医用電子と生体工学
Online ISSN : 2185-5498
Print ISSN : 0021-3292
ISSN-L : 0021-3292
多項式モデルを用いた心臓X線陰影の自動診断
上山 明英斎藤 正男山口 巌木村 栄一
著者情報
ジャーナル フリー

1975 年 13 巻 2 号 p. 78-84

詳細
抄録
The purpose of this study is to diagnose heart diseases automatically from chest X-ray films.
Cardiac shape is represented by a closed curve which is a seventh-order polynomial least-square-approximation. It is shown that the shape of the entire cardiac shadow can be represented by this model. Although this model originally has six parameters, result of principal component analysis indicates that this model can approximately be expressed (up to 98.5% in variance) by only three main components.
By reproducing cardiac shapes using three main components and diagnosing for heart diseases, the distribution of diseases is plotted in a three-dimensional space. If one is to diagnose heart disease from this distribution diagram, classification is correct for 70.0% of typical 23 cases.
Parameters used in this model can be extracted by sequentially fitting the model to an actual X-ray film. Noises such as those of the ribs and gas in stomach are eliminated in this procedure, though there are errors of 5% to 10% for three main components.
著者関連情報
© 日本生体医工学会
前の記事 次の記事
feedback
Top