Progesterone Receptor Status Analysis in Breast Cancer Patients using DCE- MR Images and Gabor Derived Anisotropy Index

Hormone receptors play a key role in female breast cancers as predictive biomarkers. Breast cancer subtype with Progesterone receptor (PgR) expression is one of the important hormone receptors in predicting prognosis and evaluating the Neoadjuvant Chemotherapy (NAC) treatment response. PgR (-) breast cancers are associated with a higher response to NAC compared to PgR (+) breast cancer patients. Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) is the widely used imaging modality in assessing the NAC response in patients. However, evaluating the treatment response of PgR breast cancers is complicated and challenging since breast cancer with positive receptor statuses will respond differently to NAC. Therefore, in this work, an attempt has been made to differentiate the PgR (+) and PgR (-) breast cancer patients due to NAC using Gabor derived Anisotropy Index (AI). A total of 50 PgR (+) and 63 PgR (-) DCE-MR images at 4 time points of NAC treatment are considered from the openly available I-SPY1 of the TCIA database. AI is calculated within the PgR status groups from Gabor energies that are acquired after designing the Gabor filter bank with 5 scales and 7 orientations. Results demonstrate that the AI values can significantly differentiate PgR (+) and PgR (-) breast cancer patients $(\mathrm{p}\leq 0.05)$ due to NAC. The mean AI values are observed to be high in PgR (+) patients $(4.14\mathrm{E}+10\pm$ 1.17E+ 11) than PgR (-) patients $(1.95\mathrm{E}+10\pm 8.06\mathrm{E}+10)$. AI could statistically differentiate visit 1 & visit 4 of NAC treatment in both PgR status patients with a p-value of 0.0246 and 0.0387 respectively. Further, the percentage difference in the mean value of AI is observed to be high in PgR (-) between visit 1 V s 4, visit 2 V s 4, visit 1 V s 3, and visit 2 Vs 3 compared to PgR (+) subjects. Hence, AI could be used as a single index value in assessing the treatment response in both PgR (+) and PgR (-) subjects.