Training Set Research Articles

Radiomics model based on dual-energy CT can determine the source of thrombus in strokes with middle cerebral artery occlusion.

To develop thrombus radiomics models based on dual-energy CT (DECT) for predicting etiologic cause of stroke. We retrospectively enrolled patients with occlusion of the middle cerebral artery who underwent computed tomography (NCCT) and DECT angiography (DECTA). 70keV virtual monoenergetic images (simulate conventional 120kVp CTA images) and iodine overlay maps (IOM) were reconstructed for analysis. Five logistic regression radiomics models for predicting cardioembolism (CE) were built based on the features extracted from NCCT, CTA and IOM images. From these, the best one was selected to integrate with clinical information for further construction of the combined model. The performance of the different models was evaluated and compared using ROC curve analysis, clinical decision curves (DCA), calibration curves and Delong test. Among all the radiomic models, model NCCT+IOM performed the best, with AUC = 0.95 significantly higher than model NCCT, model CTA, model IOM and model NCCT+CTA in the training set (AUC = 0.88, 0.78, 0.90,0.87, respectively, P < 0.05), and AUC = 0.92 in the testing set, significantly higher than model CTA (AUC = 0.71, P < 0.05). Smoking and NIHSS score were independent predictors of CE (P < 0.05). The combined model performed similarly to the model NCCT+IOM, with no statistically significant difference in AUC either in the training or test sets. (0.96 vs. 0.95; 0.94 vs. 0.92, both P > 0.05). Radiomics models constructed based on NCCT and IOM images can effectively determine the source of thrombus in stroke without relying on clinical information.

CT-Based Radiomics Models for Differentiation of Benign and Malignant Thyroid Nodules: A Multicenter Development and Validation Study.

BACKGROUND. CT is increasingly detecting thyroid nodules. Prior studies indicated a potential role of CT-based radiomics models in characterizing thyroid nodules, although these studies lacked external validation. OBJECTIVE. The purpose of this study was to develop and validate a CT-based radiomics model for the differentiation of benign and malignant thyroid nodules. METHODS. This retrospective study included 378 patients (mean age, 46.3 ± 13.9 [SD] years; 86 men, 292 women) with 408 resected thyroid nodules (145 benign, 263 malignant) from two centers (center 1: 293 nodules, January 2018 to December 2022; center 2: 115 nodules, January 2020 to December 2022) who underwent preoperative multiphase neck CT (noncontrast, arterial, and venous phases). Nodules from center 1 were divided into training (n = 206) and internal validation (n = 87) sets; all nodules from center 2 formed an external validation set. Radiologists assessed nodules for morphologic CT features. Nodules were manually segmented on all phases, and radiomic features were extracted. Conventional (clinical and morphologic CT), noncontrast CT radiomics, arterial phase CT radiomics, venous phase CT radiomics, multiphase CT radiomics, and combined (clinical, morphologic CT, and multiphase CT radiomics) models were established using feature selection methods and evaluated by ROC curve analysis, calibration-curve analysis, and decision-curve analysis. RESULTS. The combined model included patient age, three morphologic features (cystic change, "edge interruption" sign, abnormal cervical lymph nodes), and 28 radiomic features (from all three phases). In the external validation set, the combined model had an AUC of 0.923, and, at an optimal threshold derived in the training set, sensitivity of 84.0%, specificity of 94.1%, and accuracy of 87.0%. In the external validation set, the AUC was significantly higher for the combined model than for the conventional model (0.827), noncontrast CT radiomics model (0.847), arterial phase CT radiomics model (0.826), venous phase CT radiomics model (0.773), and multiphase CT radiomics model (0.824) (all p < .05). In the external validation set, the calibration curves indicated the lowest (i.e., best) Brier score for the combined model; in the decision-curve analysis, the combined model had the highest net benefit for most of the range of threshold probabilities. CONCLUSION. A combined model incorporating clinical, morphologic CT, and multiphase CT radiomics features exhibited robust performance in differentiating benign and malignant thyroid nodules. CLINICAL IMPACT. The combined radiomics model may help guide further management for thyroid nodules detected on CT.

