In order to solve the pathological grading of hepatocellular carcinomas (HCC) which depends on biopsy or surgical pathology invasively, a quantitative analysis method based on radiomics signature was proposed for pathological grading of HCC in non-contrast magnetic resonance imaging (MRI) images. The MRI images were integrated to predict clinical outcomes using 328 radiomics features, quantifying tumour image intensity, shape and text, which are extracted from lesion by manual segmentation. Least absolute shrinkage and selection operator (LASSO) were used to select the most-predictive radiomics features for the pathological grading. A radiomics signature, a clinical model, and a combined model were built. The association between the radiomics signature and HCC grading was explored. This quantitative analysis method was validated in 170 consecutive patients (training dataset: n = 125; validation dataset, n = 45), and cross-validation with receiver operating characteristic (ROC) analysis was performed and the area under the ROC curve (AUC) was employed as the prediction metric. Through the proposed method, AUC was 0.909 in training dataset and 0.800 in validation dataset, respectively. Overall, the prediction performances by radiomics features showed statistically significant correlations with pathological grading. The results showed that radiomics signature was developed to be a significant predictor for HCC pathological grading, which may serve as a noninvasive complementary tool for clinical doctors in determining the prognosis and therapeutic strategy for HCC.
ObjectiveTo characterize surgical outcomes of supra-cardiac total anomalous pulmonary venous connection (TAPVC), investigate risk factors for postoperative death, and explore informed LASSO machine learning methods to solve "small sample size problem" in research of rare congenital heart diseases.MethodsA retrospective analysis of 241 patients with supra-cardiac TAPVC who underwent surgical repair in Guangdong Provincial People's Hospital from 2009 to 2019 was conducted, including 179 males and 62 females with a median surgical age of 71 (33, 232) d.Detailed clinical data of the postoperative death-related factors were extracted. Univariable Cox proportional hazard models were used to initially screen potential risk factors for postoperative death. Factors with P≤0.05 were retained. To solve the limitation of small sample size and the "P>n" problem, we proposed a novel LASSO method for conducting multivariable Cox regression analysis that was capable of bringing in findings of related studies to improve analysis power and to reduce false-negative findings.ResultsUnivariable Cox analyses showed several potential clinical risk factors, among which highly significant factors (P<0.001) included: surgical weight≤2.5 kg (HR=16.00), main pulmonary artery diameter (HR=0.78), prolonged cardiopulmonary bypass time (HR=1.21), aortic block time (HR=1.28), and postoperative ventilator-assisted time (HR=1.13/d). LASSO multivariable analysis revealed that independent risk factors for postoperative death included cardiopulmonary bypass time (aHR=1.308/30 min), age (aHR=0.898), postoperative ventilator-assisted time (aHR=1.023/d), weight≤2.5 kg (aHR=2.545), right vertical venous return (aHR=1.977), preoperative pulmonary venous obstruction (aHR=1.633) and emergency surgery (aHR=1.383).ConclusionOur proposed informed LASSO method can use previous studies' results to improve the power of analysis and effectively solve the "P>n" and small sample size limitation. Cardiopulmonary bypass time, surgical age, postoperative ventilator-assisted time, weight, right vertical venous return, preoperative pulmonary venous obstruction, and emergency surgery are risk factors for postoperative death of supra-cardiac TAPVC.
