Search

The deep learning models such as AlexNet, VGG, and ResNet achieved a good performance in classifying the breast cancer histopathological images in BreakHis dataset. However, these models are not practically appropriate due to their computational complexity and too many parameters; as a result, they are rarely utilized on devices with limited computational resources. This paper develops a lightweight learning model based on knowledge distillation to classify the histopathological images of breast cancer in BreakHis. This method employs two teacher models based on VGG and ResNext to train two student models, which are similar to the teacher models in development but have fewer deep layers. In the proposed method, the adaptive joint learning approach is adopted to transfer the knowledge in the final-layer output of a teacher model along with the feature maps of its middle layers as the dark knowledge to a student model. According to the experimental results, the student model designed by ResNeXt architecture obtained the recognition rate 97.09% for all histopathological images. In addition, this model has ∼69.40 million fewer parameters, ∼0.93 G less GPU memory use, and 268.17 times greater compression rate than its teacher model. While in the student model the recognition rate merely dropped down to 1.75%. The comparisons indicated that the student model had a rather acceptable outputs compared with state-of-the-art methods in classifying the images of breast cancer in BreakHis., (Copyright © 2022. Published by Elsevier Ltd.)

Magnification-independent (MI) classification is considered a promising method for detecting the histopathological images of breast cancer. However, it has too many parameters for real implementation due to dependence on input images in different magnification factors. In addition, magnification-dependent (MD) classification usually performs poorly on unseen samples, although it has lower input image sizes and fewer parameters. This paper proposes a novel method based on knowledge distillation (KD) to overcome the limitation of dissociation between MI classification and MD classification of breast cancer in histopathological images. The proposed KD method includes a pre-trained MI teacher model that is responsible for training an unprepared MD student model developed through only one magnification factor. In the proposed method, the decomposed feature maps of a teacher's intermediate layers are transferred as dark knowledge to a student. According to the experimental results, the student model developed through 40X images yielded accuracy rates of 99.41%, 99.26%, 99.14%, and 99.09% in response to unseen samples of 40X, 100X, 200X, and 400X images, respectively. Moreover, comparison results indicated the competitive performance of the proposed student model as opposed to the state-of-the-art method based on deep learning on BreakHis., (Copyright © 2022 Elsevier Ltd. All rights reserved.)