MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Quantitative MR Imaging Features Associated with Pathologic Complete Response to Neoadjuvant Chemotherapy in Breast Cancer Patients

J Jimenez*, A Abeer, N Elshafeey, J Yung, J Hazle, G Rauch, UT MD Anderson Cancer Center, Houston, TX

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Identify quantitative MR imaging features associated with pathologic complete response (pCR) to neoadjuvant chemotherapy (NACT) to potentially improve prediction of treatment response in breast cancer patients.

Methods: This IRB approved retrospective study included 72 biopsy-proven HER 2 positive and triple negative (TN) breast cancer patients. The study included patients that received NACT, had pretreatment dynamic contrast enhanced (DCE) MRI and pretreatment biopsy-based pathological assessment of the tumor infiltrating lymphocytes (TIL). Patients were classified into pCR, and non-pCR groups based on post-operative pathology, where pCR was defined as the absence of residual invasive component in tumor bed. DCE-MR images were obtained from a variety of GE and Siemens platforms. Lesion segmentation and image feature subtraction was processed using FDA approved software QuantX.

Feature analysis was done using MATLAB. We first evaluated the Pearson linear correlation between individual imaging features and the pCR classifier. Next, we used minimum redundancy maximum relevance (MRMR) and binary decision trees to rank the importance of the remaining features. Finally, we linearly combined the top five imaging features into a single imaging signature number.

Results: We found 26 statistically significant quantitative imaging features out of the 145 subtracted features. The most relevant imaging features were: Contrast, Correlation, Maximal Correlation Coefficient, Homogeneity, and Difference Variance. The imaging feature model provides better accuracy than the pathology model: 71.4% (MR features) and 65.5% (TIL model) for pCR and 81.1% (MR features) and 69.8% (TIL model) for non-pCR. However, when both models are combined, there is an even higher prognostic predictively: 79.0% for pCR and 86.4% for non-pCR.

Conclusion: Our study showed that quantitative imaging features can produce similar classification accuracy to pathologic information from biopsy. Combining the imaging features with pathology information could improve prognostic prediction confidence for NAC treatment response in breast cancer patients.

Keywords

Diagnostic Radiology, Image Analysis, Breast

Taxonomy

IM- MRI : Breast MRI

Contact Email