0.84), and ANN indicated the highest AUC in males (0.92). Regarding the prediction of low ALMI, specificity values were highest in DT for females (100%), while RF performed best in males (92%). The key predictive variables varied depending on sex, with BMI and calf girth being the most relevant for females and arm girth for males. Conclusions: Anthropometry combined with machine learning provides an accurate, low-cost approach for identifying low ALMI in older adults. This method could facilitate sarcopenia screening in clinical settings with limited access to advanced diagnostic tools. Full article

Background: Acute complicated cholecystitis (ACC) is associated with prolonged hospitalization, increased morbidity, and higher mortality. However, objective imaging-based criteria to guide early clinical decision-making remain limited. This study aimed to develop a predictive scoring system integrating clinical characteristics, laboratory biomarkers, and computed tomography (CT) findings to facilitate the early identification of ACC in the emergency department (ED). Methods: We conducted a retrospective study at an urban tertiary care center in Taiwan, screening 729 patients who presented to the ED with suspected cholecystitis between 1 January 2018 and 31 December 2020. Eligible patients included adults (≥18 years) with a confirmed diagnosis of acute cholecystitis based on the Tokyo Guidelines 2018 (TG18) and who were subsequently admitted for further management. Exclusion criteria included (a) the absence of contrast-enhanced CT imaging, (b) no hospital admission, (c) alternative final diagnosis, and (d) incomplete clinical data. A total of 390 patients met the inclusion criteria. Demographic data, laboratory results, and CT imaging features were analyzed. Logistic regression and decision tree analyses were used to construct predictive models. Results: Among the 390 included patients, 170 had mild, 170 had moderate, and 50 had severe cholecystitis. Key predictors of ACC included gangrenous changes, gallbladder wall attenuation > 80 Hounsfield units, CRP > 3 mg/dL, and WBC > 11,000/μL. A novel scoring system incorporating these variables demonstrated good diagnostic performance, with an area under the curve (AUC) of 0.775 and an optimal cutoff score of ≥2 points. Decision tree analysis similarly identified these four predictors as critical determinants in stratifying disease severity. Conclusions: This CT- and biomarker-based scoring system, alongside a decision tree model, provides a practical and robust tool for the early identification of complicated cholecystitis in the ED. Its implementation may enhance diagnostic accuracy and support timely clinical intervention. Full article

Background/Objective: This study aims to evaluate the predictive performance of three supervised machine learning algorithms—decision tree (DT), support vector machine (SVM), and k-nearest neighbors (KNN) in forecasting key visual skills relevant to rhythmic gymnastics. Methods: A total of 383 rhythmic gymnasts aged 4 to 27 years were evaluated in various sports centers across Madrid, Spain. Visual assessments included clinical tests (near convergence point accommodative facility, reaction time, and hand–eye coordination) and eye-tracking tasks (fixation stability, saccades, smooth pursuits, and visual acuity) using the DIVE (Devices for an Integral Visual Examination) system. The dataset was split into training (70%) and testing (30%) subsets. Each algorithm was trained to classify visual performance, and predictive performance was assessed using accuracy and macro F1-score metrics. Results: The decision tree model demonstrated the highest performance, achieving an average accuracy of 92.79% and a macro F1-score of 0.9276. In comparison, the SVM and KNN models showed lower accuracies (71.17% and 78.38%, respectively) and greater difficulty in correctly classifying positive cases. Notably, the DT model outperformed the others in predicting fixation stability and accommodative facility, particularly in short-duration fixation tasks. Conclusion: The decision tree algorithm achieved the highest performance in predicting short-term fixation stability, but its effectiveness was limited in tasks involving accommodative facility, where other models such as SVM and KNN outperformed it in specific metrics. These findings support the integration of machine learning in sports vision screening and suggest that predictive modeling can inform individualized training and performance optimization in visually demanding sports such as rhythmic gymnastics. Full article

