Application of Lightweight Convolution Neural Network in Cancer Diagnosis
Cai Zhao, Yiyang Feng, Ruijing Liu, Wen Zheng · 2020
Acute lymphoblastic leukemia is a major problem in people's lives. At present, the most widespread detection method is mainly manual, which is time-consuming and laborious. With the development of deep learning, specifically targeting the complexity of traditional convolutional neural network models, substantial resource costs, and non-conducive deployment. This paper proposes a simple and effective classification method. The use of lightweight convolutional neural network classified microscopic images of acute leukemia can more effectively assist doctors in the diagnosis of diseases. The method is evaluated in C-NMC 2019 data set and achieves an accuracy rate is 86.29%, F1 coefficient is 89.98%, and a recall rate of an astonishing 95.73% on the test set. Meanwhile, the model parameters are one ninth of those in the traditional convolution neural network. On the premise of ensuring high accuracy, this paper discuss that reduced complexity of the model, the facilitation of deployment and application of the model in the mobile terminal and the potential to help doctors to carry out auxiliary diagnosis of diseases with the help of portable instruments or modern equipment.