Objectives To develop a machine learning (ML)-based prognostic model for predicting the risk of lower extremity deep vein thrombosis (DVT) after acute stroke, with an emphasis on limb functional assessments. Methods We conducted a retrospective analysis of 225 acute stroke patients admitted within 15 days of onset between December 1, 2015, and April 30, 2025. Predictor variables were selected using collinearity diagnostics and Least Absolute Shrinkage and Selection Operator (LASSO) regression. Three machine learning survival models—Gradient Boosting Machine (GBM), Random Survival Forest (RSF), and Generalized Linear Model (GLM)—were employed to identify the most effective model. The performance of the optimal ML model was compared with that of the traditional Cox proportional hazards model using the concordance index (C-index), cumulative/dynamic area under the curve (C/D AUC), and integrated Brier score (IBS). The optimal model was further interpreted using SurvSHAP(t). Results Six variables were selected for the model: age, stroke type, gender, tension of the muscle, Brunnstrom stage (lower limb), and sitting balance. The RSF model, implemented using the Ranger algorithm, demonstrated superior performance, with an integrated Brier score (IBS) of 0.081 and a C-index of 0.841. Age and Brunnstrom stage (lower limb) were identified as the most influential predictors. Conclusion We developed an ML-based prognostic model for predicting the risk of lower limb DVT after acute stroke. Age and Brunnstrom stage (lower limb) were the most significant predictors. This model shows promise for risk stratification in clinical practice.

PURPOSE:We aimed to develop a machine learning model to predict activities of daily living (ADL) at discharge in stroke patients and identify key predictors to guide rehabilitation decisions. MATERIALS AND METHODS:Data of 589 stroke inpatients (2019-2024) were split into good (BI ≥ 60) and poor (BI < 60) ADL groups. Continuous variables were processed using Z-score normalization, followed by preliminary univariate regression screening (P < 0.05) and final feature selection via LASSO regression (lambda.1se = 0.0488). The screened features were used to train and validate ten machine learning algorithms; 30% of the dataset (n = 177) was allocated as an independent test set for model evaluation, and SHAP analysis was performed to interpret the optimal model. RESULTS:Six of 41 features were retained. Random forest achieved the best performance (AUC = 0.958; accuracy = 0.936; sensitivity = 0.934; specificity = 0.950). SHAP identified the top drivers: admission Barthel Index, standing balance, Brunnstrom stages (upper and lower limb), dressing, and grooming abilities. CONCLUSION:The ADL risk prediction model constructed using machine learning, particularly the random forest model, shows excellent predictive performance and clinical interpretability, making it valuable for individualized risk assessment of daily living skills in stroke patients at discharge.

This study aims to develop a prediction model based on machine learning algorithms to predict the risk of stroke by analyzing physical activity and other risk factors. We conducted a retrospective analysis of 134 stroke patients treated at the Department of Neurology, the First Affiliated Hospital of Nanjing Medical University, from July 1, 2021, to May 31, 2023, and 354 non-stroke individuals recruited from the Fenghuang Community Health Screening Program in Nanjing during the same period. Eight machine learning models, including extreme gradient boosting, support vector machine, random forest (RF), neural network, Naive Bayesian, logistic regression, K-nearest neighbor, and decision tree, were used to build the prediction models. Variables were selected using the least absolute shrinkage and selection operator and the multivariable logistic regression analysis. The models were evaluated using receiver operating characteristic (ROC) curves, area under the ROC curve, precision-recall (PR) curves, area under the PR curve, accuracy, sensitivity, specificity, and precision. Shapley Additive Explanations were employed to determine feature importance. The results demonstrated that the RF algorithm performed well in terms of area under the ROC curve (0.96), area under the PR curve (0.92), specificity (0.97), and precision (0.92). Shapley Additive Explanations analysis revealed that the number of weekly exercise days had the most significant impact on stroke risk, followed by calf circumference, past medical history, gender, body mass index, and Strength, Assistance with walking, Rising from a chair, Climbing stairs, and Falls score. The RF algorithm demonstrated strong predictive performance for stroke risk and may guide clinical decision-making.

