Computer Engineering and Applications ›› 2019, Vol. 55 ›› Issue (3): 252-257.DOI: 10.3778/j.issn.1002-8331.1710-0296

Previous Articles     Next Articles

Application of Deep Learning in Identification of Throat New Organisms

TIAN Yongliang1, ZHANG Jing1, LIU Kai1, TANG Wei2, TIAN Weidong2   

  1. 1.College of Electrical Engineering and Information Technology, Sichuan University, Chengdu 610065, China
    2.West China School/Hospital of Stomatology, Sichuan University, Chengdu 610041, China
  • Online:2019-02-01 Published:2019-01-24


田永良1,张  劲1,刘  凯1,汤  炜2,田卫东2   

  1. 1.四川大学 电气信息学院,成都 610065
    2.四川大学 华西口腔医学院,成都 610041

Abstract: Throat new organism is the disease that new organisms grow in the area of vocal cord, which is one of common diseases that affect normal function of throat. At present, the throat new tissues are mainly diagnosed by doctors using laryngoscope photos. However, misdiagnosed cases happened because of the differences of individual clinical experience of doctors in the diagnosis of new tissue diseases often have different diagnostic results. And new tissue diseases if not timely diagnosed and treated is possible to evolve into laryngeal cancer. Based on this, this paper proposes a learning algorithm of throat new tissue disease based on deep learning. Through the multi-layer convolution kernels extracted features from a large number of labeled training sets, and only the effective features are retained in the process of reverse propagation. The studied model has an impressive fitting performance on training set and also a good generalization performance on test set. It has a certain value in the practical application.

Key words: deep learning, neural network, features, laryngoscope, throat new organisms

摘要: 咽喉新生物是声带及其附近区域生长有新生组织,影响正常咽喉功能的咽喉常见疾病之一。目前,咽喉新生物主要是医生通过拍摄喉镜照片来进行诊断,由于医生个体临床经验的差异,在新生物疾病的诊断上往往存在不同诊断结果,常常有误诊的情况发生。而新生物疾病若不能及时确诊和治疗就有可能演变成喉癌的可能。基于此,提出了基于深度学习的咽喉新生物疾病识别算法,通过多层较小的卷积核在大量标注的训练集上逐层提取特征,在反向传播的过程中只保留有效特征,最终得到很好的拟合效果,在测试集上也达到了很好的泛化表现。具有一定的实际应用价值。

关键词: 深度学习, 神经网络, 特征, 喉镜, 咽喉新生物