Long-COVID (LC) is characterized by diverse and persistent symptoms, potentially mirroring different molecular pathways. Recent data might offer that one of them is mediated by functional autoantibodies (fAAb) targeting G protein-coupled receptors (GPCR). Thus, the aim of this study was to investigate the clinical phenotype of patients with LC in relation to their GPCR-fAAb seropositivity. The present study recruited 194 patients with LC and profiled them based on self-reported symptoms. GPCR-fAAb seropositivity was identified by using a cardiomyocyte bioassay, testing the presence and functionality of the AAbs. Logistic regression, clustering, and decision tree analyses were applied to examine associations between GPCR-fAAb profiles and self-reported symptoms considering age and gender. The most prevalent GPCR-fAAbs in patients with LC were fAAB targeting the β2 adrenergic receptor (β2-fAAb, 92.8%), the muscarinergic M2 receptor (M2-fAAb, 87.1%), the Angiotensin II type 1 receptor (AT1-fAAb, 85.6%), and angiotensin (1–7) Mas receptor (MAS-fAAb, 85.6%). β2-fAAb showed a significant relation with dizziness, lack of concentration, and POTS, while Endothelin Type A receptor functional autoantibody (ET-A-fAAb) was significantly related to deterioration of pre-existing neurological disorders. Statistical analysis indicated a strong positive correlation between M2- and β2-fAAb; as in addition, an association of β2-fAAb and gender was observed to one of the major clinical symptoms (fatigue/PEM), a critical impact of GPCR-fAAb on LC-pathogenesis can be assumed. Summing up, the present data show that specific GPCR-fAAb are associated with distinct clinical phenotypes. Especially, the combination of M2- and β2-fAAb seemed to be essential for the LC-phenotype with a combination of fatigue/PEM and lack of concentration as major clinical symptoms. Full article

Background and Objectives: Early detection of cognitive decline (CD) is crucial for managing dementia risk factors and preventing disease progression. This study pursues two main objectives: (1) to review existing cognitive screening practices implemented in community pharmacy settings and (2) to characterize the cognitive profile of individuals eligible for screening in this context. Materials and Methods: This study was conducted in two phases. First, a scoping review of cognitive screening tools used in community pharmacies was carried out following PRISMA-ScR guidelines. Second, a cross-sectional study was performed to design and implement a CD screening protocol, assessing cognitive function. Data collection included demographic and clinical variables commonly associated with dementia risk. Decision tree analysis was applied to identify key variables contributing to the cognitive profile of patients eligible for screening. Results: The scoping review revealed that screening approaches differed by country and population, with limited pharmacy involvement suggesting implementation barriers. Cognitive screening was conducted in 18 pharmacies in Valencia, Spain (1.45%), involving 286 regular users reporting Subjective Memory Complaints (SMC). The average age of participants was 71 years, and 74.8% were women. According to the unbiased Gini impurity index, the most relevant predictors of CD—based on the corrected mean decrease in corrected impurity (MDcI), a bias-adjusted measure of variable importance—were age (MDcI: 2.60), internet and social media use (MDcI: 2.43), sleep patterns (MDcI: 1.83), and educational attainment (MDcI: 0.96). Simple decision trees can reduce the need for full screening by 53.6% while maintaining an average sensitivity of 0.707. These factors are essential for defining the profile of individuals who would benefit most from CD screening services. Conclusions: Community pharmacy-based detection of CD shows potential, though its implementation remains limited by issues of consistency and feasibility. Enhancing early dementia detection in primary care settings may be achieved by prioritizing individuals with limited internet and social media use, irregular sleep patterns, and lower education levels. Targeting these groups could significantly improve the effectiveness of CD screening programs. Full article

