Clinical Prediction Model Research Articles

Corrigendum: A clinical practical model for preoperative prediction of visual outcome for pituitary adenoma patients in a retrospective and prospective study

Corrigendum: A clinical practical model for preoperative prediction of visual outcome for pituitary adenoma patients in a retrospective and prospective study

Refining MR-guided thermal ablation for HCC within the Milan criteria: a decade of clinical outcomes and predictive modeling at a single institution

BackgroundThe appropriateness of ablation for liver cancer patients meeting the Milan criteria remains controversial.PurposeThis study aims to evaluate the long-term outcomes of MR-guided thermal ablation for HCC patients meeting the Milan criteria and develop a nomogram for predicting survival rates.MethodsA retrospective analysis was conducted from January 2009 to December 2021 at a single institution. Patients underwent MR-guided thermal ablation. Factors influencing progression-free survival (PFS) and overall survival (OS) were identified using univariate and multivariate Cox regression and stepwise regression. A nomogram was developed for survival prediction, followed by risk stratification and internal validation. Adverse events (AEs) were also analyzed.ResultsA total of 181 patients were included, with a mean follow-up of 73.8 ± 31.7 months. The cumulative local tumor progression rates at 1, 3, and 5 years were 0.80%, 1.27%, and 1.86%, respectively. The 1-, 3-, and 5-year PFS rates were 81.8%, 57.4%, and 38.1%, and OS rates were 98.3%, 87.8%, and 62.9%. Poorer outcomes were associated with age ≤ 60 years, tumor size > 2 cm, multiple tumors, cirrhosis, proximity to major vessels, and narrow ablation margins (P < 0.05). The nomogram accurately predicted 3- and 5-year survival, and internal validation confirmed the results. AEs occurred in 33.7% of patients, with pain being the most common.ConclusionMR-guided ablation is effective for HCC patients within the Milan criteria, especially for those with smaller tumors and better liver function. The nomogram and risk stratification model are valuable tools for predicting patient outcomes and guiding treatment.

Exploring the role of multimodal [18F]F-PSMA-1007 PET/CT and multiparametric MRI data in predicting ISUP grading of primary prostate cancer.

The study explores the role of multimodal imaging techniques, such as [18F]F-PSMA-1007 PET/CT and multiparametric MRI (mpMRI), in predicting the ISUP (International Society of Urological Pathology) grading of prostate cancer. The goal is to enhance diagnostic accuracy and improve clinical decision-making by integrating these advanced imaging modalities with clinical variables. In particular, the study investigates the application of few-shot learning to address the challenge of limited data in prostate cancer imaging, which is often a common issue in medical research. This study conducted a retrospective analysis of 341 prostate cancer patients enrolled between 2019 and 2023, with data collected from five imaging modalities: [18F]F-PSMA-1007 PET, CT, Diffusion Weighted Imaging (DWI), T2 Weighted Imaging (T2WI), and Apparent Diffusion Coefficient (ADC). The study compared the performance of five single-modality data sets, PET/CT dual-modality fusion data, mpMRI tri-modality fusion data, and five-modality fusion data within deep learning networks, analyzing how different modalities impact the accuracy of ISUP grading prediction. To address the issue of limited data, a few-shot deep learning network was employed, enabling training and cross-validation with only a small set of labeled samples. Additionally, the results were compared with those from preoperative biopsies and clinical prediction models to further assess the reliability of the experimental findings. The experimental results demonstrate that the multimodal model (combining [18F]F-PSMA-1007 PET/CT and multiparametric MRI) significantly outperforms other models in predicting ISUP grading of prostate cancer. Meanwhile, both the PET/CT dual-modality and mpMRI tri-modality models outperform the single-modality model, with comparable performance between the two multimodal models. Furthermore, the experimental data confirm that the few-shot learning network introduced in this study provides reliable predictions, even with limited data. This study highlights the potential of applying multimodal imaging techniques (such as PET/CT and mpMRI) in predicting ISUP grading of prostate cancer. The findings suggest that this integrated approach can enhance the accuracy of prostate cancer diagnosis and contribute to more personalized treatment planning. Furthermore, incorporating few-shot learning into the model development process allows for more robust predictions despite limited data, making this approach highly valuable in clinical settings with sparse data.

Development and validation of a novel risk-predicted model for early sepsis-associated acute kidney injury in critically ill patients: a retrospective cohort study

ObjectivesThis study aimed to develop a prediction model for the detection of early sepsis-associated acute kidney injury (SA-AKI), which is defined as AKI diagnosed within 48 hours of a sepsis...

