A Rincon¹*, L Young², Y Yang³, F Yang⁴, (1) Department Of Biomedical Engineering, University Of Miami, Coral Gables, FL, (2) Department Of Radiation Oncology, University of Washington, Seattle, WA, (3) Department of Radiation Oncology, The First Affiliated Hospital of University of Science and Technology of China, Hefei, China, (4) (1) Department Of Radiation Oncology, University Of Miami, Miami, FL

• 
  A Rincon

SU-H400-IePD-F5-5 (Sunday, 7/10/2022) 4:00 PM - 4:30 PM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 5

Purpose: Breast MRI is commonly used in the pre-operative evaluation of women with breast cancer to determine the extent of the tumor and assist in surgical planning. Imaging phenotypes extracted may provide early risk of recurrence data. The purpose of this study was to assess the discrimination power of pre-operative MRI imaging phenotypes between patients with invasive breast cancer who developed recurrence after completion of treatment and those who were recurrence-free.

Methods: The current study employed patient data (n=259) accrued within the Duke Breast Cancer MRI collection for which imaging and clinical data are publicly available from The Cancer Imaging Archive (TCIA). The event of interest was recurrence. A total of 529 imaging phenotypes characterizing ranging from size, morphology, heterogeneity, to enhancement variation were extracted from the delineated tumor and fibroglandular tissue volumes. The phenotype data obtained for patients with and without recurrence were then balanced using synthetic minority over-sampling technique (SMOTE). Top imaging phenotypes that minimized redundancy while maximizing relevance to the endpoint were selected and used to build a support vector machine (SVM)-based predictive classifier. Performance of the developed classifier was assessed by receiver operating characteristic (ROC) curve analysis.

Results: Of the 259 patients with invasive breast cancer being studied, 14.6% experienced recurrence. The selected top phenotypes fall in the categories of tumor enhancement variation, FGT enhancement variation, tumor enhancement texture, and FGT enhancement texture. SVM classifier built using the selected imaging phenotypes demonstrated a relatively strong discriminative power in differentiating patients with and without recurrence, with a ROC area under curve (AUC) of 0.87.

Conclusion: Pre-operative MRI imaging phenotype demonstrated discriminative ability in distinguishing patients with invasive breast cancer who went on to develop recurrence after completion of definitive treatment and may have the potential to risk stratify this patient population to guide the appropriate therapy.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email