Japanese
Titleシンチグラム処理における定量的評価の試み
Subtitle原著
Authors福久健二郎*, 飯沼武*, 永井輝夫**
Authors(kana)
Organization*放射線医学総合研究所, **群馬大学医学部放射線医学教室
Journal核医学
Volume17
Number2
Page155-166
Year/Month1980/2
Article原著
Publisher日本核医学会
Abstract「要旨」電子計算機によるRI像処理の研究は, 1964年ごろから急速に進展した. しかし, これらの処理法のうち, どれが最適かを客観的に判定するのは容易ではない. この客観的方法として総得点法, ROC解析, McNeilらの方法が知られているが, 本報ではこれらの特性を検討する目的でIAEAの電算機シミュレーテッド・ファントムを用い, それぞれの方法で相互比較した. 処理法はガウス分布のフィルタとオプティマム・フィルタを用い, 同一条件でプロッタに表示し, 4名のシンチグラム専門医に読影願った. また, 判定法は主にIAEAが示す方法を用いた. 結果として, いずれの比較方法によっても, 個人差や処理法による差は明確になるが, 得点法ではFNの評価法, ROC解析ではFPがない場合の比較に問題があるほか, 誤診の重大さをどう個々の臨床評価で考慮するかが大切であることを示した.
Practice臨床医学：一般
KeywordsSintigraphic image processing, Computer-simulated phantom, Image display, ROC curve
English
TitleAn Analytical Study of Scintigram Interpretation
SubtitleOriginal Articles
AuthorsKenjiro FUKUHISA*, Takeshi A. IINUMA*, Teruo NAGAI**
Authors(kana)
Organization*National Institute of Radiological Sciences, **Department of Radiology, Gunma University
JournalThe Japanese Journal of nuclear medicine
Volume17
Number2
Page155-166
Year/Month1980/2
ArticleOriginal article
PublisherTHE JAPANESE SOCIETY OF NUCLEAR MEDICINE
Abstract[Summary] : Various image processing methods are available for transforming scintigraphic image quality. However, it is not so easy to determine the best processing method in view of diagnosis of diserse by physicians. In order to know the effect of image quality on the physician's interpretation, 24 images of computer-simulated phantom which was established in the first coordinated programme on the "Intercomparison of Computer-assisted Scintigraphic Teshniques" sponsored by IAEA, are processed by using the 'Optimum Filter' with medium frequency enhancing effect (OF-method) and the Gaussian filter with simple smoothing effect (GF-method) , and are displayed as a contour map with same display factors by a curve plotter of computer system. The processed and displayed 48 images in random order are viewed by f our physicians who are specialists in scintigraphic diagnosis, with the suggestions as follows : (1) The images are consisted of 48 pictures., (2) Each images is divided into four quadrants, each of which could contain either one lesion or none, Therefore, each images could contain from 0 to 4 lesions, (3) All lesions in the phantom are either cold or hot, but there are no mixed lesions of both in a picture, (4) The probability of presence of a lesion in a given quadrant is taken to be about 2/3, and (5) The lesions observed should be entered with following figure of rating from 1 to 4 indicating the degree of confidence ; 1 : definitely positive, 2 : probably positive, 3 : possibly negative and 4 : difinitely negative. The results of their answers are analyzed with (1) total score method, (2) RCC curves and (3) values defined by McNeil. Among the results, the OF-method gives significantly better results compared with the GF-method, though the former are increased number of false positive compared with the latter. The cold lesions are more detectable compared with hot lesions of the phantom. The cold lesions are similar to a scintigram of human liver administrated 99mTc-colloid. Further investigation is being directed what kind of display can give the best results by using various displaying devices for same processed images, as the display method in this study seems to be not suitable for detecting the lesions in scintigraphic images.
PracticeClinical medicine
KeywordsSintigraphic image processing, Computer-simulated phantom, Image display, ROC curve

【全文PDF】