Developing a clinical prediction model to modify empirical antibiotics for non-typhoidal Salmonella bloodstream infection in children under-five in the Democratic Republic of Congo

BackgroundNon-typhoidal Salmonella (NTS) frequently cause bloodstream infection in children under-five in sub-Saharan Africa, particularly in malaria-endemic areas. Due to increasing drug resistance, NTS are often not covered by standard-of-care empirical antibiotics for severe febrile illness. We developed a clinical prediction model to orient the choice of empirical antibiotics (standard-of-care versus alternative antibiotics) for children admitted to hospital in settings with high proportions of drug-resistant NTS.MethodsData were collected during a prospective cohort study in children (> 28 days—< 5 years) admitted with severe febrile illness to Kisantu district hospital, DR Congo. The outcome variable was blood culture confirmed NTS bloodstream infection; the comparison group were children without NTS bloodstream infection. Predictors were selected a priori based on systematic literature review. The prediction model was developed with multivariable logistic regression; a simplified scoring system was derived. Internal validation to estimate optimism-corrected performance was performed using bootstrapping and net benefits were calculated to evaluate clinical usefulness.ResultsNTS bloodstream infection was diagnosed in 12.7% (295/2327) of enrolled children. The area under the curve was 0.79 (95%CI: 0.76–0.82) for the prediction model, and 0.78 (0.85–0.80) for the scoring system. The estimated calibration slopes were 0.95 (model) and 0.91 (scoring system). At a decision threshold of 20% NTS risk, the prediction model and scoring system had 57% and 53% sensitivity, and 85% specificity. The net benefit for decisions thresholds < 30% ranged from 2.4 to 3.9 per 100 children.ConclusionThe model predicts NTS bloodstream infection and can support the choice of empiric antibiotics to include coverage of drug-resistant NTS, in particular for decision thresholds < 30%. External validation studies are needed to investigate generalizability.Trial registrationDeNTS study, clinicaltrials.gov: NCT04473768 (registration 16/07/2020) and TreNTS study, clinicaltrials.gov: NCT04850677 (registration 20/04/2021).

Machine Learning Model for Risk Stratification of Papillary Thyroid Carcinoma Based on Radiopathomics.

This study aims to develop a radiopathomics model based on preoperative ultrasoundand fine-needle aspiration cytology (FNAC) images to enable accurate, non-invasive preoperative risk stratification for patients with papillary thyroid carcinoma (PTC). The model seeks to enhance clinical decision-making by optimizing preoperative treatment strategies. A retrospective analysis was conducted on data from PTC patients who underwent thyroidectomy between October 2022 and May 2024 across six centers. Based on lymph node dissection outcomes, patients were categorized into high-risk and low-risk groups. Initially, a clinical predictive model was established based on the maximum diameter of the thyroid nodules. Radiomics features were extracted from preoperative two-dimensional ultrasound images, and pathomics features were extracted from 400x magnification H&E-stained tumor cell images from FNAC. The most predictive radiomics and pathomics features were identified through univariate analysis, Pearson correlation analysis and LASSO algorithm. The most valuable radiopathomics features were then selected by combining these predictive features. Finally, machine learningwith the XGBoost algorithm was employed to construct radiomics, pathomics, and radiopathomics models. The performance of the models was evaluated using the area under the curve (AUC), decision curve analysis, accuracy, specificity, sensitivity, positive predictive value, and negative predictive value. A total of 688 PTC patients were included, with 344 classified as intermediate/high-risk and 344 as low-risk. The multimodal radiopathomics model demonstrated excellent predictive performance, with AUCs of 0.886 (95% CI: 0.829-0.924) and 0.828 (95% CI: 0.751-0.879) in two external validation cohorts, significantly outperforming the clinical model (AUCs of 0.662 and 0.601), radiomics model (AUCs of 0.702 and 0.697), and pathomics model (AUCs of 0.741 and 0.712). The radiopathomics model exhibits significant advantages in accurately predicting preoperative risk stratification in PTC patients. Its application is expected to reduce unnecessary lymph node dissection surgeries, optimize treatment strategies, and improve therapeutic outcomes.

The Harms of Class Imbalance Corrections for Machine Learning Based Prediction Models: A Simulation Study.