This study aimed to apply machine learning (ML) techniques to develop and validate a risk prediction model for post-stroke lower extremity deep vein thrombosis (DVT) based on patients’ limb function, activities of daily living (ADL), clinical laboratory indicators, and DVT preventive measures. We retrospectively analyzed 620 stroke patients. Eight ML models—logistic regression (LR), support vector machine (SVM), random forest (RF), decision tree (DT), neural network (NN), extreme gradient boosting (XGBoost), Bayesian (NB), and K-nearest neighbor (KNN)—were used to build the model. These models were extensively evaluated using ROC curves, AUC, PR curves, PRAUC, accuracy, sensitivity, specificity, and clinical decision curves (DCA). Shapley’s additive explanation (SHAP) was used to determine feature importance. Finally, based on the optimal ML algorithm, different functional feature set models were compared with the Padua scale to select the best feature set model. Our results indicated that the RF algorithm demonstrated superior performance in various evaluation metrics, including AUC (0.74/0.73), PRAUC (0.58/0.58), accuracy (0.75/0.77), and sensitivity (0.78/0.80) in both the training set and test set. DCA analysis revealed that the RF model had the highest clinical net benefit. SHAP analysis showed that D-dimer had the most significant influence on DVT, followed by age, Brunnstrom stage (lower limb), prothrombin time (PT), and mobility ability. The RF algorithm can predict post-stroke DVT to guide clinical practice.

Background: Cervical lymph node metastasis in papillary thyroid carcinoma plays a crucial role in the development of surgical strategy for thyroid patients. The aim of this study was to determine the predictors of cervical lymph node metastasis based on ultrasound features of papillary thyroid carcinoma, and to develop and validate nomogram to help predict cervical lymph node metastasis. Methods: Patients who underwent thyroid ultrasound examination in Department of Ultrasonography of The First Affiliated Hospital of Nanjing Medical University between January 1, 2021 and October 31, 2021 were selected. Patients with at least one Thyroid Imaging Reporting and Data System (TI-RADS) class 4 or higher nodule and postoperative pathologically confirmed primary papillary thyroid carcinoma with cervical lymph node metastasis were identified or not, and ultrasound image characteristics of the nodules were recorded to screen for cervical lymph node metastasis predictors. Subsequently, nomogram was developed and validated to help predict cervical lymph node metastasis. Results: The overall echogenicity of the thyroid gland, the number of malignant nodules, nodule left-right diameter, the location of the nodules, the relationship between the nodules and the thyroid capsule, and the elasticity score of the nodules were considered to be independent predictors of papillary thyroid carcinoma related cervical lymph node metastasis; the model had a good discrimination rate. Conclusions: We developed a nomogram to predict metastasis in the neck lymph nodes of papillary thyroid carcinoma, and the nomogram showed good performance for prediction aspects.

Objectives: To develop a nomogram for predicting calf muscle veins thrombosis (CMVT) in stroke patients during rehabilitation. Methods: We enrolled 360 stroke patients from the Rehabilitation Medicine Center from December 2015 to February 2019. Of the participants, 123 were included in the CMVT group and 237 in the no CMVT group. The least absolute shrinkage and selection operator (LASSO) regression model was applied to optimize feature selection for the model. Multivariable logistic regression analysis was applied to construct a predictive model. Performance and clinical utility of the nomogram were generated using the Harrell's concordance index, calibration curve, and decision curve analysis (DCA). Results: Age, Brunnstrom stage (lower extremity), D-dimer, and antiplatelet therapy were associated with the occurrence of CMVT. The prediction nomogram showed satisfactory performance with a concordance index of 0.718 (95% CI: 0.663-0.773) in internal verification. The Hosmer-Lemeshow test, P = .217, suggested that the model was of goodness-of-fit. In addition, the DCA demonstrated that the CMVT nomogram had a good clinical net benefit. Conclusions: We developed a nomogram that could help clinicians identify high-risk groups of CMVT in stroke patients during rehabilitation for early intervention.