Habitat radiomics and deep learning fusion nomogram to predict EGFR mutation status in stage I non-small cell lung cancer: a multicenter study

Develop a radiomics nomogram that integrates deep learning, radiomics, and clinical variables to predict epidermal growth factor receptor (EGFR) mutation status in patients with stage I non-small cell lung cancer (NSCLC). We retrospectively included 438 patients who underwent curative surgery and completed driver-gene mutation tests for stage I NSCLC from four academic medical centers. Predictive models were established by extracting and analyzing radiomic features in intratumoral, peritumoral, and habitat regions of CT images to identify EGFR mutation status in stage I NSCLC. Additionally, three deep learning models based on the intratumoral region were constructed. A nomogram was developed by integrating representative radiomic signatures, deep learning, and clinical features. Model performance was assessed by calculating the area under the receiver operating characteristic (ROC) curve. The established habitat radiomics features demonstrated encouraging performance in discriminating between EGFR mutant and wild-type, with predictive ability superior to other single models (AUC 0.886, 0.812, and 0.790 for the training, validation, and external test sets, respectively). The radiomics-based nomogram exhibited excellent performance, achieving the highest AUC values of 0.917, 0.837, and 0.809 in the training, validation, and external test sets, respectively. Decision curve analysis (DCA) indicated that the nomogram provided a higher net benefit than other radiomics models, offering valuable information for treatment.

Comparing Two Data Collection Procedures to Teach Academic Targets within a Group Instructional Format

ABSTRACT This study aimed to extend previous literature comparing continuous methods of data collection to estimation data, but this time implementing the data collection procedures within a group discrete trial teaching format with three individuals diagnosed with autism spectrum disorder. Group discrete trial teaching was conducted in a classroom setting using relevant academic targets for each child. The data collection procedures were examined in terms of accuracy of data collection, efficiency of teaching (i.e. number of trials delivered per session), and rate of child acquisition of targets. Results of the adapted alternating treatment design replicated across sets and multiple targets indicated that estimation data was found to be a valid data collection method in identifying mastery, with only two instances where the mastery probe did not confirm mastery. Trial-by-trial data indicated mastery four times in which a probe for mastery did not confirm mastery of targets. Additionally, it was found that instructors conducted more teaching trials within the estimation data condition across all learners and training sets and social validity measures indicated overall favorable responses toward the estimation data collection procedure from both instructors.

MR radiomics to predict microvascular invasion status and biological process in combined hepatocellular carcinoma-cholangiocarcinoma

ObjectivesTo establish an MRI-based radiomics model for predicting the microvascular invasion (MVI) status of cHCC-CCA and to investigate biological processes underlying the radiomics model.MethodsThe study consisted of a retrospective dataset (82 in the training set, 36 in the validation set) and a prospective dataset (25 patients in the test set) from two hospitals. Based on the training set, logistic regression analyses were employed to develop the clinical-imaging model, while radiomic features were extracted to construct a radiomics model. The diagnosis performance was further validated in the validation and test sets. Prognostic aspects of the radiomics model were investigated using the Kaplan–Meier method and log-rank test. Differential gene expression analysis and gene ontology (GO) analysis were conducted to explore biological processes underlying the radiomics model based on RNA sequencing data.ResultsOne hundred forty-three patients (mean age, 56.4 ± 10.5; 114 men) were enrolled, in which 73 (51.0%) were confirmed as MVI-positive. The radiomics model exhibited good performance in predicting MVI status, with the area under the curve of 0.935, 0.873, and 0.779 in training, validation, and test sets, respectively. Overall survival (OS) was significantly different between the predicted MVI-negative and MVI-positive groups (median OS: 25 vs 18 months, p = 0.008). Radiogenomic analysis revealed associations between the radiomics model and biological processes involved in regulating the immune response.ConclusionA robust MRI-based radiomics model was established for predicting MVI status in cHCC-CCA, in which potential prognostic value and underlying biological processes that regulate immune response were demonstrated.Critical relevance statementMVI is a significant manifestation of tumor invasiveness, and the MR-based radiomics model established in our study will facilitate risk stratification. Furthermore, underlying biological processes demonstrated in the radiomics model will offer valuable insights for guiding immunotherapy strategies.Key PointsMVI is of prognostic significance in cHCC-CCA, but lacks reliable preoperative assessment.The MRI-based radiomics model predicts MVI status effectively in cHCC-CCA.The MRI-based radiomics model demonstrated prognostic value and underlying biological processes.The radiomics model could guide immunotherapy and risk stratification in cHCC-CCA.Graphical

Development and validation of a clinical predictive model for high-volume lymph node metastasis of papillary thyroid carcinoma

