Prediction of Ki-67 expression in pituitary adenoma using a joint model based on siamese network and transfer learning
10.3969/j.issn.1002-1671.2025.11.002
- VernacularTitle:基于孪生网络和迁移学习的联合模型预测垂体腺瘤Ki-67表达
- Author:
Xue GE
1
;
Jin DUAN
;
Xiuling WANG
;
Lu TANG
;
Chunfeng HU
;
Kai XU
;
Qian XU
Author Information
1. 徐州医科大学附属医院影像科,江苏 徐州 221000
- Publication Type:Journal Article
- Keywords:
pituitary adenoma;
magnetic resonance imaging;
Ki-67 index;
siamese network;
transfer learning
- From:
Journal of Practical Radiology
2025;41(11):1769-1772,1790
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the prediction efficiency of Ki-67 expression status in pituitary adenoma(PA)based on a joint model of siamese network and transfer learning.Methods The preoperative MR T1WI enhanced sequence images of 370 patients with PA diagnosed by surgery and pathology were retrospectively collected.According to the results of immunohistochemical,all patients were divided into high proliferation index group(Ki-67≥3,n=97)and low proliferation index group(Ki-67＜3,n=273),and all the sample data were randomly divided into training set and test set according to the ratio of 7∶3.Two single predictive models,ResNet-50 and VGGNet-16 and combined them with the siamese network and transfer learning were built.The prediction efficiency of different models were evaluated via accuracy rate,precision rate,recall rate,F1 score and receiver operating characteristic(ROC)curve as the main criteria.Results Compared with a single predictive model,the model combined with siamese network and transfer learning showed a better performance for predicting the Ki-67 of PA.Additionally,the ResNet-50 joint model exhibited the superior predictive performance.The accuracy rate was 0.872 7,the precision rate was 0.812 5,the recall rate was 0.764 7,the F1 score was 0.787 9,and the area under the curve(AUC)was 0.841 6.Conclusion The joint model based on siamese network and transfer learning exhibits a higher efficiency for predicting the Ki-67 expression status in PA,which can help the clinicians to formulate more personalized treatment for the patients.