Objective:Muscle weakness and spasticity are common consequences of stroke, leading to a decrease in physical activity. The effective implementation of precision rehabilitation requires detailed rehabilitation evaluation. We aimed to analyze the surface electromyography (sEMG) signal features of elbow flexor muscle (biceps brachii and brachioradialis) spasticity in maximum voluntary isometric contraction (MVIC) and fast passive extension (FPE) in stroke patients and to explore the main muscle groups that affect the active movement and spasticity of the elbow flexor muscles to provide an objective reference for optimizing stroke rehabilitation. Methods:Fifteen patients with elbow flexor spasticity after stroke were enrolled in this study. sEMG signals of the paretic and nonparetic elbow flexor muscles (biceps and brachioradialis) were detected during MVIC and FPE, and root mean square (RMS) values were calculated. The RMS values (mean and peak) of the biceps and brachioradialis were compared between the paretic and nonparetic sides. Additionally, the correlation between the manual muscle test (MMT) score and the RMS values (mean and peak) of the paretic elbow flexors during MVIC was analyzed, and the correlation between the modified Ashworth scale (MAS) score and the RMS values (mean and peak) of the paretic elbow flexors during FPE was analyzed. Results:During MVIC exercise, the RMS values (mean and peak) of the biceps and brachioradialis on the paretic side were significantly lower than those on the nonparetic side (p < 0.01), and the RMS values (mean and peak) of the bilateral biceps were significantly higher than those of the brachioradialis (p < 0.01). The MMT score was positively correlated with the mean and peak RMS values of the paretic biceps and brachioradialis (r = 0.89, r = 0.91, r = 0.82, r = 0.85; p < 0.001). During FPE exercise, the RMS values (mean and peak) of the biceps and brachioradialis on the paretic side were significantly higher than those on the nonparetic side (p < 0.01), and the RMS values (mean and peak) of the brachioradialis on the paretic side were significantly higher than those of the biceps (p < 0.01). TheMAS score was positively correlated with the mean RMS of the paretic biceps and brachioradialis (r = 0.62, p = 0.021; r = 0.74, p = 0.004), and the MAS score was positively correlated with the peak RMS of the paretic brachioradialis (r = 0.59, p = 0.029) but had no significant correlation with the peak RMS of the paretic biceps (r = 0.49, p > 0.05). Conclusions:The results confirm that the biceps is a vital muscle in active elbow flexion and that the brachioradialis plays an important role in elbow flexor spasticity, suggesting that the biceps should be the focus of muscle strength training of the elbow flexors and that the role of the brachioradialis should not be ignored in the treatment of elbow flexor spasticity. This study also confirmed the application value of sEMG in the objective assessment of individual muscle strength and spasticity in stroke patients.

To develop and validate a nomogram for individualized prediction of lower extremity deep venous thrombosis (DVT) in stroke patients based on extremity function and daily living ability of stroke patients. In this study, 423 stroke patients admitted to the Rehabilitation Medical Center of the First Affiliated Hospital of Nanjing Medical University from December 2015 to February 2019 were taken as the subjects, who were divided into the DVT group (110) and No-DVT group (313) based on the existence of DVT. Inter-group comparison of baseline data was performed by 1-way Analysis of Variance, Kruskal-Wallis rank-sum test, or Pearson chi-square test. Data dimensions and predictive variables were selected by least absolute shrinkage and selection operator (LASSO); the prediction model was developed and the nomogram was prepared by binary logistics regression analysis; the performance of the nomogram was identified by the area under the receiver operating characteristic curve (AUC), Harrell's concordance index, and calibration curve; and the clinical effectiveness of the model was analyzed by clinical decision curve analysis. Age, Brunnstrom stage (lower extremity), and D-dimer were determined to be the independent predictors affecting DVT. The independent predictors mentioned above were developed and presented as a nomogram, with AUC and concordance index of 0.724 (95% confidence interval [CI]: 0.670-0.777), indicating the satisfactory discrimination ability of the nomogram. The P value of the results of the Hosmer-Lemeshow test was 0.732, indicating good fitting of the prediction model. Decision curve analysis showed that the clinical net benefit of this model was 6% to 50%. We developed a nomogram to predict lower extremity deep venous thrombosis in stroke patients, and the results showed that the nomogram had satisfactory prediction performance and clinical efficacy.

