Abstract
The purpose of this study is to evaluate the accuracy for classification of hepatic tumors by characterization of T1-weighted magnetic resonance (MR) images using two radiomics approaches with machine learning models: texture analysis and topological data analysis using persistent homology. This study assessed non-contrast-enhanced fat-suppressed three-dimensional (3D) T1-weighted images of 150 hepatic tumors. The lesions included 50 hepatocellular carcinomas (HCCs), 50 metastatic tumors (MTs), and 50 hepatic hemangiomas (HHs) found respectively in 37, 23, and 33 patients. For classification, texture features were calculated, and also persistence images of three types (degree 0, degree 1 and degree 2) were obtained for each lesion from the 3D MR imaging data. We used three classification models. In the classification of HCC and MT (resp. HCC and HH, HH and MT), we obtained accuracy of 92% (resp. 90%, 73%) by texture analysis, and the highest accuracy of 85% (resp. 84%, 74%) when degree 1 (resp. degree 1, degree 2) persistence images were used. Our methods using texture analysis or topological data analysis allow for classification of the three hepatic tumors with considerable accuracy, and thus might be useful when applied for computer-aided diagnosis with MR images.
Highlights
Magnetic resonance (MR) imaging is an important tool for detection and differential diagnosis of hepatic tumors
hepatocellular carcinomas (HCCs) and hepatic hemangiomas (HHs), metastatic tumors (MTs) and HH), we obtained accuracy of 92% with the sensitivity, specificity, and area under the curve (AUC) respectively being 100%, 84%, and 0.95
HCC and HH), we obtained the best accuracy of 85% when feature vectors obtained from degree 1 persistence images and XGBoost were used, with the sensitivity, specificity, and AUC, respectively being 86%, 84% and 0.85
Summary
Magnetic resonance (MR) imaging is an important tool for detection and differential diagnosis of hepatic tumors. A differential diagnosis cannot usually be made from non-contrast-enhanced T1-weighted MR images alone because most hepatic tumors appear as low signal intensity areas in the liver without pathognomonic findings[1,2,3,4,5,6,7,8]. Texture features, which describe statistical relations between voxels with distinctive contrast values, have often been used as primary quantitative features[13,14,15,16,17,18]. This article describes our first experience with hepatic tumor classification by characterization of non-contrast-enhanced 3D T1-weighted MR images using texture features and persistent homology and using analysis of the obtained data with machine learning models
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call