Risk prediction models are increasingly used in healthcare to aid in clinical decision-making. In most clinical contexts, model calibration (i.e., assessing the reliability of risk estimates) is critical. Data available for model development are often not perfectly balanced with the modeled outcome (i.e., individuals with vs. without the event of interest are not equally prevalentin the data). It is common for researchers to correct for class imbalance, yet, the effect of such imbalance corrections on the calibration of machine learning models is largely unknown. We studied the effect of imbalance corrections on model calibration for a variety of machine learning algorithms. Using extensive Monte Carlo simulations we compared the out-of-sample predictive performance of models developed with an imbalance correction to those developed without a correction for class imbalance across different data-generating scenarios (varying sample size, the number of predictors, and event fraction). Our findings were illustrated in a case study using MIMIC-III data. In all simulation scenarios, prediction models developed without a correction for class imbalance consistently had equal or better calibration performance than prediction models developed with a correction for class imbalance. The miscalibration introduced by correcting for class imbalance was characterized by an over-estimation of risk and was not always able to be corrected with re-calibration. Correcting for class imbalance is not always necessary and may even be harmful to clinical prediction models which aim to produce reliable risk estimates on an individual basis.

Statistical models versus machine learning approach for competing risks in proctological surgery.

Clinical risk prediction models are ubiquitous in many surgical domains. The traditional approach to develop these models involves the use of regression analysis. Machine learning algorithms are gaining in popularity as an alternative approach for prediction and classification problems. They can detect non-linear relationships between independent and dependent variables and incorporate many of them. In our work, we aimed to investigate the potential role of machine learning versus classical logistic regression for the preoperative risk assessment in proctological surgery. We used clinical data from a nationwide audit: the database consisted of 1510 patients affected by Goligher's grade III hemorrhoidal disease who underwent elective surgery. We collected anthropometric, clinical, and surgical data and we considered ten predictors to evaluate model-predictive performance. The clinical outcome was the complication rate evaluated at 30-day follow-up. Logistic regression and three machine learning techniques (Decision Tree, Support Vector Machine, Extreme Gradient Boosting) were compared in terms of area under the curve, balanced accuracy, sensitivity, and specificity. In our setting, machine learning and logistic regression models reached an equivalent predictive performance. Regarding the relative importance of the input features, all models agreed in identifying the most important factor. Combining and comparing statistical analysis and machine learning approaches in clinical field should be a common ambition, focused on improving and expanding interdisciplinary cooperation.

Developing predictive nomogram models using quantitative electroencephalography for brain function in type a aortic dissection: a prospective observational study.

Type A aortic dissection (TAAD) remains a significant challenge in cardiac surgery, presenting high risks of adverse outcomes such as permanent neurological dysfunction and mortality despite advances in medical technology and surgical techniques. This study investigates the use of quantitative electroencephalography (QEEG) to monitor and predict neurological outcomes during the perioperative period in TAAD patients. This prospective observational study was conducted at the hospital, involving patients undergoing TAAD surgery from February 2022 to January 2023. QEEG parameters, including the dynamic amplitude-integrated electroencephalography (aEEG) grade, which assesses changes in brain function over time, alongside aEEG and relative band power (RBP), were monitored and analyzed to assess brain function preoperatively, intraoperatively, and within 2hours postoperatively. A predictive nomogram model was developed using these QEEG metrics along with other clinical variables to forecast neurological outcomes. In this study, we analyzed the factors contributing to adverse outcomes (AO) and transient neurological dysfunction (TND) following TAAD surgery. For AO, multivariable analysis identified pre-mental status (odds ratio [OR]=4.652, 95% confidence interval [CI]=2.316-10.074, P < 0.001), cardiopulmonary bypass time (OR=1.014, 95% CI=1.006-1.023, P =0.001), and dynamic aEEG grade (OR=9.926, 95% CI=4.493-25.268, P < 0.001) as independent risk factors. The AO model showed high discriminative ability with an area under the curve of 0.888 (95% CI=0.818-0.960) and good calibration (Brier score=0.0728). For TND, significant preoperative differences included dynamic aEEG grade (P < 0.001) and Log(Post-RBP Alpha%) (6.00 vs. 4.00, P < 0.001). Multivariable analysis identified cardiopulmonary bypass time (OR=1.014, 95% CI=1.006-1.023, P =0.001), Post-RBP Alpha% (OR=0.263, 95% CI=0.121-0.532, P < 0.001), and dynamic aEEG grade (OR=12.444, 95% CI=5.337-30.814, P < 0.001) as independent risk factors. The TND model had an area under the curve of 0.893 (95% CI=0.844-0.941) and good calibration (Brier score=0.125). These findings highlight the role of QEEG in predicting postoperative neurological dysfunction in TAAD patients. Through perioperative QEEG monitoring of TAAD patients, combined with clinical indicators such as cardiopulmonary bypass time and preoperative mental status, we developed clinical predictive models for AO and TND after surgery. These models allow for early detection of postoperative brain function impairment, as assessed by QEEG parameters monitored intraoperatively and during the first 2hours after surgery, a period chosen based on clinical definitions of delayed awakening and supported by the findings of this study. This study provides evidence supporting postoperative brain function monitoring in TAAD patients, with potential clinical implications for improved outcomes.

