Abstract #0058

BI-RADS Reading of Non-Mass Lesions on DCE-MRI and Differential Diagnosis Performed by Radiomics and Deep Learning

Jiejie Zhou¹, Yan-Lin Liu², Yang Zhang², Jeon-Hor Chen^2,3, Freddie J. Combs², Ritesh Parajuli⁴, Rita S. Mehta⁴, Huiru Liu¹, Zhongwei Chen¹, Youfan Zhao¹, Meihao Wang¹, and Min-Ying Su²

¹Department of Radiology, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China, ²Department of Radiological Sciences, University of California, Irvine, CA, United States, ³Department of Radiology, E-Da Hospital and I-Shou University, Kaohsiung, Taiwan, ⁴Department of Medicine, University of California, Irvine, CA, United States

A total of 150 lesions, 104 malignant and 46 benign, presenting as non-mass-like enhancements were analyzed. Three radiologists performed BI-RADS reading for the morphological distribution and internal enhancement pattern. For each case, the 3D tumor mask was generated using Fuzzy-C-Means segmentation. Three DCE parametric maps were generated, and PyRadiomics was applied to extract features. The radiomics model was built using 5 different machine learning algorithms. ResNet50 was implemented using three parametric maps as input. SVM yielded the highest accuracy of 80.4% in training, 77.5% in testing datasets. ResNet50 had better diagnostic performance, 91.5% in training, and 83.3% in testing datasets.

This abstract and the presentation materials are available to members only; a login is required.

Join Here