A Computer-aided radiomics approach based on classifier ensemble to differentiate malignancy of hepatocellular carcinoma with Contrast-enhanced MR
2019 International Conference on Medical Imaging Physics and Engineering (ICMIPE)(2019)
Abstract
Preoperative knowledge of the malignant grade of hepatocellular carcinoma (HCC) is significant to establish a therapeutic strategy and predict the patient prognosis in clinical practice. The radiomics approach has been recently proposed to estimate the pathological grade of HCC based on non-invasive medical images. However, previous radiomics studies generally make the prediction based on a single modality of medical images, without taking full advantage of multimodalities for lesion characterization. In this study, we proposed a computer-aided radiomics approach based on classifier ensemble to differentiate malignancy for HCC with Contrast-enhanced MR. First, radiomics features derived from histogram, run-length matrix and co-occurrence matrix are separately extracted from regions of interests (ROIs) of lesions delineated from the pre-contrast, arterial, and portal vein phase of Contrast-enhanced MR. For each phase, feature selection using the the least absolute shrinkage and selection operator (LASSO) logistic regression model is adopted to obtain a subset of most discriminative features that provide optimal performance, and support vector machine (SVM) is selected as the classifier to differentiate those lesions into low-grade (grade I and II) and high-grade HCCs (grade III and IV). The classification results of the three phases are then combined by Multiple Kernel Learning(MKL) in order to make full use of contexture information from the three phases. Experimental results of 117 clinical HCCs with pathological confirmed information demonstrate the superior performance of the proposed method compared with conventional radiomics approach.
MoreTranslated text
Key words
computer-aided radiomics approach,classifier ensemble,hepatocellular carcinoma,contrast-enhanced MR,malignant grade,pathological grade,noninvasive medical images,lesion characterization,radiomics features,run-length matrix,portal vein phase,feature selection,high-grade HCCs,conventional radiomics approach,malignancy differentiation,therapeutic strategy,patient prognosis,histogram,cooccurrence matrix,precontrast vein phase,arterial phase,least absolute shrinkage and selection operator logistic regression model,discriminative features,support vector machine,SVM,low-grade HCC,multiple kernel learning,contexture information,clinical HCC,pathological confirmed information
AI Read Science
Must-Reading Tree
Example
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined