MCS Thesis Examination: Mohammad Amin Shamshiri

Abstract

    Microscopic analysis of breast cancer images is the primary task in diagnosing cancer malignancy, which requires high expertise and precision. Recent attempts to automate this highly subjective task have employed deep learning models whose success has depended on large volumes of data, while acquiring annotated data in biomedical domains is time-consuming and may not always be feasible. A typical strategy to address this is to apply transfer learning using pre-trained models on a large natural image database (e.g., ImageNet) instead of training a model from scratch. This approach, however, has not been effective in several previous studies due to fundamental differences in patterns, size and data feature between natural and medical images. In this study, we propose the Compatible-domain Transfer Learning approach that uses a compatible data set (i.e., histopathological images) instead of natural images to classify cytological biopsy specimens of breast cancer. To our best knowledge, this is the first reported effort to employ a histopathology data source to classify cytological images of breast cancer. Despite the inconsistency of histopathological and cytological images in some respects, we demonstrate that the features learned by networks during the pre-training procedure are compatible with those obtained throughout fine-tuning with the target data set. To comprehensively investigate this assertion, we explore three different strategies for training as well as two different approaches for fine-tuning deep learning models. The proposed method is compared with five state-of-the-art studies previously conducted on the same data set, and we demonstrate that the proposed approach significantly outperforms all of them in terms of classification accuracy. Specifically, the proposed method has improved classification accuracy by 6% to 17% compared to the state-of-the-art studies which were based on traditional machine learning techniques, and also enhanced accuracy by roughly 7% compared to those who utilized deep learning methods, eventually achieving 94.55% test set accuracy and 98.73% validation accuracy. Experimental results show that our approach, despite using a very limited number of images, has achieved performance comparable to that of experienced pathologists and has the potential to be applied in clinical settings.

Examining Committee

• Dr. Charalambos Poullis (Chair) 
• Dr. Adam Krzyzak (Supervisor)
  
• Dr. Sudhir Mudur (Examiner)
• Dr. Charalambos Poullis (Examiner)