Room: Exhibit Hall | Forum 2
Purpose: to evaluate breast MRI radiomics in predicting, prior to any treatment, the response to neoadjuvant chemotherapy (NAC) in patients with invasive lymph node-positive breast cancer in terms of post-NAC lymph node (LN) status.
Methods: Our study included 154 women; 41 women were post-NAC LN-negative (8 hormone receptor-positive/18 HER2-enriched/15 triple negative breast cancers) and 113 (69/21/23) remained LN-positive post-NAC. Only pre-NAC MRIs underwent computer analysis, initialized by an expert breast radiologist indicating index cancers and metastatic axillary sentinel lymph nodes on DCE-MRI images. We used 18 radiomics features, both for the primary cancers and for the metastatic sentinel lymph nodes, that showed promise in an earlier study (3 size, 3 shape, 2 morphology, 1 texture, 1 kinetics, 1 variance kinetics, and 7 statistics features). Analysis was performed by hormone receptor subtype of the primary breast cancer. Features promising for the prediction of LN response to NAC were identified using the Mann-Whitney U-test. ROC analysis was used to assess performance in the pre-NAC prediction of nodal response to NAC.
Results: For the hormone receptor-positive and the HER2-enriched breast cancers, 5 and 6 lymph node features, respectively, demonstrated a statistically significant difference (p<0.05) between lymph nodes going to respond to NAC and those unresponsive to NAC. Areas under the ROC curve were up to 0.81 [0.67; 0.94] and 0.78 [0.64; 0.93] for hormone receptor positive and HER2-enriched subgroups, respectively. For triple negative breast cancers no useful features could be identified, however. We were unable to identify any mass features useful for predicting post-NAC lymph node status for any of the hormone receptor subtypes.
Conclusion: In patients with hormone receptor-positive and HER2-enriched node positive breast cancer, radiomics of lymph nodes imaged pre-NAC could be helpful in treatment planning of NAC and potential avoidance of axillary dissection which is associated with substantial morbidity.
Funding Support, Disclosures, and Conflict of Interest: This project was funded in part by NIH grant U01CA195564 in the QIN (Quantitative Imaging Network). Karen Drukker receives royalties from Hologic. Maryellen Giger is a stockholder in Hologic Inc., is co-founder and equity holder in Quantitative Insights Inc., and receives royalties from Hologic Inc., General Electric Company, MEDIAN Technologies, Riverain Technologies LLC, Mitsubishi Corporation and Toshiba Corporation.
IM/TH- Image Analysis (Single modality or Multi-modality): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)