Background: Parkinson’s disease (PD) is the second most common neurodegenerative disorder after Alzheimer’s disease, affecting countless individuals worldwide. PD is characterized by the onset of a marked motor symptomatology in association with several non-motor manifestations. The clinical phase of the disease is usually preceded by a long prodromal phase, devoid of overt motor symptomatology but often showing some conditions such as sleep disturbance, constipation, anosmia, and phonatory changes. To date, speech analysis appears to be a promising digital biomarker to anticipate even 10 years before the onset of clinical PD, as well serving as a useful prognostic tool for patient follow-up. That is why, the voice can be nominated as the non-invasive method to detect PD from healthy subjects (HS). Methods: Our study was based on cross-sectional study to analysis voice impairment. A dataset comprising 81 voice samples (41 from healthy individuals and 40 from PD patients) was utilized to train and evaluate common machine learning (ML) models using various types of features, including long-term (jitter, shimmer, and cepstral peak prominence (CPP)), short-term features (Mel-frequency cepstral coefficient (MFCC)), and non-standard measurements (pitch period entropy (PPE) and recurrence period density entropy (RPDE)). The study adopted multiple machine learning (ML) algorithms, including random forest (RF), K-nearest neighbors (KNN), decision tree (DT), naïve Bayes (NB), support vector machines (SVM), and logistic regression (LR). Cross-validation technique was applied to ensure the reliability of performance metrics on train and test subsets. These metrics (accuracy, recall, and precision), help determine the most effective models for distinguishing PD from healthy subjects. Result: Among all the algorithms used in this research, random forest (RF) was the best-performing model, achieving an accuracy of 82.72% with a ROC-AUC score of 89.65%. Although other models, such as support vector machine (SVM), could be considered with an accuracy of 75.29% and a ROC-AUC score of 82.63%, RF was by far the best one when evaluated across all metrics. The K-nearest neighbor (KNN) and decision tree (DT) performed the worst. Notably, by combining a comprehensive set of long-term, short-term, and non-standard acoustic features, unlike previous studies that typically focused on only a subset, our study achieved higher predictive performance, offering a more robust model for early PD detection. Conclusions: This study highlights the potential of combining advanced acoustic analysis with ML algorithms to develop non-invasive and reliable tools for early PD detection, offering substantial benefits for the healthcare sector. Full article

Interest in explainable machine learning has grown, particularly in healthcare, where transparency and trust are essential. We developed a semi-automated evaluation framework within a clinical decision support system (CDSS-EQCM) that integrates LIME and SHAP explanations with multi-criteria decision-making (TOPSIS and Borda count) to rank model interpretability. After two-phase preprocessing of 2934 COVID-19 patient records spanning four epidemic waves, we applied five classifiers (Random Forest, Decision Tree, Logistic Regression, k-NN, SVM). Five infectious disease physicians used a Streamlit interface to generate patient-specific explanations and rate models on accuracy, separability, stability, response time, understandability, and user experience. Random Forest combined with SHAP consistently achieved the highest rankings in Borda count. Clinicians reported reduced evaluation time, enhanced explanation clarity, and increased confidence in model outputs. These results demonstrate that CDSS-EQCM can effectively streamline interpretability assessment and support clinician decision-making in medical diagnostics. Future work will focus on deeper electronic medical record integration and interactive parameter tuning to further enhance real-time diagnostic support. Full article

