Objective To construct an innovative CNN-Transformer dual-stream parallel network architecture integrating clinical data and imaging features for improving the predictive accuracy of pectoralis major myofascial metastasis in breast cancer,and to optimize the model performance by screening the optimal feature subset through genetic algorithm.Methods The proposed architecture concurrently processed clinical records and imaging data,including physical characteristics such as resolution,contrast,grayscale distribution,and texture features to identify their latent correlations.Meanwhile,genetic algorithms were employed to remove redundant features while retaining the most clinically and physically relevant features for pectoralis major myofascial metastasis prediction.Results The CNN-Transformer model that integrated imaging and clinical features showed superior performance across all evaluation metrics such as weighted F1 score and AUCROC,outperforming models relying only on imaging or clinical data.Conclusion The proposed dual-stream parallel network architecture combined with feature selection strategy significantly enhances the predictive accuracy of pectoralis major myofascial metastasis in breast cancer,and demonstrates the critical role of imaging features in improving model performance.

Objective To construct an innovative CNN-Transformer dual-stream parallel network architecture integrating clinical data and imaging features for improving the predictive accuracy of pectoralis major myofascial metastasis in breast cancer,and to optimize the model performance by screening the optimal feature subset through genetic algorithm.Methods The proposed architecture concurrently processed clinical records and imaging data,including physical characteristics such as resolution,contrast,grayscale distribution,and texture features to identify their latent correlations.Meanwhile,genetic algorithms were employed to remove redundant features while retaining the most clinically and physically relevant features for pectoralis major myofascial metastasis prediction.Results The CNN-Transformer model that integrated imaging and clinical features showed superior performance across all evaluation metrics such as weighted F1 score and AUCROC,outperforming models relying only on imaging or clinical data.Conclusion The proposed dual-stream parallel network architecture combined with feature selection strategy significantly enhances the predictive accuracy of pectoralis major myofascial metastasis in breast cancer,and demonstrates the critical role of imaging features in improving model performance.

Objective To construct and validate the risk prediction model for the occurrence of caesarean scar pregnancy(CSP)in women with re-pregnancy after cesarean section.Methods A total of 663 women with re-pregnancy after cesarean section in Urumqi Maternal and Child Health Hospital from 2018 to 2022 were collected,and randomly divided the training set(n=460)and the test set(n=203)according to 7∶3,the cases of the training set were divided into the CSP group(n=239)and the non-CSP group(n=221),and the risk factors for the occurrence of CSP were evaluated by univariate and multivariate Logistic regression analysis.Based on the a-bove results,a nomogram model was constructed,validated and evaluated in the test set and the training set,respectively.The predictive efficacy of the model was evaluated by area under the curve(AUC)of receiver operating characteristic(ROC)and the Hosmer-Leme-show test,and the clinical application value of the model was evaluated by clinical decision curve analysis(DCA).Results The results of multivariate Logistic regression analysis showed that the number of cesarean section＞1,posterior uterine position,the number of mis-carriages＞1,CSD,the history of miscarriage between the current pregnancy and the previous cesarean section were the risk factors for the occurrence of CSP(P＜0.05),and the timing of cesarean section was the protective factor for the occurrence of CSP in the course of labor(P＜0.05).Based on the above results,the nomogram prediction model was constructed,the AUC of the model in the training set was 0.813(95％CI:0.773-0.852),and the AUC of the model in the test set was 0.817(95％CI:0.755-0.878).Hosmer-Lemeshow goodness-of-fit test for the training set and the test set model was well fitted(x2=7.647,P=0.469;x=6.162,P=0.629).The calibration curve showed that the model had good consistency in predicting the occurrence of CSP in re-pregnancy after cesarean section,and the DCA curve showed that the model had high clinical efficacy in both the training set and the test set.Conclusion The prediction model constructed in this study can effectively predict the occurrence of CSP,which can provide references for early identification and pre-ventive treatment for high-risk populations.

Reflex tests are ordered when a particular test result indicates that additional testing should be performed according to the guidelines or the feedback process formulated by clinical consultation. The application scope of the reflex tests involves various subspecialties of laboratory medicine. The clinical application needs the support of qualified laboratory doctors, comprehensive information and financial system, clinical guidelines, and so on. Active application of reflex tests can promote the standardization of evidence-based medicine in clinical practice, save medical resources, and shorten the diagnosis and treatment time of patients.