Texture analysis in assessment and prediction of chemotherapy response in breast cancer

Abstract

Purpose: To assess the efficacy of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI)-based textural analysis in predicting response to chemotherapy in a cohort of breast cancer patients.

Materials and methods: In all, 100 patients were scanned on a 3.0T HDx scanner immediately prior to neoadjuvant chemotherapy treatment. A software application to use texture features based on co-occurrence matrices was developed. Texture analysis was performed on precontrast and 1-5 minutes postcontrast data. Patients were categorized according to their chemotherapeutic response: partial responders corresponding to a decrease in tumor diameter over 50% (40) and nonresponders corresponding to a decrease of less than 50% (4). Data were also split based on factors that influence response: triple receptor negative phenotype (TNBC) (22) vs. non-TNBC (49); node negative (45) vs. node positive (46); and biopsy grade 1 or 2 (38) vs. biopsy grade 3 (55).

Results: Parameters f2 (contrast), f4 (variance), f10 (difference in variance), f6 (sum average), f7 (sum variance), f8 (sum entropy), f15 (cluster shade), and f16 (cluster prominence) showed significant differences between responders and partial responders of chemotherapy. Differences were mainly seen at 1-3 minutes postcontrast administration. No significant differences were found precontrast administration. Node +ve, high grade, and TNBC are associated with poorer prognosis and appear to be more heterogeneous in appearance according to texture analysis.

Conclusion: This work highlights that textural differences between groups (based on response, nodal status, and triple negative groupings) are apparent and appear to be most evident 1-3 minutes postcontrast administration. The fact that significant differences for certain texture parameters and groupings are consistently observed is encouraging.

Keywords: Haralick co-occurrence matrices; MRI contrast enhancement; breast cancer chemotherapy prediction; computer science informatics; image processing; texture analysis software.