Elizabeth S. Burnside MD, MPH, Presenter: Stockholder, Cellectar Biosciences, Inc Stockholder, NeuWave Medical Inc

Maryellen L. Giger PhD, Abstract Co-Author: Stockholder, Hologic, Inc Shareholder, Quantitative Insights, Inc Royalties, Hologic, Inc Royalties, General Electric Company Royalties, MEDIAN Technologies Royalties, Riverain Technologies, LLC Royalties, Mitsubishi Corporation Royalties, Toshiba Corporation Researcher, Koninklijke Philips NV Researcher, U-Systems, Inc

Elizabeth A. Morris MD, Abstract Co-Author: Nothing to Disclose

Gary J. Whitman MD, Abstract Co-Author: Nothing to Disclose

Ermelinda Bonaccio MD, Abstract Co-Author: Nothing to Disclose

Margarita Louise Zuley MD, Abstract Co-Author: Research Grant, Hologic, Inc

Hui Li PhD, Abstract Co-Author: Nothing to Disclose

Erich Huang PhD, Abstract Co-Author: Nothing to Disclose

Jose Miguel Net MD, Abstract Co-Author: Nothing to Disclose

Marie Adele Ganott MD, Abstract Co-Author: Nothing to Disclose

Kathleen Rae Brandt MD, Abstract Co-Author: Nothing to Disclose

Elizabeth J. Sutton MD, Abstract Co-Author: Nothing to Disclose

Justin Kirby, Abstract Co-Author: Stockholder, Myriad Genetics, Inc

John B. Freymann BS, Abstract Co-Author: Nothing to Disclose

Basak Erguvan Dogan MD, Abstract Co-Author: Nothing to Disclose

C. Carl Jaffe MD, Abstract Co-Author: Nothing to Disclose

Brenda Fevrier-Sullivan BA, Abstract Co-Author: Nothing to Disclose

Yuan Ji, Abstract Co-Author: Nothing to Disclose

One of the most important roles of imaging in women with breast cancer is to accurately predict stage in order to direct patients to appropriate treatment. Our goal in this study was to demonstrate that computer extracted features of biopsy-proven breast cancer (computer-extracted tumor phenotype-CETP) on MRI can accurately predict breast cancer stage.

We used a retrospectively collected dataset of de-identified breast MRIs from multiple institutions organized by the National Cancer Institute (NCI) in a centralized repository called The Cancer Imaging Archive (TCIA) which includes outcomes collected from cancer center tumor registries. For each case, we characterized tumors on MRI by (a) qualitative semantic features from multiple radiologists’ interpretations and (b) automated computerized image analyses (CTEP) including volumetrics, texture (homogeneity), and kinetics. We built a linear discriminant analysis model (LDA) to predict tumor stage and lymph nodes involvement on pathology. We evaluated each LDA model in turn by calculating a risk score for each patient (using cross validation); used this risk score to construct ROC curves; and compared the AUC of each model to baseline chance (AUC=0.5) using the DeLong method.

We analyzed a total of 98 biopsy proven breast cancer cases. Pathologic outcomes revealed: negative nodes (n= 49), >1 positive node (n = 48; 1 missing), stage I (n= 23), stage II (n= 62) and stage III (n= 13). Use of CTEP to distinguish between tumors at stage 1 (N=23) and stage III (N=13) demonstrated an AUC = 0.7, significantly better than chance (p = 0.017). We also found that CTEP could distinguish between tumor without (N=49) and with (N=48) positive lymph nodes AUC = 0.59.

The results from this study indicate that quantitative MRI analysis shows promise as a means for predicting breast cancer stage and lymph node status.

In an era of personalized treatment based on genetics, demonstrating that image based (MRI) phenotyping can contribute to prediction of cancer stage is important.

Burnside, E, Giger, M, Morris, E, Whitman, G, Bonaccio, E, Zuley, M, Li, H, Huang, E, Net, J, Ganott, M, Brandt, K, Sutton, E, Kirby, J, Freymann, J, Dogan, B, Jaffe, C, Fevrier-Sullivan, B, Ji, Y, Using Computer-extracted Features from Tumors on Breast MRI to Predict Stage.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045585.html