基于模式识别技术的眼科疾病辅助诊断系统的研究
|
摘要
近年来我国政府越来越重视改善农村的医疗条件,而目前我国乡镇医院的眼科专业医师还比较缺乏,因此,为乡镇医院医生设计一套即能辅助诊疗,又能供经验不足的实习医生学习眼科诊疗过程的眼科辅助诊断系统是十分有意义的。
专家系统在医学领域的应用已经很多见了,但是能被临床医生认可并投入使用的医学专家系统却很少,原因是以往的医学专家系统在设计上存在缺陷。本文在认真分析了医生诊断疾病的实际过程的基础上,指出了传统医学专家系统存在的两个缺陷,并针对以往的医学专家系统的两个缺陷做出了改进。
首先,针对传统医学专家系统的工作过程和医生诊断疾病的过程不一致这个缺陷,本文尝试提出了设计医学专家系统的新思路。其次,以往的医学专家系统,将医学专家的经验整理成多条推理规则,导致了医学专家系统被设计成了一个静止的系统,无法应付多变的疾病症状。针对这一缺陷,本文将模式识别方法应用于医学专家系统,以达到让医学专家系统具有自我学习能力的目的。最后,笔者以眼科专家系统为例,实践了以上提到的对传统医学专家系统缺陷的改进方法。
As China's economic development and the accelerated pace of life in modern society, people work and study are also used to increase the strength of the eye. So the burden of Ophthalmology is growing. Therefore, by using technological means to reduce the burden of ophthalmologists, thereby to improve Ophthalmology's diagnosis and treatment efficiency is very important.
In recent years, Rural People's health problem earns more and more national attention, and in China's rural hospitals for ophthalmic physician is still relatively lacking. In response to the call to improve our country's rural medical conditions, so that let the majority of farmers to enjoy the results of modern technology and modern medicine has great social significance. Also, design an computer-aided diagnosis system for eye disease for the township hospital doctors,and the computer-aided diagnosis system for eye disease could not only Assist ophthalmologists to diagnose patients, but also for the less experienced interns to learn the process of eye disease diagnose and treatment is very necessary.
Expert system applications in the medical field have many met, but little of them can be accepted by the clinical doctors. The reason of this phenomenon is traditional medical expert system has two flaws. The first flaw is:the experience of medical experts to be compiled to a section of standardized, formal rules. Another drawback is:ignore the real process of doctors to diagnose the disease is:first step "consultation" next step "conclusion" and is also "consultation"-"conclusion" alternative process. This paper presented two flaws of traditional medical expert system, and then this paper proposed a medical expert system designed new ideas. This paper argues that a practical and truly Can be accepted medical expert system by doctors should have the following three points:
First, medical experts system should be able to simulate the real process of doctor diagnose a patient.
Second, the positioning of medical expert system should be to support and assist doctors to diagnose illnesses, inspired thinking, not to substitute doctors to make a diagnosis.
Third, the medical expert system has to be self-learning, self-ability to sum up experience.
Based on this new idea, this paper designed an ophthalmic diagnosis system to correction the two existing defects of traditional medical expert systems. This paper improved traditional medical expert systems as follows:
First, use pattern recognition method to instead rule-based reasoning method to diagnose patients'disease. The advantage of this is:a large number of medical experts'valuable time will be saved and let the diagnosis system has the ability to learn. The system is a self-learning system, and with the increase of patient data the system will be more intelligent, more perfect, diagnostic accuracy will be higher.
This paper used matlab simulation software to establish two eye disease diagnosis models. One of them is based on BP neural network method and the other one is based on support vector machine pattern recognition method. In order to simulate doctors'diagnose process, the assisted diagnosis model has two levels and three BP neural networks or three support vector machine. Then we use the patient case data obtained from the Chaoyang District Hospital of Jilin University to train the two diagnosis models and test the two models' ability of classification. According to the test results we optimized the diagnostic model respectively. Finally, test and comparison the two models, analyze test results and choice the better one to complete the Computer-aided Diagnosis System for Eye Disease.
Traditional medical expert systems can only make "conclusion" can not "consultation". To make up for this deficiency the author according to doctors'actual process diagnosis, proposed that a medical experts system working process should be:first step "consultation" next step "conclusion" and also "consultation"-"conclusion" alternative process. The positioning of medical expert system should be to support and assist doctors to diagnose illnesses, inspired thinking.
Finally, the author according to the new medical experts system design ideas, design and implementation an eye disease diagnosis system.