Objective To evaluate the risk factors for postoperative in-hospital mortality in elderly patients receiving cardiac valvular surgery, and develop a new prediction models using the least absolute shrinkage and selection operator (LASSO)-logistic regression. Methods The patients≥65 years who underwent cardiac valvular surgery from 2016 to 2018 were collected from the Chinese Cardiac Surgery Registry (CCSR). The patients who received the surgery from January 2016 to June 2018 were allocated to a training set, and the patients who received the surgery from July to December 2018 were allocated to a testing set. The risk factors for postoperative mortality were analyzed and a LASSO-logistic regression prediction model was developed and compared with the EuroSCOREⅡ. Results A total of 7 163 patients were collected in this study, including 3 939 males and 3 224 females, with a mean age of 69.8±4.5 years. There were 5 774 patients in the training set and 1389 patients in the testing set. Overall, the in-hospital mortality was 4.0% (290/7163). The final LASSO-logistic regression model included 7 risk factors: age, preoperative left ventricular ejection fraction, combined coronary artery bypass grafting, creatinine clearance rate, cardiopulmonary bypass time, New York Heart Association cardiac classification. LASSO-logistic regression had a satisfying discrimination and calibration in both training [area under the curve (AUC)=0.785, 0.627] and testing cohorts (AUC=0.739, 0.642), which was superior to EuroSCOREⅡ. Conclusion The mortality rate for elderly patients undergoing cardiac valvular surgery is relatively high. LASSO-logistic regression model can predict the risk of in-hospital mortality in elderly patients receiving cardiac valvular surgery.
ObjectiveTo explore the risk factors for postoperative respiratory failure (RF) in patients with esophageal cancer, construct a predictive model based on least absolute shrinkage and selection operator (LASSO)-logistic regression, and visualize the constructed model. MethodsA retrospective analysis was conducted on patients with esophageal cancer who underwent surgical treatment in the Department of Thoracic Surgery, Sun Yat-sen University Cancer Center Gansu Hospital from 2020 to 2023. Patients were divided into a RF group and a non-RF (NRF) group according to whether RF occurred after surgery. Clinical data of the two groups were collected, and LASSO-logistic regression was used to optimize feature selection and construct the predictive model. The model was internally validated by repeated sampling 1000 times based on the Bootstrap method. ResultsA total of 217 patients were included, among which 24 were in the RF group, including 22 males and 2 females, with an average age of (63.33±9.10) years; 193 were in the NRF group, including 161 males and 32 females, with an average age of (62.14±8.44) years. LASSO-logistic regression analysis showed that the ratio of forced expiratory volume in one second/forced vital capacity (FEV1/FVC) to predicted value (FEV1/FVC%pred) [OR=0.944, 95%CI (0.897, 0.993), P=0.026], postoperative anastomotic fistula [OR=4.106, 95%CI (1.457, 11.575), P=0.008], and postoperative lung infection [OR=3.776, 95%CI (1.373, 10.388), P=0.010] were risk factors for postoperative RF in patients with esophageal cancer. Based on the above risk factors, a predictive model was constructed, with an area under the receiver operating characteristic curve of 0.819 [95%CI (0.737, 0.901)]. The Hosmer-Lemeshow test for the calibration curve showed that the model had good goodness of fit (P=0.527). The decision curve showed that the model had good clinical net benefit when the threshold probability was between 5% and 50%. Conclusion FEV1/FVC%pred, postoperative anastomotic fistula, and postoperative lung infection are risk factors for postoperative RF in patients with esophageal cancer. The predictive model constructed based on LASSO-logistic regression analysis is expected to help medical staff screen high-risk patients for early individualized intervention.
Features and interaction between features of liver disease is of great significance for the classification of liver disease. Based on least absolute shrinkage and selection operator (LASSO) and interaction LASSO, the generalized interaction LASSO model is proposed in this paper for liver disease classification and compared with other methods. Firstly, the generalized interaction logistic classification model was constructed and the LASSO penalty constraints were added to the interactive model parameters. Then the model parameters were solved by an efficient alternating directions method of multipliers (ADMM) algorithm. The solutions of model parameters were sparse. Finally, the test samples were fed to the model and the classification results were obtained by the largest statistical probability. The experimental results of liver disorder dataset and India liver dataset obtained by the proposed methods showed that the coefficients of interaction features of the model were not zero, indicating that interaction features were contributive to classification. The accuracy of the generalized interaction LASSO method is better than that of the interaction LASSO method, and it is also better than that of traditional pattern recognition methods. The generalized interaction LASSO method can also be popularized to other disease classification areas.