Fig. 2.

Breast cancer cohort (BRCA) single- and multi- radiogenomic datasets organization flowcharts. TCGA provides genomic data and clinical data of a cohort with 1097 BRCA patients. TCIA provides medical images of partial TCGA-BRCA cohort (137 patients). Four of the 1097 BRCA patients have no meaningful survival information (death days or last contact days are negative numbers) thus were excluded. Nighty-one of 137 TCIA patients have annotated dynamic contrast-enhanced magnetic resonance images (DCE-MRI). A: BRCA single-radiogenomic dataset. One-thousand and nighty-three (1097-4) TCGA-BRCA patients all have gene expression data. Thus, gene expression data were used as baseline single-genomic information in this study to compare with the multi-genomic information. Those 91 patients with annotated DCE-MRI are all included in the 1093 TCGA-BRCA patients. This means, 91 patients have paired gene expression data and annotated image data. However, no survival difference is observed among these 91 patients, because they were all alive according to the last follow-up. Therefore, we cannot perform survival analysis using the paired data. The rest of 1002 (1093-91) patients only have gene expression data (no DCE-MRI data), but there exists survival difference among them. B: BRCA multi-radiogenomic dataset. Only 762 of the 1093 TCGA-BRCA patients have matched gene expression, copy number alteration (CNA), and DNA methylation data. Sixty-one of those 91 TCIA-BRCA patients with annotated DCE-MRI are included in the 762 TCGA-BRCA patients with multi-genomic data. This means, 61 patients have paired multi-genomic data and annotated image data. However, no survival difference is observed among these 61 patients. The rest of 701 (762-61) patients only have multi-genomic data (no medical image data), but there exists survival difference among them.