The central lymph node metastasis (CLNM) status in the cervical region serves as a pivotal determinant for the extent of surgical intervention and prognosis in papillary thyroid carcinoma (PTC). This paper seeks to devise and validate a predictive model based on clinical parameters for the early anticipation of high-volume CLNM (hv-CLNM, > 5 nodes) in high-risk patients. A retrospective analysis of the pathological and clinical data of patients with PTC who underwent surgical treatment at Medical Centers A and B was conducted. The data from Center A was randomly divided into training and validation sets in an 8:2 ratio, with those from Center B serving as the test set. Multifactor logistic regression was harnessed in the training set to select variables and construct a predictive model. The generalization ability of the model was assessed in the validation and test sets. The model was evaluated through the receiver operating characteristic area under the curve (AUC) to predict the efficiency of hv-CLNM. The goodness of fit of the model was examined via the Brier verification technique. The incidence of hv-CLNM in 5897 PTC patients attained 4.8%. The occurrence rates in males and females were 9.4% (128/1365) and 3.4% (156/4532), respectively. Multifactor logistic regression unraveled male gender (OR = 2.17, p < .001), multifocality (OR = 4.06, p < .001), and lesion size (OR = 1.08 per increase of 1 mm, p < .001) as risk factors, while age emerged as a protective factor (OR = 0.95 per an increase of 1 year, p < .001). The model constructed with four predictive variables within the training set exhibited an AUC of 0.847 ([95%CI] 0.815–0.878). In the validation and test sets, the AUCs were 0.831 (0.783–0.879) and 0.845 (0.789–0.901), respectively, with Brier scores of 0.037, 0.041, and 0.056. Subgroup analysis unveiled AUCs for the prediction model in PTC lesion size groups (≤ 10 mm and > 10 mm) as 0.803 (0.757–0.85) and 0.747 (0.709–0.785), age groups (≤ 31 years and > 31 years) as 0.778 (0.720–0.881) and 0.837 (0.806–0.867), multifocal and solitary cases as 0.803 (0.767–0.838) and 0.809 (0.769–0.849), and Hashimoto’s thyroiditis (HT) and non-HT cases as 0.845 (0.793–0.897) and 0.845 (0.819–0.871). Male gender, multifocality, and larger lesion size are risk factors for hv-CLNM in PTC patients, whereas age serves as a protective factor. The clinical predictive model developed in this research facilitates the early identification of high-risk patients for hv-CLNM, thereby assisting physicians in more efficacious risk stratification management for PTC patients.

Radiomics Based on Multimodal magnetic resonance imaging for the Differential Diagnosis of Benign and Malignant Vertebral Compression Fractures.

Recent studies have indicated that radiomics may have excellent performance and clinical application prospects in the differential diagnosis of benign and malignant vertebral compression fractures (VCFs). However, multimodal magnetic resonance imaging (MRI)-based radiomics model is rarely used in the differential diagnosis of benign and malignant VCFs, and is limited to lumbar. Herein, this study intends to develop and validate MRI radiomics models for differential diagnoses of benign and malignant VCFs in patients. This cross-sectional study involved 151 adult patients diagnosed with VCF in The First Affiliated Hospital of Soochow University in 2016-2021. The study was conducted in three steps: (i) the original MRI images were segmented, and the region of interest (ROI) was marked out; (ii) among the extracted features, those features with Pearson's correlation coefficient lower than 0.9 and the top 15 with the highest variance and Lasso regression coefficient less than and more than 0 were selected; (iii) MRI images and combined data were studied by logistic regression, decision tree, random forest and extreme gradient boosting (XGBoost) models in training set and the test set (ratio of 8:2), respectively; and the models were further verified and evaluated for the differential diagnosis performance. The evaluated indexes included area under receiver (AUC) of operating characteristic curve, accuracy, sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and 95% confidence intervals (CIs). The AUCs were used to assess the predictive performance of different machine learning modes for benign and malignant VCFs. A total of 1144 radiomics features, and 14 clinical features were extracted. Finally, 12 radiomics features were included in the radiomics model, and 12 radiomics features with 14 clinical features were included in the combined model. In the radiomics model, the differential diagnosis performance in the logistic regression model with the AUC of 0.905 ± 0.026, accuracy of 0.817 ± 0.057, sensitivity of 0.831 ± 0.065, and negative predictive value of 0.813 ± 0.042, was superior to the other three. In the combined model, XGBoost model had the superior differential diagnosis performance with specificity (0.979 ± 0.026) and positive predictive value (0.971 ± 0.035). The multimodal MRI-based radiomics model performed well in the differential diagnosis of benign and malignant VCFs, which may provide a tool for clinicians to differentially diagnose VCFs.

