Development of an intelligent full-field polarization sensitive optical coherence tomography for breast cancer classification.

Biopsy is the gold standard for cancer detection, however the surge in breast cancer cases has made manual haematoxylin and eosin stained histopathological image examination difficult. Automatic cancer diagnosis is vital for a healthy life. It allows fast diagnosis without specific skills. This research proposes an intelligent full-field polarization-sensitive optical coherence tomography (FF-PS-OCT) system for ex-vivo breast classification using ensemble model corroborated by technique for order preference by similarity to ideal solution (TOPSIS). 220 samples image were scanned using the FF-PS-OCT to extract the phase information. The multilevel ensemble classifier has 94.8% precision, 92.5% recall, 93.7% F-score and 82.3% Mathews correlation coefficient on the testing dataset. The developed ensemble model corroborated by TOPSIS, outperforms the single model in terms of performance metrics. The initial results indicate that the rapid, non-contact and label-free FF-PS-OCT imaging modality using birefringent information is beneficial for making interventional decisions by clinicians.