Development of 3 dimensional medical image diagnosis system using artificial intelligence

利用人工智能开发3维医学图像诊断系统

• 批准号: 22560403
• 负责人: KONDO Tadashi
• 金额: $ 1.41万
• 依托单位: The University of Tokushima
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2010
• 资助国家: 日本
• 起止时间: 2010 至 2012
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-22560403/
• 关键词: システム情報(知識)処理; 人工ニューラルネットワーク; ＧＭＤＨ; コンピュータ支援診断; GMDH; マルチスライスCT; 知識工学

In this study, three dimensional medical image diagnosissystem using artificial intelligence was developed. The artificial intelligence developed in this study, are the revised GMDH-type neural network algorithms and these algorithms can automatically organize the optimum artificial neural network architectures fitting the complexity of the medical images using X-ray CT images. In the conventional GMDH-type neural network algorithms, multi-colinearity occurred and prediction values become unstable. In the revised GMDH-type neural network algorithms developed in this study, the principal component-regression analysis is used and multi-colinearity does not occur and accurate prediction values are obtained. The revised GMDH-typeneural networks ware applied to medical image diagnosis of the lung cancer and the results were compared with the conventional sigmoid function neural network trained using the back propagation method and it was shown that these algorithms were useful for medical image diagnosisof the lung cancer.

本研究开发了基于人工智能的三维医学图像诊断系统。本研究开发的人工智能是修正的GMDH型神经网络算法，这些算法可以自动组织出适合医学图像复杂性的最优人工神经网络结构，使用X射线CT图像。在传统的GMDH型神经网络算法中，存在多重共线性和预测值不稳定的问题。在本文提出的改进的GMDH型神经网络算法中，使用了主成分回归分析，没有出现多个共线性，得到了准确的预测值。将改进的GMDH-典型神经网络应用于肺癌的医学图像诊断，并与传统的用反向传播方法训练的Sigmoid函数神经网络进行了比较，结果表明这些算法对肺癌的医学图像诊断是有效的。

项目成果

A new multi-layered GMDH-type neural network algorithm using principal component-regression analysis

一种新的基于主成分回归分析的多层GMDH型神经网络算法

• 发表时间: 2012
• 期刊: Proceedings of the 43th ISCIE International Symposium on Stochastic Systems Theory and Its Applications
• 作者: Tadashi Kondo;Junji Ueno and Schoichiro Takao
• 通讯作者: Junji Ueno and Schoichiro Takao

Revised GMDH-type neural network using principal component-regression analysis

使用主成分回归分析修正 GMDH 型神经网络

• 发表时间: 2011
• 期刊: Proceedings of SICE annual conference 2011
• 作者: Takuya Kurihara;Kenya Jin'no;Tadashi Kondo and Junji Ueno
• 通讯作者: Tadashi Kondo and Junji Ueno

Nonlinear system identification by feedback GMDH-type neural network with architecture self-selecting function

• DOI: 10.1109/isic.2010.5612889
• 发表时间: 2010-10
• 期刊: 2010 IEEE International Symposium on Intelligent Control
• 作者: T. Kondo;J. Ueno

Feedback GMDH-type Neural Network Self-selecting Various Functions and Its Application to Medical Image Diagnosis of Lung Cancer

反馈GMDH型自选函数神经网络及其在肺癌医学图像诊断中的应用

• 发表时间: 2012
• 期刊: Proceedings of 13thACIS international conference on software engineering, artificial intelligence, networking and parallel distributed computing (SNPD2012)
• 作者: Kondo T.;et al
• 通讯作者: et al

Feedback GMDH-Type Neural Network Algorithm Using Prediction Error Criterion Defined as AIC

• DOI: 10.1007/978-3-642-29920-9_32
• 发表时间: 2012-05
• 期刊: Quality and Reliability Engineering International
• 影响因子: 2.3
• 作者: T. Kondo;J. Ueno;S. Takao