Background/Objectives: This study evaluates whether combining ⁶⁸Ga-PSMA-11-PET/CT derived imaging biomarkers with clinical risk factors improves the prediction of early biochemical recurrence (eBCR) or clinical progress in patients with high-risk prostate cancer (PCa) after primary treatment, using machine learning (ML) models. Methods: We analyzed data from 93 high-risk PCa patients who underwent ⁶⁸Ga-PSMA-11 PET/CT and received primary treatment at a single center. Two predictive models were developed: a logistic regression (LR) model and an ML derived probabilistic graphical model (PGM) based on a naïve Bayes framework. Both models were compared against each other and against the CAPRA risk score. The models’ input variables were selected based on statistical analysis and domain expertise including a literature review and expert input. A decision tree was derived from the PGM to translate its probabilistic reasoning into a transparent classifier. Results: The five key input variables were as follows: binarized CAPRA score, maximal intraprostatic PSMA uptake intensity (SUVmax), presence of bone metastases, nodal involvement at common iliac bifurcation, and seminal vesicle infiltration. The PGM achieved superior predictive performance with a balanced accuracy of 0.73, sensitivity of 0.60, and specificity of 0.86, substantially outperforming both the LR (balanced accuracy: 0.50, sensitivity: 0.00, specificity: 1.00) and CAPRA (balanced accuracy: 0.59, sensitivity: 0.20, specificity: 0.99). The decision tree provided an explainable classifier with CAPRA as a primary branch node, followed by SUVmax and specific PET-detected tumor sites. Conclusions: Integrating ⁶⁸Ga-PSMA-11 imaging biomarkers with clinical parameters, such as CAPRA, significantly improves models to predict progression in patients with high-risk PCa undergoing primary treatment. The PGM offers superior balanced accuracy and enables risk stratification that may guide personalized treatment decisions. Full article

Background: Chronic kidney disease (CKD) is a prevalent complication among individuals with type 2 diabetes (T2D), posing significant diagnostic challenges in resource-limited settings due to infrequent testing and missed hospital visits. This study aimed to develop a simple, effective ML model to identify T2D patients at high risk for reduced kidney function. Methods: We retrospectively analyzed data from 3471 T2D patients collected over a ten-year period at a university hospital in Bangkok, Thailand. Two models were developed using readily available clinical features: one including hemoglobin A1c (HbA1c) levels (the “with-HbA1c” model) and one excluding HbA1c levels (the “non–HbA1c” model). Three tree-based ML algorithms—decision tree, random forest, and extreme gradient boosting (XGBoost) algorithms—were employed. The outcome label was CKD, defined as an estimated Glomerular Filtration Rate (eGFR) < 60 mL/min/1.73 m² that persisted for more than 90 days. The model performance was evaluated using the AUROC. The feature importance was assessed using Shapley additive explanations (SHAP). Results: The XGBoost algorithm demonstrated a strong predictive performance. The “with-HbA1c” model achieved an AUROC of 0.824, while the “non–HbA1c” model attained a comparable AUROC of 0.819. Both models were well-calibrated. SHAP analysis identified age, HbA1c, and systolic blood pressure as the most influential predictors. Conclusions: Our simplified, interpretable ML models can effectively stratify the risk of reduced kidney function in patients with T2D using minimal, routine data. These models represent a promising step toward integration into clinical practice, such as through EHR-based alerts or patient-facing mobile applications, to improve early CKD detection, particularly in resource-limited settings. Full article

Delirium is a common and underrecognized complication among critically ill patients, associated with prolonged ICU stays, cognitive dysfunction, and increased mortality. Its multifactorial causes and fluctuating course hinder early prediction, limiting timely management. Predictive models based on data available at ICU admission may help to identify high-risk patients and guide early interventions. This study evaluated machine learning models used to predict delirium in critically ill patients with SARS-CoV-2 infections using a prospective cohort of 426 patients. The dataset included demographic characteristics, clinical data (e.g., comorbidities, medication, reason for ICU admission, interventions), and routine lab test results. Five models—Logistic Regression, Support Vector Machine, Decision Tree, Random Forest, and Naïve Bayes—were developed using 112 features. Feature selection relied on Information Gain, and model performance was assessed via 10-fold cross-validation. The Naïve Bayes model showed moderate predictive performance and high interpretability, achieving an AUC of 0.717, accuracy of 65.3%, sensitivity of 62.4%, specificity of 68.1%, and precision of 66.2%. Key predictors included invasive mechanical ventilation, deep sedation with benzodiazepines, SARS-CoV-2 as the reason for ICU admission, ECMO use, constipation, and male sex. These findings support the use of interpretable models for early delirium risk stratification using routinely available ICU data. Full article