AUTHOR=Wu Huaiyu , Ye Xiuqin , Jiang Yitao , Tian Hongtian , Yang Keen , Cui Chen , Shi Siyuan , Liu Yan , Huang Sijing , Chen Jing , Xu Jinfeng , Dong Fajin 

TITLE=A Comparative Study of Multiple Deep Learning Models Based on Multi-Input Resolution for Breast Ultrasound Images

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.869421

DOI=10.3389/fonc.2022.869421

ISSN=2234-943X

ABSTRACT=Purpose The purpose of this study was to explore the performance of different parameter combinations of deep learning (DL) models (Xception, DenseNet121, MobileNet, ResNet50 and EfficientNetB0) and input image resolutions (REZs) (224 × 224, 320 × 320 and 488 × 488 pixels) for breast cancer diagnosis.
Methods This multicenter study retrospectively studied gray-scale ultrasound breast images enrolled from 2 Chinese hospitals. The data are divided into training, validation, internal testing and external testing set. Three-hundreds images were randomly selected for the comparison test between physicians and models. The Wilcoxon test was used to compare the diagnose error of physicians and models. The significance level was set at P=0.05 and 0.1. The specificity, sensitivity, accuracy, area under the curve (AUC) were used for evaluation. 
Results A total of 13,684 images of 3447 female patients are finally included. The 224 and 320 REZ achieve the best performance in MobileNet and EfficientNetB0 respectively (AUC: 0.893 and 0.907). The 448 REZ achieve the best performance in Xception, DenseNet121 and ResNet50 (AUC: 0.900, 0.883 and 0.871 respectively). The 320 REZ for EfficientNetB0 (AUC: 0.896, P < 0.1) is better than senior physicians. The 224 REZ for MobileNet (AUC: 0.878, P < 0.1), 448 REZ for Xception (AUC: 0.895, P < 0.1) are better than junior physicians. The 448 REZ for DenseNet121 (AUC: 0.880, P < 0.05) and ResNet50 (AUC: 0.838, P < 0.05) are only better than entry physicians. 
Conclusion Based on the gray-scale ultrasound breast images, we obtained the best DL combination which was better than the physicians.