Developing a deep learning model for predicting ovarian cancer in Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US) Category 4 lesions: A multicenter study

PurposeTo develop a deep learning (DL) model for differentiating between benign and malignant ovarian tumors of Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US) Category 4 lesions, and validate its diagnostic performance.MethodsA retrospective analysis of 1619 US images obtained from three centers from December 2014 to March 2023. DeepLabV3 and YOLOv8 were jointly used to segment, classify, and detect ovarian tumors. Precision and recall and area under the receiver operating characteristic curve (AUC) were employed to assess the model performance.ResultsA total of 519 patients (including 269 benign and 250 malignant masses) were enrolled in the study. The number of women included in the training, validation, and test cohorts was 426, 46, and 47, respectively. The detection models exhibited an average precision of 98.68% (95% CI: 0.95–0.99) for benign masses and 96.23% (95% CI: 0.92–0.98) for malignant masses. Moreover, in the training set, the AUC was 0.96 (95% CI: 0.94–0.97), whereas in the validation set, the AUC was 0.93(95% CI: 0.89–0.94) and 0.95 (95% CI: 0.91–0.96) in the test set. The sensitivity, specificity, accuracy, positive predictive value, and negative predictive values for the training set were 0.943,0.957,0.951,0.966, and 0.936, respectively, whereas those for the validation set were 0.905,0.935, 0.935,0.919, and 0.931, respectively. In addition, the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value for the test set were 0.925, 0.955, 0.941, 0.956, and 0.927, respectively.ConclusionThe constructed DL model exhibited high diagnostic performance in distinguishing benign and malignant ovarian tumors in O-RADS US category 4 lesions.

Accurately Predicting Spatiotemporal Variations of Near-Surface Nitrous Acid (HONO) Based on a Deep Learning Approach.

Gaseous nitrous acid (HONO) is identified as a critical precursor of hydroxyl radicals (OH), influencing atmospheric oxidation capacity and the formation of secondary pollutants. However, large uncertainties persist regarding its formation and elimination mechanisms, impeding accurate simulation of HONO levels using chemical models. In this study, a deep neural network (DNN) model was established based on routine air quality data (O3, NO2, CO, and PM2.5) and meteorological parameters (temperature, relative humidity, solar zenith angle, and season) collected from four typical megacity clusters in China. The model exhibited robust performance on both the train sets [slope = 1.0, r2 = 0.94, root mean squared error (RMSE) = 0.29 ppbv] and two independent test sets (slope = 1.0, r2 = 0.79, and RMSE = 0.39 ppbv), demonstrated excellent capability in reproducing the spatiotemporal variations of HONO, and outperformed an observation-constrained box model incorporated with newly proposed HONO formation mechanisms. Nitrogen dioxide (NO2) was identified as the most impactful features for HONO prediction using the SHapely Additive exPlanation (SHAP) approach, highlighting the importance of NO2 conversion in HONO formation. The DNN model was further employed to predict the future change of HONO levels in different NOx abatement scenarios, which is expected to decrease 27-44% in summer as the result of 30-50% NOx reduction. These results suggest a dual effect brought by abatement of NOx emissions, leading to not only reduction of O3 and nitrate precursors but also decrease in HONO levels and hence primary radical production rates (PROx). In summary, this study demonstrates the feasibility of using deep learning approach to predict HONO concentrations, offering a promising supplement to traditional chemical models. Additionally, stringent NOx abatement would be beneficial for collaborative alleviation of O3 and secondary PM2.5.

Prediction of tumor regression grade in far-advanced gastric cancer after preoperative immuno-chemotherapy using dual-energy CT-derived extracellular volume fraction.

This study examines the effectiveness of dual-energy CT (DECT) delayed-phase extracellular volume (ECV) fraction in predicting tumor regression grade (TRG) in far-advanced gastric cancer (FAGC) patients receiving preoperative immuno-chemotherapy. A retrospective analysis was performed on far-advanced gastric adenocarcinoma patients treated with preoperative immuno-chemotherapy at our institution from August 2019 to March 2023. Patients were categorized based on their TRG into pathological complete response (pCR) and non-pCR groups. ECV was determined using the delayed-phase iodine maps. In addition, tumor iodine densities and standardized iodine ratios were meticulously analyzed using the triple-phase enhanced iodine maps. Univariate analysis with five-fold cross-validation and Spearman correlation determined DECT parameters and clinical indicators association with pCR. The predictive accuracy of these parameters for pCR was evaluated using a weighted logistic regression model with five-fold cross-validation. Of the 88 patients enrolled (mean age 60.8 ± 11.1 years, 63 males), 21 (23.9%) achieved pCR. Univariate analysis indicated ECV's significant role in differentiating between pCR and non-pCR groups (average p value = 0.021). In the logistic regression model, ECV independently predicted pCR with an average odds ratio of 0.911 (95% confidence interval, 0.798-0.994). The model, incorporating ECV, tumor area, and IDAV (the relative change rate of iodine density from venous phase to arterial phase), showed an average area under curves (AUCs) of 0.780 (0.770-0.791) and 0.766 (0.731-0.800) for the training and validation sets, respectively, in predicting pCR. DECT-derived ECV fraction is a valuable predictor of TRG in FAGC patients undergoing preoperative immuno-chemotherapy. This study demonstrates that DECT-derived extracellular volume fraction is a reliable predictor for pathological complete response in far-advanced gastric cancer patients receiving preoperative immuno-chemotherapy, offering a noninvasive tool for identifying potential treatment beneficiaries.

