Parameters Estimation of Impulse Response in Compartment Model Using Cumulative Function and Linear Regression Analysis

カテゴリ: 論文誌(論文単位)

グループ名: 【C】電子・情報・システム部門

発行日: 2023/02/01

タイトル(英語): Parameters Estimation of Impulse Response in Compartment Model Using Cumulative Function and Linear Regression Analysis

著者名: Toshiki Tanaka (KINKEI SYSTEM CORPORATION), Tetsuo Hattori (Kagawa University), Yusuke Kawakami (National Institute of Technology, Kagawa College), Yo Horikawa (Kagawa University), Yoshiro Imai (Kagawa University)

著者名(英語): Toshiki Tanaka (KINKEI SYSTEM CORPORATION), Tetsuo Hattori (Kagawa University), Yusuke Kawakami (National Institute of Technology, Kagawa College), Yo Horikawa (Kagawa University), Yoshiro Imai (Kagawa University)

キーワード: parameter estimation，compartment model，cumulative function，linear regression analysis，PET inspection

要約(英語): We propose a novel and simple estimation method of parameters that appear in the impulse response of compartment model, using cumulative function and Linear Regression Analysis, taking the case of PET (Positron Emission Tomography) inspection for example. In PET inspection, the parameters estimation is a very important for the analysis of dynamic function and the presence of cancer, etc. Since the important parameters to be estimated are intricately included in the exponent part of exponential function in impulse response of the compartment model, the Conjugate Gradient Method (CGM), and the extended Newton one, etc., are normally used. However, there is a problem that these iterative computation methods do not converge or take much time, depending on the initial values of those parameters. In this paper, we present an alternative method that does not need to set the initial values of parameters in computing. This paper also shows that the proposed method works well, illustrating the experimental results in comparison with the CGM.

本誌: 電気学会論文誌C（電子・情報・システム部門誌） Vol.143 No.2 （2023） 特集：エネルギー分野へ適用されたAI・IoT技術

本誌掲載ページ: 185-191 p

原稿種別: 論文／英語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejeiss/143/2/143_185/_article/-char/ja/