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

Min-Ying Su,Ritesh Parajuli,Huiru Liu,Risheng Yu,Jeon-Hor Chen,Rita S. Mehta,Yan-Lin Liu,Freddie J. Combs,Yang Zhang,Meihao Wang,Youfan Zhao,Zhongwei Chen,Jiejie Zhou,Zhifang Pan

doi:10.3389/fonc.2021.728224

Abstract

BackgroundA wide variety of benign and malignant processes can manifest as non-mass enhancement (NME) in breast MRI. Compared to mass lesions, there are no distinct features that can be used for differential diagnosis. The purpose is to use the BI-RADS descriptors and models developed using radiomics and deep learning to distinguish benign from malignant NME lesions.Materials and MethodsA total of 150 patients with 104 malignant and 46 benign NME were analyzed. Three radiologists performed reading for morphological distribution and internal enhancement using the 5th BI-RADS lexicon. For each case, the 3D tumor mask was generated using Fuzzy-C-Means segmentation. Three DCE parametric maps related to wash-in, maximum, and wash-out were generated, and PyRadiomics was applied to extract features. The radiomics model was built using five machine learning algorithms. ResNet50 was implemented using three parametric maps as input. Approximately 70% of earlier cases were used for training, and 30% of later cases were held out for testing.ResultsThe diagnostic BI-RADS in the original MRI report showed that 104/104 malignant and 36/46 benign lesions had a BI-RADS score of 4A–5. For category reading, the kappa coefficient was 0.83 for morphological distribution (excellent) and 0.52 for internal enhancement (moderate). Segmental and Regional distribution were the most prominent for the malignant group, and focal distribution for the benign group. Eight radiomics features were selected by support vector machine (SVM). Among the five machine learning algorithms, SVM yielded the highest accuracy of 80.4% in training and 77.5% in testing datasets. ResNet50 had a better diagnostic performance, 91.5% in training and 83.3% in testing datasets.ConclusionDiagnosis of NME was challenging, and the BI-RADS scores and descriptors showed a substantial overlap. Radiomics and deep learning may provide a useful CAD tool to aid in diagnosis.

Highlights

Breast MRI is an important modality for the detection and characterization of lesions
The detection and diagnosis of non-mass enhancement (NME) have been known as a more challenging problem compared to mass lesions, which may be addressed by advanced machine learning methods [6]
The common histopathology that may manifest as NME includes ductal carcinoma in situ (DCIS), invasive ductal cancer (IDC), invasive lobular cancer (ILC), benign adenosis, fibrocystic changes, and inflammation [12, 26]

Summary

Introduction

Breast MRI is an important modality for the detection and characterization of lesions. It has become a clinical examination routinely used with mammography and ultrasound for diagnosis of breast cancer [1, 2]. Dynamic contrast-enhancement MRI (DCE-MRI) is a well-established imaging method to evaluate the vascular properties, which can be used for distinguishing benign from malignant lesions [3, 4]. While mass lesions can be detected by all imaging modalities and diagnosed with a high accuracy, the diagnosis of NME lesions are more challenging [6]. To further designate the detected NME as likely to be malignant or benign is more difficult, and machine learning-based computer-aided diagnosis (CAD) may provide a feasible tool [6]. The purpose is to use the BI-RADS descriptors and models developed using radiomics and deep learning to distinguish benign from malignant NME lesions

Methods

Results

Conclusion