Evolutionary Strategies Enable Systematic and Reliable Uncertainty Quantification: A Proof-of-Concept Pilot Study on Resting-State Functional MRI Language Lateralization.

Reliable and trustworthy artificial intelligence (AI), particularly in high-stake medical diagnoses, necessitates effective uncertainty quantification (UQ). Existing UQ methods using model ensembles often introduce invalid variability or computational complexity, rendering them impractical and ineffective in clinical workflow. We propose a UQ approach based on deep neuroevolution (DNE), a data-efficient optimization strategy. Our goal is to replicate trends observed in expert-based UQ. We focused on language lateralization maps from resting-state functional MRI (rs-fMRI). Fifty rs-fMRI maps were divided into training/testing (30:20) sets, representing two labels: "left-dominant" and "co-dominant." DNE facilitated acquiring an ensemble of 100 models with high training and testing set accuracy. Model uncertainty was derived from distribution entropies over the 100 model predictions. Expert reviewers provided user-based uncertainties for comparison. Model (epistemic) and user-based (aleatoric) uncertainties were consistent in the independently and identically distributed (IID) testing set, mainly indicating low uncertainty. In a mostly out-of-distribution (OOD) holdout set, both model and user-based entropies correlated but displayed a bimodal distribution, with one peak representing low and another high uncertainty. We also found a statistically significant positive correlation between epistemic and aleatoric uncertainties. DNE-based UQ effectively mirrored user-based uncertainties, particularly highlighting increased uncertainty in OOD images. We conclude thatDNE-based UQ correlates with expert assessments, making it reliable for our use case and potentially for other radiology applications.

Patient-specific deep learning for 3D protoacoustic image reconstruction and dose verification in proton therapy.

Protoacoustic (PA) imaging has the potential to provide real-time 3D dose verification of proton therapy. However, PA images are susceptible to severe distortion due to limited angle acquisition. Our previous studies showed the potential of using deep learning to enhance PA images. As the model was trained using a limited number of patients' data, its efficacy was limited when applied to individualpatients. In this study, we developed a patient-specific deep learning method for protoacoustic imaging to improve the reconstruction quality of protoacoustic imaging and the accuracy of dose verification for individualpatients. Our method consists of two stages: in the first stage, a group model is trained from a diverse training set containing all patients, where a novel deep learning network is employed to directly reconstruct the initial pressure maps from the radiofrequency (RF) signals; in the second stage, we apply transfer learning on the pre-trained group model using patient-specific dataset derived from a novel data augmentation method to tune it into a patient-specific model. Raw PA signals were simulated based on computed tomography (CT) images and the pressure map derived from the planned dose. The reconstructed PA images were evaluated against the ground truth by using the root mean squared errors (RMSE), structural similarity index measure (SSIM) and gamma index on 10 specific prostate cancer patients. The significance level was evaluated by t-test with the p-value threshold of 0.05 compared with the results from the groupmodel. The patient-specific model achieved an average RMSE of 0.014 ( ), and an average SSIM of 0.981 ( ), out-performing the group model. Qualitative results also demonstrated that our patient-specific approach acquired better imaging quality with more details reconstructed when comparing with the group model. Dose verification achieved an average RMSE of 0.011 ( ), and an average SSIM of 0.995 ( ). Gamma index evaluation demonstrated a high agreement (97.4% [ ] and 97.9% [ ] for 1%/3 and 1%/5mm) between the predicted and the ground truth dose maps. Our approach approximately took 6 s to reconstruct PA images for each patient, demonstrating its feasibility for online 3D dose verification for prostate protontherapy. Our method demonstrated the feasibility of achieving 3D high-precision PA-based dose verification using patient-specific deep-learning approaches, which can potentially be used to guide the treatment to mitigate the impact of range uncertainty and improve the precision. Further studies are needed to validate the clinical impact of thetechnique.

Predictive modeling of lower extreme deep vein thrombosis following radical gastrectomy for gastric cancer: based on multiple machine learning methods.

