Malaria Cell Detection Using Evolutionary Convolutional Deep Networks
Bin Qin, Yufan Wu, Zhili Wang, Haocheng Zheng · 2019
With the rapid development of deep learning and computer-vision, accurate identification of medical imaging has become one of the most important factors in medical diagnosis and decision-making. To this end, we propose a data-driven approach named Evolutionary Convolutional Deep Network (ECDN) to detect malaria parasites, which can use evolutionary algorithms to automatically generate deep neural networks, and optimized its network topology structure during the evolution process. Extensive experiments based on the large-scale thin-blood smear images data validate the effectiveness of ECDN for detecting malaria. To be specific, it has the advantage of being able to automatically generate an optimal network structure without the need for any prior knowledge of constructing a neural network, as compared to a traditional artificial convolution network. The experimental results show that the model robustness of ECDN is better. When the training set and test set are divided according to the ratio of 6 and 4, the best result is achieved, and the accuracy rate reaches 99.98%, which provides an important basis for this research.