Combination of Tumor Kinetics and Morphometry in Breast DCE-MRI

Abstract

CONTENT ORGANIZATION 1. Introduction of our proposed perfusion index which maximizes the difference of enhancement kinetic features between a lesion and normal parenchyma 2. Automatic tumor segmentation using Otsu thresholding, 3D region growing algorithm, hole-filling and iterative erosion and dilation 3. Integrated feature extraction that combines enhancement kinetics and 3D intratumoral morphometry 4. Tumor discrimination using least square support vector machine with radial basis function

SUMMARY The major teaching points of this exhibit are: 1. Automatic tumor segmentation in breast DCE-MRI is successfully carried out from our proposed perfusion index image which maximizes the difference of enhancement kinetic features between a lesion and normal parenchyma 2. Combining kinetics and 3D morphometry unfolds the spatio-temporal complexities of intra-tumoral enhancements 3. This combined approach enables the high performance discrimination of benign and malignant tumors.