Postoperative venous thromboembolic events (VTEs), such as lower extremity deep vein thrombosis (DVT), are major risk factors for gastric cancer (GC) patients following radical gastrectomy. Accurately predicting and managing these risks is crucial for optimal patient care.This retrospective case‒control study involved 693 GC patients from our hospital who underwent radical gastrectomy. We collected plentiful and comprehensive clinical indicators including a total of 49 baseline, preoperative, surgical and pathological clinical data. Using univariate logistic regression, we identified potential risk factors, followed by feature selection through the Boruta algorithm. We then constructed the final predictive model using multivariate logistic regression and evaluated it using receiver operating characteristic (ROC) curve analysis, calibration plots, decision curve analysis, and other methods. Additionally, we applied various machine learning techniques, including decision trees and random forests, to assess our model's predictive strength.This retrospective case‒control study involved 693 GC patients from our hospital who underwent radical gastrectomy. We collected plentiful and comprehensive clinical indicators including a total of 49 baseline, preoperative, surgical and pathological clinical data. Using univariate logistic regression, we identified potential risk factors, followed by feature selection through the Boruta algorithm. We then constructed the final predictive model using multivariate logistic regression and evaluated it using receiver operating characteristic (ROC) curve analysis, calibration plots, decision curve analysis, and other methods. Additionally, we applied various machine learning techniques, including decision trees and random forests, to assess our model's predictive strength.Univariate logistic analysis revealed 14 risk factors associated with postoperative lower limb DVT. Based on the Boruta algorithm, six significant clinical factors were selected, namely, age, D-dimer (D-D) level, low-density lipoprotein, CA125, and calcium and chloride ion levels. A nomogram was developed using the outcomes from the multivariate logistic regression analysis. The predictive model showed high accuracy, with an area under the curve of 0.936 in the training set and 0.875 in the validation set. Various machine learning algorithms confirmed its strong predictive capacity. MR analysis revealed meaningful causal relationships between key clinical factors and DVT risk.Based on various machine learning methods, we developed an effective predictive diagnostic model for postoperative lower extremity DVT in GC patients. This model demonstrated excellent predictive value in both the training and validation sets. This novel model is a valuable tool for clinicians to use in identifying and managing thrombotic risks in this patient population.

Development and validation of an interpretable machine learning for mortality prediction in patients with sepsis

IntroductionSepsis is a leading cause of death. However, there is a lack of useful model to predict outcome in sepsis. Herein, the aim of this study was to develop an explainable machine learning (ML) model for predicting 28-day mortality in patients with sepsis based on Sepsis 3.0 criteria.MethodsWe obtained the data from the Medical Information Mart for Intensive Care (MIMIC)-III database (version 1.4). The overall data was randomly assigned to the training and testing sets at a ratio of 3:1. Following the application of LASSO regression analysis to identify the modeling variables, we proceeded to develop models using Extreme Gradient Boost (XGBoost), Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) techniques with 5-fold cross-validation. The optimal model was selected based on its area under the curve (AUC). Finally, the Shapley additive explanations (SHAP) method was used to interpret the optimal model.ResultsA total of 5,834 septic adults were enrolled, the median age was 66 years (IQR, 54–78 years) and 2,342 (40.1%) were women. After feature selection, 14 variables were included for developing model in the training set. The XGBoost model (AUC: 0.806) showed superior performance with AUC, compared with RF (AUC: 0.794), LR (AUC: 0.782) and SVM model (AUC: 0.687). SHAP summary analysis for XGBoost model showed that urine output on day 1, age, blood urea nitrogen and body mass index were the top four contributors. SHAP dependence analysis demonstrated insightful nonlinear interactive associations between factors and outcome. SHAP force analysis provided three samples for model prediction.ConclusionIn conclusion, our study successfully demonstrated the efficacy of ML models in predicting 28-day mortality in sepsis patients, while highlighting the potential of the SHAP method to enhance model transparency and aid in clinical decision-making.

Coarse to superfine: can hyperspectral soil organic carbon models predict higher-resolution information?