Adult patients with Philadelphia chromosome positive acute lymphoblastic leukemia undergoing allogeneic hematopoietic stem cell transplantation and tyrosine kinase inhibitors: development and validation of a clinical prediction model based on cytogenetics, IKZF1 deletions and minimal residual disease.

The aim of this study was to develop and validate a nomogram predicting progression-free survival (PFS) for adult patients with positive acute lymphoblastic leukemia(Ph + ALL) who have undergone allogeneic hematopoietic stem cell transplantation(allo-HSCT) and tyrosine kinase inhibitor(TKI) treatment. Data were retrospectively collected from 176 adult patients diagnosed with Ph + ALL and treated with allo-HSCT and TKIs at The First Affiliated Hospital, Zhejiang University School of Medicine, between January 2015 and May 2023. 70% of the patients were randomly assigned to the training group(n = 124) and 30% of the patients were assigned to the validation group(n = 52). Univariate Cox regression analysis and Akaike Information Criterion(AIC) were utilized to identify significant predictive factors, leading to the development of a nomogram designed to forecast the probability of PFS at 6, 9, and 12 months post-transplantation. The final nomogram incorporated three key variables: presence of complex additional cytogenetic abnormalities (ACAs), minimal residual disease (MRD) status prior to allo-HSCT, and IKZF1 gene deletions. The calibration curves showed excellent consistency between the nomogram prediction and actual observation for 6-, 9- and 12-month PFS in the training set and validation set. The C-index of the training set was 0.726(95%CI: 0.635-0.816), which was no significantly different from the validation set(C-index = 0.774, 95%CI: 0.674-0.875, P > 0.05). This study may provide a simple and efficient prediction model for patients with Ph + ALL undergoing allo-HSCT and TKIs, which can accurately predict PFS subsequent to transplantation. This tool could potentially aid clinicians in decision-making processes and improve patient outcomes.

CiThroModel Improves Prediction of Symptomatic Venous Thromboembolism in Hospitalized Patients With Cirrhosis Without Hepatocellular Carcinoma.

Venous thromboembolism (VTE) is a recognized complication of acutely ill patients, but its incidence and risk factors in those with cirrhosis are uncertain. We retrospectively studied a consecutive cohort of cirrhosis patients non-electively admitted to our medical unit to determine the rates of symptomatic VTE during hospitalization. Firstly, we explored associations with baseline, clinical and laboratory characteristics using logistic regression. Secondly, we developed a clinical prediction model that could predict the risk of in-hospital VTE. We included 687 patients (median age 61years old; 68% male; Child-Pugh A/B/C, 13%/40%/47%). During hospitalization, 34 patients (4.9%) experienced VTE. Multivariate analysis showed that male sex (OR: 2.56, p=0.05), AKI (OR: 3.1, p=0.001), bacterial infections (OR: 2.6, p=0.008), Pugh score (OR: 1.6. p<0.001), family history of thrombosis (OR: 3.1, p=0.04), reduced mobility (OR: 4.6, p<0.001), and C-reactive protein (OR: 1.1, p=0.005) were independent predictors of VTE. We combined these variables in a prediction model (CirrhosisThrombosisModel) that accurately discriminated between high- and low-risk patients. The AUROC of CiThroModel was significantly higher than that of Padua prediction score (0.882 vs. 0.742). After validating the CiThroModel using bootstrapping, the adjusted model maintained optimal discrimination ability (0.862) and calibration. The adjusted formula to calculate the in-hospital risk of VTE was -9.00+0.82 [Male sex]+1.14 [AKI]+0.98 [Infection]+0.48*Child Pugh score+1.14 [VTE family history]+1.54 [Reduced mobility] + 0.15*PCR/10. The CiThroModel seems a valuable tool for identifying hospitalized patients with cirrhosis at risk of VTE (https://majinzin.shinyapps.io/vterisk/).

