Machine learning-based fusion model in predicting HER2 expression of breast cancer by Sonazoid-enhanced ultrasound: A multicenter study

Huiting Zhang¹Manlin Lang¹Huiming Shen²Hang Li³Ning Yang⁴Bo Chen⁵Yixu Chen⁶Hong Ding⁷Weiping Yang⁸Xiaohui Ji⁹Ping Zhou¹⁰Cui Li Gang¹¹Jiandong Wang¹²Wentong Xu¹²Xiuqin Ye^13,14Liu Zhixing¹⁵Yu Yang¹⁶Tianci Wei¹⁷Hui Wang¹⁸Yuanyuan Yan¹⁹Changjun Wu²⁰Yi Yun Wu²¹Jingwen Shi²²Yaxi Wang²³Xiuxia Fang²³Ran Li²⁴Ping Liang¹Jie Yu^1*

• ¹PLA Medical College & Chinese PLA General Hospital, Beijing, China
• ²Zhongda Hospital, Southeast University, Nanjing, Jiangsu Province, China
• ³Affiliated Hospital of Putian University, Putian, Fujian Province, China
• ⁴Xingcheng People’s Hospital, Xingcheng, China
• ⁵Lu’ an people’s Hospital of Anhui Province, Liuan, China
• ⁶Chengdu Fifth People's Hospital, Chengdu, Sichuan Province, China
• ⁷Huashan Hospital, Fudan University, Shanghai, Shanghai Municipality, China
• ⁸Guangxi Medical University Cancer Hospital, Nanning, Guangxi Zhuang Region, China
• ⁹Fourth Hospital of Hebei Medical University, Shijiazhuang, Hebei Province, China
• ¹⁰Department of Ultrasound, The Third Xiangya Hospital, Central South University, Changsha, Hunan Province, China
• ¹¹Peking University Third Hospital, Haidian, Beijing Municipality, China
• ¹²People's Liberation Army General Hospital, Beijing, Beijing Municipality, China
• ¹³First Affiliated Hospital of Southern University of Science and Technology, Shenzhen, China
• ¹⁴Shenzhen People's Hospital, Jinan University, Shenzhen, Guangdong Province, China
• ¹⁵The First Affiliated Hospital of Nanchang University, Nanchang, Jiangxi Province, China
• ¹⁶Beijing Friendship Hospital, Capital Medical University, Beijing, Beijing Municipality, China
• ¹⁷The Second Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang Province, China
• ¹⁸China-Japan Union Hospital, Jilin University, Changchun, Jilin Province, China
• ¹⁹Zhengzhou Central Hospital, Zhengzhou, Henan Province, China
• ²⁰First Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang Province, China
• ²¹Affiliated Hospital of Nanjing University of Chinese Medicine, Nanjing, Liaoning Province, China
• ²²Sheng Jing Hospital Affiliated, China Medical University, Shenyang, Liaoning Province, China
• ²³The Affiliated Hospital of Inner Mongolia Medical University, Hohhot, Inner Mongolia Autonomous Region, China
• ²⁴The First Affiliated Hospital of Xinxiang Medical University, Xinxiang, Henan Province, China

Purpose: To predict human epidermal growth factor receptors 2 (HER2) expression of breast cancer by Sonazoid-enhanced ultrasound in machine learning based model.: From August 2020 to February 2021, breast cancer patients underwent surgical treatment without neoadjuvant chemotherapy were enrolled prospectively from 17 hospitals in China. HER2 expression status was assessed by immunohistochemistry or fluorescence in situ hybridization (FISH). The training set contained data from 11 hospitals, and the validation set contained 6 hospitals. Features of clinical, B-mode ultrasound, contrast enhanced ultrasound and time intensity curve were selected by Least Absolute Shrinkage and Selection Operator.Based on the selected features, six prediction models including logistic regression (LR), support vector machine (SVM), random forest (RF), eXtreme Gradient Boosting (XGB), XGB combined with LR, and the fusion model were established to predict HER2 3+ and 2+/1+ expression, respectively.Results: A total of 140 breast cancer patients were enrolled in the study. Seven features related to HER2 3+ and six features related to HER2 2+/1+ were selected to establish the prediction models. Among six models, LR, SVM and XGB showed better prediction performance for both HER2 3+ and HER2 2+/1+ cases. The three models were targeted into a fusion model. In the validation, the fusion model achieved the highest value of area under the receiver operating characteristic curve as 0.869 (95%CI: 0.715 -0.958) for predicting HER2 3+ and 0.747 (95%CI: 0.548 -0.891) for predicting HER2 2+/1+ cases. The model could upgrade HER2 2+ cases to