Abstract PD5-04: Tumor-infiltrating lymphocytes evaluation using machine learning image analysis based on core needle biopsy in HER2-positive breast cancer

Abstract

Abstract Purpose: To evaluate tumor-infiltrating lymphocytes (TILs) quantitatively in breast cancer and to verify the utility of machine learning image analysis in TIL assessment. Methods: Stromal TILs in 105 Human epidermal growth factor receptor 2 (HER2)-positive invasive breast cancers, diagnosed from patients treated with neoadjuvant anti-HER2 therapy, were evaluated by image analysis of hematoxylin and eosin-stained slides from core needle biopsies. TIL level was determined as the number of TILs per square millimeter of stromal tissue. Associations of TIL level with clinicopathological parameters, pathological response, and clinical outcomes were assessed. Results: Median TIL level were 1287/mm2 (range, 123 - 8101/mm2). Higher TIL level was associated with higher histological grade (P = 0.02), estrogen receptor (ER) negativity (P = 0.036), and pathological complete response (pCR) (P< 0.0001). According to analysis of the receiver operating characteristic curve, a threshold TIL level of 2420/mm2 stroma best discriminated pCR from non-pCR. In multivariate analysis, high TIL levels (>2420/mm2) were significantly associated with pCR (P< 0.0001). On Kaplan-Meier analysis, prognosis was not associated with high or low TIL levels. However, high TIL levels tended to be associated with better prognosis in the HER2+/ER+ subgroup (P = 0.131), but not in the HER2+/ER- subgroup. Conclusions: A machine learning image analysis algorithm could assess TIL level as the number of TILs per square millimeter, contrary to semiquantitative evaluation. TIL level derived from this method could offer an independent predictor of pCR. Citation Format: Norie Abe, Hirofumi Matsumoto, Reika Takamatsu, Kentaro Tamaki, Naoko Takigami, Kanou Uehara, Yoshihiko Kamada, Nobumitsu Tamaki, Tokiwa Motonari, Norihiro Nakada, Hisamitsu Zaha, Naoki Yoshimi. Tumor-infiltrating lymphocytes evaluation using machine learning image analysis based on core needle biopsy in HER2-positive breast cancer [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr PD5-04.