Evaluating the added value of multiparametric functional MRI in predicting renal function response to renal artery stenting in severe atherosclerotic renal artery stenosis.

To explore the potential of Intravoxel Incoherent Motion Diffusion (IVIM) and Arterial Spin Labeling (ASL) in predicting the short-term effectiveness of post-revascularization for severe atherosclerotic renal artery stenosis. A retrospective analysis of 88 cases from October 2018 to February 2023 was conducted. Patients were divided into Responder and Non-Responder groups based on renal function outcomes at their last follow-up. Clinical data were compared between the groups, and preoperative functional MRI images were analyzed. ROIs were outlined for the affected and both kidneys. Measurements included ASL-derived renal blood flow (RBF), IVIM's pseudo-diffusion coefficient (D*), perfusion fraction (f), true diffusion coefficient (D), and the conventional apparent diffusion coefficient (ADC).Multivariate logistic regression identified independent clinical predictors of benefit, and a clinical prediction model was developed. Model performance was assessed using Receiver Operating Characteristic (ROC) curves and Decision Curve Analysis(DCA) curves. In the training cohort of 54 non-responders and 34 responders, no quantitative parameters of bilateral kidneys showed statistical significance in predicting Responders (all p > 0.05). Pre-treatment eGFR, presence of diabetes, and the D value of the affected kidney were identified as independent factors for predicting short-term treatment effectiveness. The combined clinical-functional imaging model yielded the higher AUC at 0.796 (95 % CI: 0.690-0.897) . Decision curve analysis further confirmed the better net benefit of combined model. Beyond clinical characteristics, functional MRI had the potential to predict response of stenting for severe atherosclerotic renal artery stenosis.

P1051 Establishment and Validation of Prediction Models for Short-term and Long-term Effectiveness of Ustekinumab in Patients with Crohn’s Disease

Abstract Background Ustekinumab (UST), a novel biological agent for Crohn’s disease (CD), has shown efficacy in CD patients. Our objective was to establish clinical prediction models to accurately forecast the short- and long-term effectiveness of UST in individuals diagnosed with CD. Methods We established a derivation cohort comprising patients diagnosed with CD and treated with UST at the Sixth Affiliated Hospital of Sun Yat-sen University from May 2020 to July 2023. The patients were assigned in a 7:3 ratio at random to the training cohort and the validation cohort. Predictors were compressed through the Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis. Logistic regression analysis was used to create the final prediction models. Results A total of 464 patients with CD were included in the statistical analysis, 171 of whom had complete endoscopic data. Four models were constructed: two to predict the probability of clinical or endoscopic remission at week 16 or 20 based on baseline data, and two to predict the probability of clinical or endoscopic remission at week 52 based on data obtained at week 16 or 20. The C-index was 0.714, 0.866, 0.814 and 0.841, respectively. Because of poor calibration, we constructed another model to predict the probability of endoscopic remission at week 52 based on baseline data (C-index:0.748). In the internal validation, the above models showed good performance. Conclusion We have constructed several models to predict the short- and long-term effectiveness of UST in patients with CD, promising to be a simple and practical clinical tool.

A nomogram for the prediction of co-infection in MDA5 dermatomyositis: A rapid clinical assessment model.

Patients with anti-melanoma differentiation-associated gene 5-positive dermatomyositis (MDA5 DM) are prone to infections, but there is a lack of rapid methods to assess infection risk, which greatly affects patient prognosis. This study aims to analyze the clinical features of MDA5 DM patients systematically and develop a predictive model for infections. Retrospective analysis was performed on clinical data from 118 hospitalized patients with MDA5 DM. According to the results of pathogen detection and clinical manifestations, the patients were divided into infected group and non-infected group. LASSO analysis and multivariate logistic regression were used to establish the prediction model of infection in MAD5 DM patients. The resulting model was visualized using a Nomogram. We used methods such as Receiver Operating Characteristic (ROC) curve analysis, Area Under the Curve (AUC) calculation to evaluate the model. The Cough, interstitial lung disease, moist rales, positive anti-RO-52, carcinoembryonic antigen, triglyceride, hydroxybutyrate dehydrogenase and erythrocyte sedimentation rate were significantly associated with infection risk in MDA5 DM patients. A prediction model was developed using these eight risk factors, achieving an AUC of 0.851 in determining co-infection status. Further analysis based on infection site and pathogen classification demonstrated strong discrimination performance of the model in identifying pulmonary infection (AUC: 0.844) and fungal infection (AUC: 0.822). This study aimed to develop a clinical prediction model and visualize it using Nomogram to assess the risk of infection in MDA5 DM. The model provides an effective tool for determining infection status in patients and serves as a reference for formulating clinical medication regimens.

