E2650. Deep Learning-Based Magnetic Resonance Radiomics Analysis for Predicting Low- and High-Grade Clear Cell Renal Cell Carcinoma
Authors
Ki Choon Sim;
Korea University Anam Hospital
Na Yeon Han;
Korea University Anam Hospital
Yongwon Cho;
Korea University Anam Hospital
Deuk Jae Sung;
Korea University Anam Hospital
Beom Jin Park;
Korea University Anam Hospital
Min Ju Kim;
Korea University Anam Hospital
Yeo Eun Han;
Korea University Anam Hospital
Objective:
Explore whether high- and low-grade clear cell renal cell carcinomas (H-ccRCC and L-ccRCC, respectively) can be distinguished using radiomic features extracted from magnetic resonance imaging (MRI).
Materials and Methods:
In this retrospective study, 154 patients with pathologically proven clear cell renal cell carcinoma (ccRCC) underwent contrast-enhanced (CE) 3T MRI and were assigned to the development (n = 122) and test (n = 32) cohorts in a temporal-split setup. A total of 834 radiomic features were extracted from whole tumor volumes using three sequences: T2-weighted imaging (T2WI), diffusion-weighted imaging, and CE T1-weighted imaging. A random forest regressor was used to extract important radiomic features that were subsequently used for model development using the random forest algorithm. Tumor size, apparent diffusion coefficient (ADC) value, and percentage of tumor-to-renal parenchymal signal intensity drop in the tumors were recorded by two radiologists for quantitative analysis. The area under the receiver operating characteristic curve (AUC) was generated to predict ccRCC grade.
Results:
In the development cohort, the T2WI-based radiomics model demonstrated the highest performance (AUC = 0.82). The T2WI-based radiomics and radiologic features hybrid model showed an AUC of 0.79 and 0.83. In the test cohort, the T2WI-based radiomics model achieved an AUC of 0.82. The hybrid model of T2WI-based radiomics and radiologic features achieved an AUC of 0.79 and 0.83.
Conclusion:
MRI-based classifier models using radiomic features and machine learning showed satisfactory diagnostic performance in distinguishing between H- and L-ccRCC, thereby, serving as a helpful noninvasive tool for predicting ccRCC grade.