专家系统在医学领域的应用已经很多见了,但是能被临床医生认可并投入使用的医学专家系统却很少,原因是以往的医学专家系统在设计上存在缺陷。本文在认真分析了医生诊断疾病的实际过程的基础上,指出了传统医学专家系统存在的两个缺陷,并针对以往的医学专家系统的两个缺陷做出了改进。
首先,针对传统医学专家系统的工作过程和医生诊断疾病的过程不一致这个缺陷,本文尝试提出了设计医学专家系统的新思路。其次,以往的医学专家系统,将医学专家的经验整理成多条推理规则,导致了医学专家系统被设计成了一个静止的系统,无法应付多变的疾病症状。针对这一缺陷,本文将模式识别方法应用于医学专家系统,以达到让医学专家系统具有自我学习能力的目的。最后,笔者以眼科专家系统为例,实践了以上提到的对传统医学专家系统缺陷的改进方法。
As China's economic development and the accelerated pace of life in modern society, people work and study are also used to increase the strength of the eye. So the burden of Ophthalmology is growing. Therefore, by using technological means to reduce the burden of ophthalmologists, thereby to improve Ophthalmology's diagnosis and treatment efficiency is very important.
In recent years, Rural People's health problem earns more and more national attention, and in China's rural hospitals for ophthalmic physician is still relatively lacking. In response to the call to improve our country's rural medical conditions, so that let the majority of farmers to enjoy the results of modern technology and modern medicine has great social significance. Also, design an computer-aided diagnosis system for eye disease for the township hospital doctors,and the computer-aided diagnosis system for eye disease could not only Assist ophthalmologists to diagnose patients, but also for the less experienced interns to learn the process of eye disease diagnose and treatment is very necessary.
Expert system applications in the medical field have many met, but little of them can be accepted by the clinical doctors. The reason of this phenomenon is traditional medical expert system has two flaws. The first flaw is:the experience of medical experts to be compiled to a section of standardized, formal rules. Another drawback is:ignore the real process of doctors to diagnose the disease is:first step "consultation" next step "conclusion" and is also "consultation"-"conclusion" alternative process. This paper presented two flaws of traditional medical expert system, and then this paper proposed a medical expert system designed new ideas. This paper argues that a practical and truly Can be accepted medical expert system by doctors should have the following three points:
First, medical experts system should be able to simulate the real process of doctor diagnose a patient.
Second, the positioning of medical expert system should be to support and assist doctors to diagnose illnesses, inspired thinking, not to substitute doctors to make a diagnosis.
Third, the medical expert system has to be self-learning, self-ability to sum up experience.
Based on this new idea, this paper designed an ophthalmic diagnosis system to correction the two existing defects of traditional medical expert systems. This paper improved traditional medical expert systems as follows:
First, use pattern recognition method to instead rule-based reasoning method to diagnose patients'disease. The advantage of this is:a large number of medical experts'valuable time will be saved and let the diagnosis system has the ability to learn. The system is a self-learning system, and with the increase of patient data the system will be more intelligent, more perfect, diagnostic accuracy will be higher.
This paper used matlab simulation software to establish two eye disease diagnosis models. One of them is based on BP neural network method and the other one is based on support vector machine pattern recognition method. In order to simulate doctors'diagnose process, the assisted diagnosis model has two levels and three BP neural networks or three support vector machine. Then we use the patient case data obtained from the Chaoyang District Hospital of Jilin University to train the two diagnosis models and test the two models' ability of classification. According to the test results we optimized the diagnostic model respectively. Finally, test and comparison the two models, analyze test results and choice the better one to complete the Computer-aided Diagnosis System for Eye Disease.
Traditional medical expert systems can only make "conclusion" can not "consultation". To make up for this deficiency the author according to doctors'actual process diagnosis, proposed that a medical experts system working process should be:first step "consultation" next step "conclusion" and also "consultation"-"conclusion" alternative process. The positioning of medical expert system should be to support and assist doctors to diagnose illnesses, inspired thinking.
Finally, the author according to the new medical experts system design ideas, design and implementation an eye disease diagnosis system.
引文
[1]王翰虎.基于数据挖掘技术的导诊系统的研究与实现[D].贵阳：贵州大学,2009.
[2]梁嘉骅,王双惠,李常洪等.医疗诊断专家系统开发的新思想与新方法[J].系统工程学报,1999,14(1)：83-90.
[3]周春光,梁艳春.计算智能[M].长春：吉林大学出版社,2006：5.
[4]Simon Haykin.神经网络原理[M].叶世伟,史忠植,译.北京：机械工业出版社,2004：10-11.