Development of a clinical prediction model for benign and malignant pulmonary nodules with a CTR ≥ 50% utilizing artificial intelligence-driven radiomics analysis

ObjectiveIn clinical practice, diagnosing the benignity and malignancy of solid-component-predominant pulmonary nodules is challenging, especially when 3D consolidation-to-tumor ratio (CTR) ≥ 50%, as malignant ones are more invasive. This study aims to develop and validate an AI-driven radiomics prediction model for such nodules to enhance diagnostic accuracy.MethodsData of 2,591 pulmonary nodules from five medical centers (Zhengzhou People’s Hospital, etc.) were collected. Applying exclusion criteria, 370 nodules (78 benign, 292 malignant) with 3D CTR ≥ 50% were selected and randomly split 7:3 into training and validation cohorts. Using R programming, Lasso regression with 10-fold cross-validation filtered features, followed by univariate and multivariate logistic regression to construct the model. Its efficacy was evaluated by ROC, DCA curves and calibration plots.ResultsLasso regression picked 18 non-zero coefficients from 108 features. Three significant factors—patient age, solid component volume and mean CT value—were identified. The logistic regression equation was formulated. In the training set, the ROC AUC was 0.721 (95%CI: 0.642–0.801); in the validation set, AUC was 0.757 (95%CI: 0.632–0.881), showing the model’s stability and predictive ability.ConclusionThe model has moderate accuracy in differentiating benign from malignant 3D CTR ≥ 50% nodules, holding clinical potential. Future efforts could explore more to improve its precision and value.Clinical trial numberNot applicable.

Unveiling Silent Atherosclerosis in Type 1 Diabetes: The Role of Glycoprotein and Lipoprotein Lipidomics, and Cardiac Autonomic Neuropathy.

Introduction: This study aimed to evaluate whether glycoprotein and lipoprotein lipidomics profiles could enhance a clinical predictive model for carotid subclinical atherosclerosis in patients with type 1 diabetes (T1D). Additionally, we assessed the influence of cardiac autonomic neuropathy (CAN) on these predictive models. Methods: We conducted a cross-sectional study including 256 patients with T1D. Serum glycoprotein and lipoprotein lipidomics profiles were determined using 1H-NMR spectroscopy. Subclinical atherosclerosis was defined as carotid intima-media thickness (cIMT) ≥ 1.5 mm. CAN was identified using the Clarke score. Predictive models were built and their performance evaluated using receiver operating characteristic curves and cross-validation. Results: Subclinical atherosclerosis was detected in 32% of participants. Patients with both CAN and atherosclerosis were older, had a longer duration of diabetes, and were more likely to present with bilateral carotid disease. Clinical predictors such as age, duration of diabetes, and smoking status remained the strongest determinants of subclinical atherosclerosis [AUC = 0.88 (95%CI: 0.84-0.93)]. While glycoprotein and lipoprotein lipidomics profiles were associated with atherosclerosis, their inclusion in the clinical model did not significantly improve its diagnostic performance. Stratification by the presence of CAN revealed no impact on the model's ability to predict subclinical atherosclerosis, underscoring its robustness across different risk subgroups. Conclusions: In a cohort of patients with T1D, subclinical atherosclerosis was strongly associated with traditional clinical risk factors. Advanced glycoprotein and lipoprotein lipidomics profiling, although associated with atherosclerosis, did not enhance the diagnostic accuracy of predictive models beyond clinical variables. The predictive model remained effective even in the presence of CAN, highlighting its reliability as a screening tool for identifying patients at risk of subclinical atherosclerosis.

