MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Radiomics for Breast MRI in Pre-Treatment Prediction of Nodal Response to Neoadjuvant Chemotherapy in Node-Positive Breast Cancer Patients; a Pilot Study

K Drukker*, C Doyle , A Edwards , J Papaioannou , M Giger , university Chicago, Chicago, IL

Presentations

(Sunday, 7/29/2018) 4:00 PM - 4:30 PM

Room: Exhibit Hall | Forum 8

Purpose: To evaluate radiomics of pre-treatment dynamic contrast-enhanced breast MR images in the prediction of lymph node (LN) response to neoadjuvant chemotherapy (NAC) in patients with invasive LN-positive breast cancer.

Methods: Forty-seven patients with LNs positive for metastasis were included in this retrospective study; 32 post-NAC LN-positive (non-responders) and 15 post-NAC LN-negative (responders, all lymph nodes dissected post-NAC were negative for metastasis). We analyzed only pre-NAC MRIs. Findings were localized on the images by an expert breast radiologist. Subsequent automated analysis included segmentation and extraction of 38 radiomic features describing, for index cancers and metastatic axillary sentinel LNs, (i) size, (ii) shape, (iii) margin, (iv) kinetic curve, (v) contrast-enhancement texture, and (vi) variance kinetics. For cancers and LNs separately, each radiomic feature was evaluated to determine whether a statistically significant difference between the post-NAC LN-positive and post-NAC LN-negative subgroups was demonstrated using the Mann-Whitney U-test (without and with correction of p-values using Holm-Bonferroni for 38 comparisons). The area under the ROC curve was calculated for the task of distinguishing between the post-NAC LN-positive and LN-negative groups.

Results: All radiomics features describing index cancers failed to show a statistically significant difference between post-NAC LN-positive and LN-negative groups (p>0.05). Eight radiomic features describing pre-treatment metastatic LNs demonstrated statistically significant differences between the two groups: 4 kinetic curve- and 4 enhancement variance kinetics features. After correction for multiple comparisons, a single lymph node feature remained statistically significant: time to maximum contrast-enhancement variance (p=0.0012) with an area under the ROC curve of 0.78 (standard error 0.07) in the prediction of nodal response.

Conclusion: Radiomics for breast MRI shows promise in the pre-treatment prediction of nodal response to neoadjuvant chemotherapy in patients with lymph-node positive invasive breast cancer. This could positively impact patient management since axillary dissection and radiation are associated with significant morbidity.

Funding Support, Disclosures, and Conflict of Interest: Funding: U01CA195564. Karen Drukker receives royalties from Hologic. Maryellen Giger is a stockholder in Hologic Inc. and Quantitative Insights Inc., is co-founder of Quantitative Insights Inc., and receives royalties from Hologic Inc., General Electric Company, MEDIAN Technologies, Riverain Technologies LLC, Mitsubishi Corporation and Toshiba Corporation

Keywords

Breast, CAD

Taxonomy

IM- MRI : Quantitative imaging/analysis

Contact Email