基于深度神经网络的肿瘤细胞分类器的研究

Research of Tumor Cell Classifier based on Deep Neural Network

近年来随着深度神经网络(Deep Neural Networks,DNN)的模型不断完善,DNN在肿瘤的数字图像识别方面精度越来越高,DNN受到了国内外医学界的广泛研究和重视,DNN为识别良性肿瘤与恶性肿瘤的临床诊断提供了客观、准确、快速、经济的解决方案。本文主要综述了DNN技术在识别良性肿瘤与恶性肿瘤方向上使用的几种常见分类器:卷积神经网络、生成对抗网络、深度残差网络和深度信念网络。分析了这几种分类器的原理及其应用的效果,分析了基于不同神经网络分类器的精准度和性能,提出了DNN的各种分类器在识别良性肿瘤与恶性肿瘤领域中面临的问题及未来发展的趋势。

In recent years,with the improvement of deep neural networks(DNN)models,the accuracy of digital image recognition of tumors has been improved continuously.DNN provides an objective,accurate,fast and economic solution for the clinical diagnosis of benign and malignant tumors.This technology has been widely studied and paid attention to by the medical community at home and abroad.This paper reviews several DNN classifiers(Convolution Neural Network,Generative Adversarial Network,Deep Residual Network,and Deep Belief Network),analyzes the accuracy and performance of these neural networks based on different classifiers,and forecast the future development trend of DNN classifiers.