[5]丰伟.基于贝叶斯网络的基因调控网络的构建[D].长春：吉林大学,2008.
[6]刘昱昊.融合多数据源构建基因调控网络[D].长春：吉林大学,2010.
[7]Fengrong Sun, Drew Morris, Paul Babyn. The optimal linear transformation-based fMRI feature space analysis [J]. Medical and Biological Engineering and Computing,2009, 47(11):19-29.
[8]林开平.人工神经网络的泛化性能与降水预报的应用研究[D].南京：南京信息工程大学,2007.
[9]吴昌友.神经网络的研究及应用[D].哈尔滨：东北农业大学,2007.
[10]史峰,王小川,郁磊,李洋,张延亮.Matlab神经网络30个案例分析[M].北京：北京航空航天大学出版社,2009：15-17.
[11]温变珍.BP神经网络在大肠癌愈后分析中的应用[D].太原：山西医科大学,2010.
[12]张治国.人工神经网络及其在地质学中的应用研究[D].长春：吉林大学,2006.
[13]刘芬.基于变化检测技术的脑瘤计算机辅助诊断方法应用研究[D].济南：山东大学,2010.
[14]徐振东.人工神经网络的数学模型建立及成矿预测BP网络的实现[D].长春：吉林大学,2004.
[15]邓乃扬,田英杰.数据挖掘中的新方法—支持向量机[M].北京：科学出版社,2004：17-56.
[16]陈金玉.面向英语学习的词重音自动检测与纠错的研究与应用[D].北京：中国科学院自动化研究所,2010.
[17]C. Cortes, V. Vapnik. Support-vector network [J]. Machine Learning,2002,20(9): 273-297.
[18]Chih-Wei Hsu, Chih-Chung Chang, Chih-Jen Lin. A Practical Guide to Support Vector Classification [C]. Taiwan:National Taiwan University,2008.
[19]Chih-Chung Chang, Chih-Jen Lin. LIBSVM:a Library for Support Vector Machines [C]. Taiwan:National Taiwan University,2009.
[20]毛勇.基于支持向量机的特征选择方法的研究与应用[D].杭州：浙江大学,2006.
[21]赵家良.眼科疾病临床诊疗规范教程[M].北京：北京大学医学出版社,2007：11-98.
[22]李玉林,刘敏,高长城,于军,李胜文.中国全科医生实用手册[M].吉林：吉林科学技术出版社,2000：452-458.
[23]张璐璐.基于遗传进化神经网络的人民币号码识别方法研究[D].长春：吉林大学,2009.
[24]姬翔.基于SVM的多病类诊断基因选择方法研究[D].西安：西安电子科技大学,2005.
[25]董国君,哈力木拉提·买买提,余辉.基于RBF核的SVM核参数优化算法[J].新疆：新疆大学学报(自然科学版),2009,8(3)：356.
[26]王小平,曹立明.遗传算法理论、应用与软件实现[M].陕西：西安交通大学出版社,2002：1-50.
[27]Michalewicz Z. Genetic Algorithms+Data Structures=Evolution Programs [M]. Berlin: SpringerVerlag,1992:8-23.
[28]李航.遗传算法求解多模态优化问题的研究[D].天津：天津大学,2007.
[29]罗均,谢少荣等.智能控制工程及其应用实例[M].北京：化工工业出版社,2005：35-46.
[30]金芬.遗传算法在函数优化中的应用研究[D].苏州：苏州大学,2008.
[31]周明,孙树栋.遗传算法原理及应用[M].北京：国防工业出版社,1999：4-43.
[32]邓超.基于支持向量机的流量预测和状态判别研究[D].大连：大连理工大学,2009.
[33]凌萍.基于支持向量技术的聚类分类研究[D].长春：吉林大学,2010.
[34]张学工.关于统计学习理论与支持向量机[J].北京：自动化学报,2000,1(1)：33-35.
[35]董西国.支持向量机在数据挖掘中的应用[D].哈尔滨：哈尔滨工程大学,2002.
[36]白裔峰.偏最小二乘算法及其在基于结构风险最小化的机器学习中的应用[D].成都：西南交通大学,2007.
[37]唐孝.支持向量机(SVM)及其在心电图(ECG)分类识别中的应用[D].成都：四川师范大 学,2007.
[38]夏玫.BP神经网络泛化能力改进研究[D].太原：太原科技大学,2009.
[2]梁嘉骅,王双惠,李常洪等.医疗诊断专家系统开发的新思想与新方法[J].系统工程学报,1999,14(1)：83-90.