Deep learning of noncontrast CT for fast prediction of hemorrhagic transformation of acute ischemic stroke: a multicenter study

BackgroundHemorrhagic transformation (HT) is a complication of reperfusion therapy following acute ischemic stroke (AIS). We aimed to develop and validate a model for predicting HT and its subtypes with poor prognosis—parenchymal hemorrhage (PH), including PH-1 (hematoma within infarcted tissue, occupying < 30%) and PH-2 (hematoma occupying ≥ 30% of the infarcted tissue)—in AIS patients following intravenous thrombolysis (IVT) based on noncontrast computed tomography (NCCT) and clinical data.MethodsIn this six-center retrospective study, clinical and imaging data from 445 consecutive IVT-treated AIS patients were collected (01/2018–06/2023). The training cohort comprised 344 patients from five centers, and the test cohort included 101 patients from the sixth center. A clinical model was developed using eXtreme Gradient Boosting, an NCCT-based imaging model was created using deep learning, and an ensemble model integrated both models. Comparison with existing clinical scores (MSS, SEDAN, GRASPS) was performed using the DeLong test.ResultsOf the 445 individuals, 202 (45.4%) had HT, 79 (17.8%) had hemorrhagic infarction, and 123 (27.6%) had PH. In the test cohort, the area under the receiver operating characteristic curve (AUROC) of the clinical, imaging, and ensemble model for HT prediction was 0.877, 0.920, and 0.937, respectively. The ensemble model for HT prediction outperformed MSS, SEDAN, and GRASPS scores (p ≤ 0.023). The ensemble model predicted PH and PH-2 with AUROC of 0.858 and 0.806, respectively.ConclusionDeveloping and validating an integrated model that can predict HT and its subtypes in AIS patients following IVT based on NCCT and clinical data is feasible.Relevance statementThe clinical, imaging, and ensemble models based on noncontrast CT and clinical data outperformed existing clinical scores in predicting hemorrhagic transformation of AIS and its subtypes with poor prognosis, facilitating personalized treatment decisions.Key PointsThe models demonstrated the capability to predict hemorrhagic transformation of acute ischemic stroke quickly, accurately, and reliably.The proposed models outperformed existing clinical scores in predicting hemorrhagic transformation.The ensemble model provided risk assessment of parenchymal hemorrhage and parenchymal hemorrhage-2 outperforming existing clinical scores.Graphical

Derivation and validation of a clinical predictive model for longer duration diarrhea among pediatric patients in Kenya using machine learning algorithms

BackgroundDespite the adverse health outcomes associated with longer duration diarrhea (LDD), there are currently no clinical decision tools for timely identification and better management of children with increased risk. This study utilizes machine learning (ML) to derive and validate a predictive model for LDD among children presenting with diarrhea to health facilities.MethodsLDD was defined as a diarrhea episode lasting ≥ 7 days. We used 7 ML algorithms to build prognostic models for the prediction of LDD among children < 5 years using de-identified data from Vaccine Impact on Diarrhea in Africa study (N = 1,482) in model development and data from Enterics for Global Health Shigella study (N = 682) in temporal validation of the champion model. Features included demographic, medical history and clinical examination data collected at enrolment in both studies. We conducted split-sampling and employed K-fold cross-validation with over-sampling technique in the model development. Moreover, critical predictors of LDD and their impact on prediction were obtained using an explainable model agnostic approach. The champion model was determined based on the area under the curve (AUC) metric. Model calibrations were assessed using Brier, Spiegelhalter’s z-test and its accompanying p-value.ResultsThere was a significant difference in prevalence of LDD between the development and temporal validation cohorts (478 [32.3%] vs 69 [10.1%]; p < 0.001). The following variables were associated with LDD in decreasing order: pre-enrolment diarrhea days (55.1%), modified Vesikari score(18.2%), age group (10.7%), vomit days (8.8%), respiratory rate (6.5%), vomiting (6.4%), vomit frequency (6.2%), rotavirus vaccination (6.1%), skin pinch (2.4%) and stool frequency (2.4%). While all models showed good prediction capability, the random forest model achieved the best performance (AUC [95% Confidence Interval]: 83.0 [78.6–87.5] and 71.0 [62.5–79.4]) on the development and temporal validation datasets, respectively. While the random forest model showed slight deviations from perfect calibration, these deviations were not statistically significant (Brier score = 0.17, Spiegelhalter p-value = 0.219).ConclusionsOur study suggests ML derived algorithms could be used to rapidly identify children at increased risk of LDD. Integrating ML derived models into clinical decision-making may allow clinicians to target these children with closer observation and enhanced management.