Introduction: Modeling and mapping of soil organic carbon concentration and distribution at the pedon scale is a current knowledge gap that can be addressed by laboratory-based hyperspectral imaging and chemometric analysis of soil cores. Despite the advancements in soil organic carbon models based on hyperspectral images, it is not clear how these models will perform upon input with images at higher resolutions than those of their training sets. This study aims to measure the generalizability of a soil organic carbon model based on a test set with higher resolution hyperspectral images than that of its training set.Methods: Organic carbon contents were measured at 10 cm intervals on eight soil cores for use as the training set and at 1 cm intervals on a single core for use as the test set. Three regression models, namely, multilayer perceptron, partial least-squares, and support vector regressions, were trained and tested with the median of each hyperspectral image for each of these intervals as the training and test predictors. Permutation importance analysis was performed to explain the models.Results: The results show that although all three models had the same validation R2 of 0.92 for cross-validation on the 10 cm data, multilayer perceptron regression allowed the best generalization with a test R2 of 0.96 compared to the partial least-squares regression (0.81) and support vector regression (0.86). It was demonstrated that the multilayer perceptron model is more robust to soil surface anomalies and that it predicts soil organic carbon on the test set by learning the spectral features related to soil organic matter chromophore activity in the 950–1,150 nm region along with clay mineralogy derived from peaks at 1,400, 1,900, 2,200, 2,250, and 2,350 nm.Conclusions: This study shows that while the regression models based on hyperspectral images perform well at the 10-cm-resolution cross validation, multilayer perceptron regression shows superior generalization and robustness for a higher 1-cm-resolution test set without much loss of prediction power.

Machine learning models to predict systemic inflammatory response syndrome after percutaneous nephrolithotomy

ObjectiveThe objective of this study was to develop and evaluate the performance of machine learning models for predicting the possibility of systemic inflammatory response syndrome (SIRS) following percutaneous nephrolithotomy (PCNL).MethodsWe retrospectively reviewed the clinical data of 337 patients who received PCNL between May 2020 and June 2022. In our study, 80% of the data were used as the training set, and the remaining data were used as the testing set. Separate prediction models based on the six machine learning algorithms were created using the training set. The predictive performance of each machine learning model was determined by the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity and specificity using the testing set. We used coefficients to interpret the contribution of each variable to the predictive performance.ResultsAmong the six machine learning algorithms, the support vector machine (SVM) delivered the best performance with accuracy of 0.868, AUC of 0.942 (95% CI 0.890–0.994) in the testing set. Further analysis using the SVM model showed that prealbumin contributed the most to the prediction of the outcome, followed by preoperative urine culture, systemic immune-inflammation (SII), neutrophil to lymphocyte ratio (NLR), staghorn stones, fibrinogen, operation time, preoperative urine white blood cell (WBC), preoperative urea nitrogen, hydronephrosis, stone burden, sex and preoperative lymphocyte count.ConclusionMachine learning-based prediction models can accurately predict the possibility of SIRS after PCNL in advance by learning patient clinical data, and should be used to guide surgeons in clinical decision-making.

A diagnostic model for Parkinson’s disease based on circadian rhythm-related genes

BackgroundCircadian rhythm (CR) disturbance is intricately associated with Parkinson’s disease (PD). However, the involvement of CR-related mechanisms in the pathogenesis and progression of PD remains elusive.MethodsA total of 141 PD patients and 113 healthy participants completed CR-related clinical examinations in this study. To further investigate the CR-related mechanisms in PD, we obtained datasets (GSE7621, GSE20141, GSE20292) from the Gene Expression Omnibus database to identify differentially expressed genes between PD patients and healthy controls and further selected CR-related genes (CRRGs). Subsequently, the least absolute shrinkage and selection operator (LASSO) followed by logistic algorithms were employed to identify the hub genes and construct a diagnostic model. The predictive performance was evaluated by area under the curve (AUC), calibration curve, and decision curve analyses in the training set and external validation sets. Finally, RT‒qPCR and Western blotting were conducted to verify the expression of these hub genes in blood samples. In addition, Pearson correlation analysis was utilized to validate the association between expression of hub genes and circadian rhythm function.ResultsOur clinical observational study revealed that even early-stage PD patients exhibited a higher likelihood of experiencing sleep disturbances, nocturnal hypertension, reverse-dipper blood pressure, and reduced heart rate variability compared to healthy controls. Furthermore, 4 CR-related hub genes (AGTR1, CALR, BRM14, and XPA) were identified and subsequently incorporated as candidate biomarkers to construct a diagnostic model. The model showed satisfactory diagnostic performance in the training set (AUC = 0.941), an external validation set GSE20295 (AUC = 0.842), and our clinical centre set (AUC = 0.805). Additionally, the up-regulation of CALR, BRM14 and the down-regulation of AGTR1, XPA were associated with circadian rhythm disruption.ConclusionCR disturbance seems to occur in the early stage of PD. The diagnostic model based on CR-related genes demonstrated robust diagnostic efficacy, offering novel insights for future clinical diagnosis of PD and providing a foundation for further exploration into the role of CR-related mechanisms in the progression of PD.