Hemophagocytic lymphohistiocytosis (HLH) is a rare systematic immune disease manifested with excessive activation of lymphocytes and macrophages. This study was designed to explore the feasible prognostic factors of secondary HLH (sHLH). We retrospectively analyzed 179 patients with newly diagnosed sHLH from January 2016 to May 2019 according to the HLH-2004 protocol. Baseline characteristics and laboratory results were reviewed. The median age of all patients was 53 years, with a male/female ratio of 1.45. The commonest cause of HLH was malignancy. Of the 179 patients, 48.6% presented with Epstein-Barr virus (EBV) infection, 92.8% with hemocytopenia (at least 2 lineages), 60.3% with hypofibrinogenemia, 43.0% with hypertriglyceridemia (≥ 3 mmol/L), 99.4% with high ferritin, 97.8% with fever, 72.1% with splenomegaly, and 72.6% with hemophagocytosis. As to their prognosis, 122 patients died; the median survival was 88 days, with a 2-year survival rate of 26.72%. Univariate analysis confirmed neutrophil-to-lymphocyte ratio ˃ 2.53, lymphocyte-to-monocyte ratio (LMR) ≤ 4.43, platelet-to-lymphocyte ratio ˃ 227.27, red blood cell distribution width ˃ 14.6, red blood cell distribution width-to-platelet ratio (RPR) > 0.33, EBV infection, platelet ≤ 34 × 109 /L, fibrinogen ≤ 1.34 g/L, alkaline phosphatase ˃ 182.4 U/L, adenosine deaminase ˃ 69.2 U/L, and ferritin ˃ 2318 ng/mL were associated with an inferior survival. In a multivariate model, LMR, RPR, and ferritin were considered as three independent factors. Some blood-based inflammatory markers, which can be easily and cheaply detected, are significantly associated with the OS of HLH patients. LMR and RPR, superior to NLR, PLR, RDW, can be taken to predict the OS of patients with HLH.

OBJECTIVE:To investigate the expression and clinical significance of soluble B7-H3 (sB7-H3) in patients with secondary hemophagocytic lymphohistiocytosis (sHLH). METHODS:The plasma samples of 85 newly diagnosed sHLH patients from December 2012 to April 2018 were collected. The patients were divided into lymphoma-related HLH（LHLH）group and infection-related HLH（IHLH）group. The expression of sB7-H3 in plasma was detected by ELISA, and the clinical data were collected for analysis. Fifteen healthy people were chosen as control group. RESULTS:The expression level of sB7-H3 in lymphoma-related HLH and infection-related HLH group significant increased as compared with the control group, (P＜0.05), and the expression level of sB7-H3 in lymphoma-related HLH group was significant higher than that in infection-related HLH group [(35.75± 9.90) vs (28.70±8.95) ng/ml)] (P＜0.001). There were no significant statistical difference in the expression of some clinical factors (including age, fever, splenomegaly, ANC, Plt, FIB, calcium ion, serum albumin, LDH, serum ferritin, sCD25) in lymphoma-related HLH and infection-related HLH group (P＞0.05). The evaluation of expression and significance of sB7-H3 in sHLH by using ROC curve, showed that the area under ROC curve comparison of patients in lymphoma-related HLH group and infection-related HLH group was 0.718 (95% CI 0.610-0.810) (P=0.0002), and predicting the sensitivity and specificity of the lymphoma-related HLH patients were 77.36% and 59.38%, respectively. The best cut-off value of patients in sB7-H3 was 29.81 ng/ml, the overall survival time of sB7-H3 high-expression group (≥29.81 ng/ml) was significant shorter than that in low-expression group (＜29.81 ng/ml) (24 vs 440 d) (P＜0.001). The clinical factors affecting the survival status of sHLH patients were neutrophils, albumin, serum ferritin, serum calcium ions and sB7-H3 levels. CONCLUSION:sB7-H3 associates with poor prognosis of sHLH patients, and may be involved in disease progression.

The clinical features of patients with secondary hemophagocytic lymphohistiocytosis (sHLH) complicated with capillary leak syndrome (CLS) remain controversial. The data of 259 sHLH patients were retrospectively analyzed. The clinical manifestations, laboratory findings, treatment, and prognosis of the CLS-sHLH group and non-CLS-sHLH group were compared. The levels of fibrinogen, albumin, and serum calcium in the CLS-sHLH group were lower than in the non-CLS-sHLH group, and serum triglycerides in the CLS-sHLH group were higher than in the non-CLS-sHLH group (P < 0.05). Univariate analysis showed that fibrinogen level was an independent prognostic factor in sHLH patients complicated with CLS. The median survival time was significantly shorter in patients with fibrinogen ≤ 1.3 g/L than in patients with fibrinogen > 1.3 g/L (P < 0.05). Patients with improved CLS conditions in the CLS-sHLH group had significantly increased albumin and serum calcium after treatment (P < 0.05); patients without improved conditions in the CLS-sHLH group also had significantly increased albumin after treatment (P < 0.05), but the serum calcium did not change significantly (P > 0.05). sHLH patients complicated with CLS had significantly worse prognosis than without CLS. Significant reduction in fibrinogen may be an independent prognostic factor for poor prognosis in sHLH patients complicated with CLS.