[3]周春光,梁艳春.计算智能[M].长春：吉林大学出版社,2006：5.
[4]Simon Haykin.神经网络原理[M].叶世伟,史忠植,译.北京：机械工业出版社,2004：10-11.
[5]丰伟.基于贝叶斯网络的基因调控网络的构建[D].长春：吉林大学,2008.
[6]刘昱昊.融合多数据源构建基因调控网络[D].长春：吉林大学,2010.
[7]Fengrong Sun, Drew Morris, Paul Babyn. The optimal linear transformation-based fMRI feature space analysis [J]. Medical and Biological Engineering and Computing,2009, 47(11):19-29.
[8]林开平.人工神经网络的泛化性能与降水预报的应用研究[D].南京：南京信息工程大学,2007.
[9]吴昌友.神经网络的研究及应用[D].哈尔滨：东北农业大学,2007.
[10]史峰,王小川,郁磊,李洋,张延亮.Matlab神经网络30个案例分析[M].北京：北京航空航天大学出版社,2009：15-17.
[11]温变珍.BP神经网络在大肠癌愈后分析中的应用[D].太原：山西医科大学,2010.
[12]张治国.人工神经网络及其在地质学中的应用研究[D].长春：吉林大学,2006.
[13]刘芬.基于变化检测技术的脑瘤计算机辅助诊断方法应用研究[D].济南：山东大学,2010.
[14]徐振东.人工神经网络的数学模型建立及成矿预测BP网络的实现[D].长春：吉林大学,2004.
[15]邓乃扬,田英杰.数据挖掘中的新方法—支持向量机[M].北京：科学出版社,2004：17-56.
[16]陈金玉.面向英语学习的词重音自动检测与纠错的研究与应用[D].北京：中国科学院自动化研究所,2010.
[17]C. Cortes, V. Vapnik. Support-vector network [J]. Machine Learning,2002,20(9): 273-297.
[18]Chih-Wei Hsu, Chih-Chung Chang, Chih-Jen Lin. A Practical Guide to Support Vector Classification [C]. Taiwan:National Taiwan University,2008.
[19]Chih-Chung Chang, Chih-Jen Lin. LIBSVM:a Library for Support Vector Machines [C]. Taiwan:National Taiwan University,2009.
[20]毛勇.基于支持向量机的特征选择方法的研究与应用[D].杭州：浙江大学,2006.
[21]赵家良.眼科疾病临床诊疗规范教程[M].北京：北京大学医学出版社,2007：11-98.
[22]李玉林,刘敏,高长城,于军,李胜文.中国全科医生实用手册[M].吉林：吉林科学技术出版社,2000：452-458.
[23]张璐璐.基于遗传进化神经网络的人民币号码识别方法研究[D].长春：吉林大学,2009.
[24]姬翔.基于SVM的多病类诊断基因选择方法研究[D].西安：西安电子科技大学,2005.
[25]董国君,哈力木拉提·买买提,余辉.基于RBF核的SVM核参数优化算法[J].新疆：新疆大学学报(自然科学版),2009,8(3)：356.
[26]王小平,曹立明.遗传算法理论、应用与软件实现[M].陕西：西安交通大学出版社,2002：1-50.
[27]Michalewicz Z. Genetic Algorithms+Data Structures=Evolution Programs [M]. Berlin: SpringerVerlag,1992:8-23.
[28]李航.遗传算法求解多模态优化问题的研究[D].天津：天津大学,2007.
[29]罗均,谢少荣等.智能控制工程及其应用实例[M].北京：化工工业出版社,2005：35-46.
[30]金芬.遗传算法在函数优化中的应用研究[D].苏州：苏州大学,2008.
[31]周明,孙树栋.遗传算法原理及应用[M].北京：国防工业出版社,1999：4-43.
[32]邓超.基于支持向量机的流量预测和状态判别研究[D].大连：大连理工大学,2009.
[33]凌萍.基于支持向量技术的聚类分类研究[D].长春：吉林大学,2010.
[34]张学工.关于统计学习理论与支持向量机[J].北京：自动化学报,2000,1(1)：33-35.
[35]董西国.支持向量机在数据挖掘中的应用[D].哈尔滨：哈尔滨工程大学,2002.
[36]白裔峰.偏最小二乘算法及其在基于结构风险最小化的机器学习中的应用[D].成都：西南交通大学,2007.
[37]唐孝.支持向量机(SVM)及其在心电图(ECG)分类识别中的应用[D].成都：四川师范大 学,2007.
[38]夏玫.BP神经网络泛化能力改进研究[D].太原：太原科技大学,2009.