“…Prior to the computer extraction of the various image phenotypes, the tumor was segmented on the MRI using the radiologist-indicated tumor center and a computational fuzzy c-means algorithm. 15 Quantitative radiomics analysis was then conducted, [16][17][18][19][20][21][22][23][24][25][26]27 yielding 38 radiomic features characterizing the size, shape, morphology, enhancement texture, kinetics, and variance kinetics of each tumor. These radiomic features can be sorted into six MRI phenotype categories: (1) size, giving the tumor dimensions, such as volume and surface area, (2) shape, characterizing the tumor geometry, such as sphericity and irregularity, (3) morphology, combining tumor shape and margin characteristics, such as spiculation and margin sharpness, (4) enhancement texture, characterizing tumor textural properties based on the gray-level co-occurrence matrix, such as energy, entropy, and contrast, (5) kinetic curve assessment, characterizing the physiological process of the uptake and washout nature of the contrast agent in a breast tumor during the dynamic imaging series, such as uptake rate, washout rate, and signal enhancement ratio, and (6) enhancement-variance kinetic features, characterizing the time course of the spatial variance of the enhancement within a breast tumor, such as variance increase rate and variance decrease rate.…”