Z Ren, F Pineda*, F Howard, K Kulkarni, H Abe, R Nanda, N Jaskowiak, G Karczmar, University of Chicago, Chicago, IL

• 
  F Pineda

WE-B-201-4 (Wednesday, 7/13/2022) 8:30 AM - 9:30 AM [Eastern Time (GMT-4)]

Room 201

Purpose: Develop logistic regression models using the kinetics from tumor, normal parenchymal enhancement (kBPE) and bilateral kBPE asymmetry from ultrafast dynamic contrast-enhanced MRI (DCE-MRI) for predicting breast tumor response to neoadjuvant therapy (NAT).

Methods: Forty-five females with histological confirmed unilateral invasive breast cancer who underwent ultrafast DCE-MRI (temporal resolution = 3 – 7 seconds) prior to NAT were enrolled. Twenty-one patients achieved pathological complete response (pCR) and 24 patients were classified as residual cancer burden (RCB) I (n = 9) or RCB II (n = 15). An analytical pipeline was developed to segment ROIs of breast, tumor, blood vessels and background parenchyma on registered images acquired from the pre-NAT ultrafast DCE-MRI series (Figure 1). Pre-NAT kinetic parameters were calculated in the top 20% enhancing voxels in segmented tumors, and in the ipsi-and contralateral normal parenchyma (Table 1), and then selected for training logistic regression models that were evaluated using the area under a receiver operating characteristic curve (ROC-AUC). A leave-one-case-out validation method was applied to test each model.

Results: The ROC-AUC of Model 3 created from features related to kBPE asymmetry only was higher than the ROC-AUCs of models with features from tumor (Model 1) or contralateral kBPE (Model 2) only (Table 2). Model 4 using kinetic features from tumor, contralateral kBPE and kBPE asymmetry showed highest ROC-AUC: 0.821 ± 0.009.

Conclusion: Previous work [1] shows that BPE could be a predictor of NAT response. Here we build on this previous work by measuring kBPE and bilateral asymmetry of kBPE. This adds important new information that can only be measured accurately on ultrafast DCE-MRI. The results show that bilateral asymmetry of kBPE can predict breast tumor response to NAT prior to NAT. Patients that exhibit greater asymmetry in kBPE prior to NAT are less likely to respond to therapy.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email