Differentiation of breast tissue types for surgical margin assessment using machine learning and polarization-sensitive optical coherence tomography

• DOI: 10.1364/boe.423026
• Publication Date: 2021-04-29T18:45:40Z
• Authors: Dan Zhu; Jianfeng Wang; Marina Marjanovic; Eric J. Chaney; Kimberly A. Cradock; Anna M. Higham; Zheng G. Liu; Zhishan Gao; Stephen A. Boppart
• Keywords: 0103 physical sciences; 01 natural sciences

We report an automated differentiation model for classifying malignant tumor, fibro-adipose, and stroma in human breast tissues based on polarization-sensitive optical coherence tomography (PS-OCT). A total of 720 PS-OCT images from 72 sites 41 patients with H&E histology-confirmed diagnoses as the gold standard were employed this study. The is trained by features extracted both one OCT-based metric (i.e., intensity) four PS-OCT-based metrics phase difference between two channels ( PD ), retardation PR local LPR degree polarization uniformity DOPU )). Further optimized forward searching validated leave-one-site-out-cross-validation (LOSOCV) method, best feature subset was acquired highest overall accuracy 93.5% model. Furthermore, to show superiority our over OCT images, intensity-only (usually obtained systems) also 82.9%, demonstrating significance information tissue differentiation. high performance suggests potential using intraoperative during surgical resection cancer.