Presentation | 2010-11-15 Choice Method of Classifier for Computer Aided Diagnosis Shunsuke HORIE, Tomoko MATSUBARA, Satoshi KASAI, Yoshikazu UCHIYAMA, Chisako MURAMATSU, Xiangrong ZHOU, Takeshi HARA, Hiroshi FUJITA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Various classifiers are used on the elimination of false positives or diagnosis of benign and malignant lesion. However, no classifier shows consistently superior performance regardless of the nature of the data. Therefore, it is necessary to clarify the selection criterion of the classifiers that shows optimal performance based on the nature of the feature used. Moreover, the available data for training is finite-sized in an experiment step of CAD development. Consequently, it is important to select the classifier that indicate steady classification performance applying to unknown large data set, based on one obtained by training. In this paper, we conducted the comparison study for diagnosis performance of five classifiers, namely, Linear Discriminant Analysis, Quadratic Discriminant Analysis, Neural Network, Support Vector Machine, and AdaBoost. Our dataset consists of benign and malignant lesions extracted clinical cases. The distribution analysis of features was performed. As a result, it becomes clear that the effects of nature of the features and number of training data on classifier performance are different each classifier. We propose the possibility of the selection method for classifier based on our results. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Classifier / Clinical Data / nature of the data / AUC / Computer-aided diagnosis(CAD) |
Paper # | MI2010-71 |
Date of Issue |
Conference Information | |
Committee | MI |
---|---|
Conference Date | 2010/11/8(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Medical Imaging (MI) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Choice Method of Classifier for Computer Aided Diagnosis |
Sub Title (in English) | |
Keyword(1) | Classifier |
Keyword(2) | Clinical Data |
Keyword(3) | nature of the data |
Keyword(4) | AUC |
Keyword(5) | Computer-aided diagnosis(CAD) |
1st Author's Name | Shunsuke HORIE |
1st Author's Affiliation | Graduate School of Medicine, Gifu University() |
2nd Author's Name | Tomoko MATSUBARA |
2nd Author's Affiliation | School of Information Culture, Nagoya Bunri University |
3rd Author's Name | Satoshi KASAI |
3rd Author's Affiliation | Konica Minolta Medical Imaging USA, INC. |
4th Author's Name | Yoshikazu UCHIYAMA |
4th Author's Affiliation | Oita National College of Technology |
5th Author's Name | Chisako MURAMATSU |
5th Author's Affiliation | Graduate School of Medicine, Gifu University |
6th Author's Name | Xiangrong ZHOU |
6th Author's Affiliation | Graduate School of Medicine, Gifu University |
7th Author's Name | Takeshi HARA |
7th Author's Affiliation | Graduate School of Medicine, Gifu University |
8th Author's Name | Hiroshi FUJITA |
8th Author's Affiliation | Graduate School of Medicine, Gifu University |
Date | 2010-11-15 |
Paper # | MI2010-71 |
Volume (vol) | vol.110 |
Number (no) | 280 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |