Objective To explore the diagnostic efficacy of machine learning in predicting perineural invasion(PNI)of invasive breast cancer based on MRI imaging features of breast cancer.Methods The data of 294 patients with invasive breast cancer confirmed by surgical pathology were retrospectively analyzed,and the patients were randomly divided into training set(205 cases,PNI 77 cases)and validation set(89 cases,PNI 33 cases)at a ratio of 7∶3.10 machine learning models were constructed by selecting training set clinical and radiographic features using single factor logistic regression.The area under the curve(AUC),accuracy(ACC),sensitivity(SE),specificity(SP),positive predictive value(PPV),and negative predictive value(NPV)were used to evaluate the predictive effi-cacy of different models for PNI,and the best model was determined.SHapley Additive exPlanation(SHAP)was used to visuaize the diagnosis process of the model.Results In the validation set,the multi-layer perceptron(MLP)model performed best,with AUC,ACC,SE,SP,PPV,and NPV of 0.91,0.89,0.79,0.95,0.90,and 0.88,respectively.Conclusion The model of MRI imaging fea-tures of breast cancer constructed by MLP machine learning model can effectively predict the preoperative PNI of invasive breast cancer.

Objective To explore the diagnostic efficacy of machine learning in predicting perineural invasion(PNI)of invasive breast cancer based on MRI imaging features of breast cancer.Methods The data of 294 patients with invasive breast cancer confirmed by surgical pathology were retrospectively analyzed,and the patients were randomly divided into training set(205 cases,PNI 77 cases)and validation set(89 cases,PNI 33 cases)at a ratio of 7∶3.10 machine learning models were constructed by selecting training set clinical and radiographic features using single factor logistic regression.The area under the curve(AUC),accuracy(ACC),sensitivity(SE),specificity(SP),positive predictive value(PPV),and negative predictive value(NPV)were used to evaluate the predictive effi-cacy of different models for PNI,and the best model was determined.SHapley Additive exPlanation(SHAP)was used to visuaize the diagnosis process of the model.Results In the validation set,the multi-layer perceptron(MLP)model performed best,with AUC,ACC,SE,SP,PPV,and NPV of 0.91,0.89,0.79,0.95,0.90,and 0.88,respectively.Conclusion The model of MRI imaging fea-tures of breast cancer constructed by MLP machine learning model can effectively predict the preoperative PNI of invasive breast cancer.

Objective To explore the value of spiral CT in venous phase in measuring the diameter of bilateral transverse sinus.Methods The CT vascular imaging findings of 200 cases at torcular herophili area in our hospital were analyzed retrospectively.The resource images, volume rendering (VR) and maximum intensity projection (MIP) were performed to observe the presentation of vascular anatomy in the torcular herophili area, and to determine the dominant transverse sinus and types of torcular herophili (typeⅠ-Ⅵ).The diameters of bilateral transverse sinus were measured by original CT images.Two groups were categorized according to the genders, and four ones (20-30 years, 30-40 years, 40-50 years and 50-60 years) according to the age.Results Between different genders, there was significant difference in right transverse diameter (P<0.05), and there was no significant difference in left transverse diameter (P>0.05).There was no significant difference among four age groups in bilateral transverse diameter (P>0.05).Conclusion Spiral CT is helpful for the evaluation of the diameter of bilateral transverse sinus.These findings can provide anatomical basis for clinical disease with significant implication.