Multimodal Machine Learning-Based Marker Enables Early Detection and Prognosis Prediction for Hyperuricemia.

Hyperuricemia (HUA) has emerged as the second most prevalent metabolic disorder characterized by prolonged and asymptomatic period, triggering gout and metabolism-related outcomes. Early detection and prognosis prediction for HUA and gout are crucial for pre-emptive interventions. Integrating genetic and clinical data from 421287 UK Biobank and 8900 Nanfang Hospital participants, a stacked multimodal machine learning model is developed and validated to synthesize its probabilities as an in-silico quantitative marker for hyperuricemia (ISHUA). The model demonstrates satisfactory performance in detecting HUA, exhibiting area under the curves (AUCs) of 0.859, 0.836, and 0.779 within the train, internal, and external test sets, respectively. ISHUA is significantly associated with gout and metabolism-related outcomes, effectively classifying individuals into low- and high-risk groups for gout in the train (AUC, 0.815) and internal test (AUC, 0.814) sets. The high-risk group shows increased susceptibility to metabolism-related outcomes, and participants with intermediate or favorable lifestyle profiles have hazard ratios of 0.75 and 0.53 for gout compared with those with unfavorable lifestyles. Similar trends are observed for other metabolism-related outcomes. The multimodal machine learning-based ISHUA marker enables personalized risk stratification for gout and metabolism-related outcomes, and it is unveiled that lifestyle changes can ameliorate these outcomes within high-risk group, providing guidance for preventive interventions.

Construction of the prognostic nomogram and treatment recommendation in patients with mixed endometrial carcinoma treated with hysterectomy.

The study aimed to identify the independent prognostic factors of mixed endometrial carcinoma (MEC) patients treated with hysterectomy and to explore the optimal treatment modalities for overall survival (OS) and cancer-specific survival (CSS). Using the Surveillance, Epidemiology, and End Results (SEER) database, a total of 12,848 MEC patients treated with hysterectomy were screened out. Independent prognostic factors were identified by Cox regression analysis and used to construct the nomogram. The concordance index (C-index) of OS and CSS were 0.807 and 0.834 in the training set. Validation of the nomogram revealed that the receiver operating curve (ROC) maintained good discrimination, the decision curve analysis (DCA) had a high net benefit rate, and the calibration curves showed high consistency. Patients were grouped by the nomogram formula and the number of positive regional lymph nodes (NPR-Lymph node) to evaluate the therapeutic outcomes of chemotherapy, radiotherapy, neoadjuvant treatment and lymph node operation. Survival analysis revealed that chemotherapy could improve the prognosis for OS and CSS in the high-risk group and in the group with NPR-Lymph node counts above 1 (P < 0.05). Radiotherapy was associated with better OS and CSS in the intermediate-risk and high-risk groups, and in the group with NPR-Lymph node counts above 0 (P < 0.05). Lymphadenectomy was found to prolong OS and CSS in the high-risk group (P < 0.05), while neoadjuvant treatment did not prolong OS and CSS in any group. Thus, in this study, the nomogram for MEC patients treated with hysterectomy was successfully built and validated which could effectively predict the prognosis and identify at-risk population to guide clinical decision-making. The NPR-Lymph node was identified as a potentially strong prognostic indicator with good clinical value.

An Adaptive Noisy Label-Correction Method Based on Selective Loss for Hyperspectral Image-Classification Problem

Due to the intricate terrain and restricted resources, hyperspectral image (HSI) datasets captured in real-world scenarios typically contain noisy labels, which may seriously affect the classification results. To address this issue, we work on a universal method that rectifies the labels first and then trains the classifier with corrected labels. In this study, we relax the common assumption that all training data are potentially corrupted and instead posit the presence of a small set of reliable data points within the training set. Under this framework, we propose a novel label-correction method named adaptive selective loss propagation algorithm (ASLPA). Firstly, the spectral–spatial information is extracted from the hyperspectral image and used to construct the inter-pixel transition probability matrix. Secondly, we construct the trusted set with the known clean data and estimate the proportion of accurate labels within the untrusted set. Then, we enlarge the trusted set according to the estimated proportion and identify an adaptive number of samples with lower loss values from the untrusted set to supplement the trusted set. Finally, we conduct label propagation based on the enlarged trusted set. This approach takes full advantage of label information from the trusted and untrusted sets, and moreover the exploitation on the untrusted set can adjust adaptively according to the estimated noise level. Experimental results on three widely used HSI datasets show that our proposed ASLPA method performs better than the state-of-the-art label-cleaning methods.