基于深度学习的全景片自动牙位标识
Automatic tooth position identification of dental panoramic radiograph based on deep learning
耿飙 1齐莎莎 2魏炜3
作者信息
- 1. 中国矿业大学计算机科学与技术学院,江苏徐州 221116;苏州卫生职业技术学院健康管理学院,江苏 苏州 215009
- 2. 山东交通技师学院人事处,山东 临沂 276000
- 3. 苏州卫生职业技术学院健康管理学院,江苏 苏州 215009
- 折叠
摘要
根据国际牙科联盟系统的全景片影像实际特征,提出一种基于参数优化的用于自动牙齿检测和标号分类的方法.运用先进的深度学习方法构建创新以及实用的三阶段牙科全景片牙齿标识方法.使用全景片图像将其分为几个阶段,以SqueezeNet的基于掩膜区域卷积神经网络作为基线模型进行特征提取过程,使用燕群优化算法进行参数优化,应用基于SoftMax分类器的牙齿预测和加权极限学习机的阶段分类模型确定牙齿编号类别标签,在图像数据集上进行评估,所提方法具有性能竞争力.
Abstract
According to the actual characteristics of panoramic images of the federation dentaire internationale,a method for auto-matic tooth detection and labeling classification based on parameter optimization was proposed.Advanced deep learning methods were used to construct the innovative and practical three-stage dental panoramic film tooth identification method.The panoramic image was divided into several stages,and the feature extraction process was carried out using SqueezeNet's mask-based convo-lution neural network as the baseline model.The parameter optimization was carried out using swallow swarm optimization algo-rithm.The stage classification model based on SoftMax classifier's tooth prediction and weighted extreme learning machine was applied to determine the tooth number category label.The image data set was evaluated,and the competitive performance was demonstrated.
关键词
深度学习/参数优化/全景片/牙齿检测/牙位标号/燕群优化/加权极限学习机Key words
deep learning/parameter optimization/dental panoramic radiograph/dental examination/tooth position label/swal-low swarm optimization/weighted extreme learning machine引用本文复制引用
基金项目
中国博士后科学基金(2021T140707)
苏州卫生职业技术学院校级领雁培育重点基金(szwzy202004)
出版年
2024
NETL
NSTL
维普
万方数据