Radiomic Signatures Derived from Diffusion-Weighted Imaging for the Assessment of Breast Cancer Receptor Status and Molecular Subtypes

• DOI: 10.1007/s11307-019-01383-w
• Publication Date: 2019-06-17T20:37:42Z
• Authors: Doris Leithner; Blanca Bernard-Da...; Danny F. Martinez; Joao V. Horvat; Maxine S. Jochelson; Maria Adele Marino; Daly Avendano; R. Elena Ochoa-Al...; Elizabeth J. Sutton; Elizabeth A. Morris; Sunitha B. Thakur; Katja Pinker
• Keywords: Breast MRI; Subtyping

To compare annotation segmentation approaches and to assess the value of radiomics analysis applied diffusion-weighted imaging (DWI) for evaluation breast cancer receptor status molecular subtyping. In this IRB-approved HIPAA-compliant retrospective study, 91 patients with treatment-naïve malignancies proven by image-guided biopsy, (luminal A, n = 49; luminal B, 8; human epidermal growth factor 2 [HER2]-enriched, 11; triple negative [TN], 23) underwent multiparametric magnetic resonance (MRI) at 3 T dynamic contrast-enhanced MRI, T2-weighted DW imaging. Lesions were manually segmented on high b-value images ROIS propagated apparent diffusion coefficient (ADC) maps. addition in a subgroup (n 79) where lesions discernable ADC maps alone, these also directly there. derive signatures, following features extracted analyzed: first-order histogram (HIS), co-occurrence matrix (COM), run-length (RLM), absolute gradient, autoregressive model (ARM), discrete Haar wavelet transform (WAV), lesion geometry. Fisher, probability error average correlation, mutual information coefficients used feature selection. Linear discriminant followed k-nearest neighbor classification leave-one-out cross-validation was pairwise differentiation Histopathologic results considered gold standard. For that DWI ROIs accuracies > 90% obtained: B vs. HER2-enriched, 94.7 % (based COM features); others, 92.3 (COM, HIS); HER2-enriched 90.1 (RLM, COM). maps, better achieved yielding accuracies: 100 WAV); A 91.5 91.1 (WAV, ARM, Radiomic signatures from mapping allows subtyping diagnostic accuracy. Better obtained when tumor segmentations could be performed