Machine learning for predicting neoadjuvant chemotherapy effectiveness using ultrasound radiomics features and routine clinical data of patients with breast cancer

BackgroundThis study explores the clinical value of a machine learning (ML) model based on ultrasound radiomics features of primary foci, combined with clinicopathologic factors to predict the pathological complete response (pCR) of neoadjuvant chemotherapy (NAC) for patients with breast cancer (BC).MethodWe retrospectively analyzed ultrasound images and clinical information from 231 participants with BC who received NAC. These patients were randomly assigned to training and validation cohorts. Tumor regions of interest (ROI) were delineated, and radiomics features were extracted. Z-score normalization, Pearson correlation analysis, and the least absolute shrinkage selection operator (LASSO) were utilized for further screening ultrasound radiomics and clinical features. Univariate and multivariate logistic regression analysis were performed to identify the CFs that were independently associated with pCR. We compared 10 ML models based on radiomics features: support vector machine (SVM), logistic regression (LR), random forest, extra trees (ET), naïve Bayes (NB), k-nearest neighbor (KNN), multilayer perceptron (MLP), gradient boosting ML (GBM), light GBM (LGBM), and adaptive boost (AB). Diagnostic performance was evaluated using the receiver operating characteristic (ROC) area under the curve (AUC), accuracy, sensitivity, and specificity, and the Rad score was calculated. Subsequently, construction of clinical predictive models and Rad score joint clinical predictive models using ML algorithms for optimal diagnostic performance. The diagnostic process of the ML model was visualized and analyzed using SHapley Additive exPlanation (SHAP).ResultsOut of 231 participants with BC, 98 (42.42%) achieved pCR, and 133 (57.58%) did not. Twelve radiomics features were identified, with the GBM model demonstrating the best predictive performance (AUC of 0.851, accuracy of 0.75, sensitivity of 0.821, and specificity of 0.698). The clinical feature prediction model using the GBM algorithm had an AUC of 0.819 and an accuracy of 0.739. Combining the Rad score with clinical features in the GBM model resulted in superior predictive performance (AUC of 0.939 and an accuracy of 0.87). SHAP analysis indicated that participants with a high Rad score, PR-negative, ER-negative and human epidermal growth factor receptor-2 (HER-2) positive were more possibly to reach pCR. Based on the decision curve analysis, it was shown that the combined model of GBM provided higher clinical benefits.ConclusionThe GBM model based on ultrasound radiomics features and routine clinical date of BC patients had high performance in predicting pCR. SHAP analysis provided a clear explanation for the prediction results of the GBM model, revealing that patients with a high Rad score, PR-negative status, ER-negative status and HER-2-positive status are more likely to achieve pCR.

Classification Prediction of Hydrocephalus After Intercerebral Haemorrhage Based on Machine Learning Approach.

In order to construct a clinical classification prediction model for hydrocephalus after intercerebral haemorrhage(ICH) to guide clinical treatment decisions, this paper retrospectively analyses the clinical data of 844 cases of ICH and hydrocephalus inpatients admitted to Yueyang People's Hospital from May 2019 to October 2022, of which 95 cases of hydrocephalus occurred after ICH and no hydrocephalus in 749 cases. The following indicators were compared between the two groups of patients: gender, age, Glasgow Coma Scale(GCS)score, whether the amount of bleeding was greater than 30ml, whether it broke into the ventricle or not, modified Graeb score(MGS), modified Rankin Scale (MRS) score, whether surgery was performed or not, red blood cells, white blood cells, and platelets. After variable screening, the following six variables were selected: GCS score, MGS, MRS score, whether the bleeding volume was greater than 30ml, whether it broke into the ventricle or not, and whether surgery was performed or not were modelled and analysed using logistic regression model and support vector machine model in machine learning. The results showed that under the same conditions, the accuracy of the support vector machine model was 0.89 and F1 was 0.838 ,the value of the AUC of the support vector machine model is 0.888; the accuracy of the logistic regression model was 0.902 and F1 was 0.89, the value of the AUC of the support vector machine model is 0.903. Compared with the group without hydrocephalus, patients in the group with hydrocephalus had bleeding volume greater than 30ml, haemorrhage into the ventricles of the brain, and had undergone surgery in the brain, and the difference was statistically significant (P 0.001). Statistical analysis showed that GCS score ≤ 8.8, modified Graeb score (MGS) ≥ 10 and MRS score ≥ 3 were independent risk factors for the development of hydrocephalus after spontaneous ventricular haemorrhage. Therefore, patients with lower GCS score, higher modified Graeb score, higher MRS score, bleeding volume > 30ml, haemorrhage into the ventricles of the brain, and experience of having undergone surgery in the brain should be operated on early to remove the intraventricular haematoma in order to reduce the incidence of hydrocephalus.