99mTc-TRODAT-1單光子電腦斷層影像活性分佈及多巴胺轉運體活性體積之合理分類模型

Feasible Classified Models for TRODAT-1 SPECT Activity Distribution and Dopamine Transporter Activity Volume

涂博偉(Po-Wei Tu)

指導教授 ： 陳泰賓

義守大學/電機資訊學院/資訊工程學系/碩士(2015年)

https://doi.org/10.6343/ISU.2015.00130

摘要

99mTc-TRODAT-1單光子電腦斷層(Single Photon Emission Computed Tomography, SPECT)造影技術，已成為臨床常規評估腦部多巴胺傳導物質系統相關疾病之檢查。文獻上經常採用感興趣區域(Region of Interest, ROI)比值方式，量化紋狀體特異性攝取率，評估腦部神經傳導物質相關疾病，此方法仍有改善之空間。
本研究採用回顧性分組實驗設計，共收集202筆99mTc-TRODAT-1 SPECT影像及診斷報告，包括正常腦部6筆及帕金森氏病(Parkinson's Disease, PD)196筆。根據帕金森氏病嚴重程度又分為帕金森氏病HYS(Hoehn and Yahr Scale) I、II、III組(102筆)及帕金森氏病HYS IV、V組(94筆)。利用三維方式估算活性分佈及紋狀體活性體積之影像特徵共六個，採用統計無母數檢定找出具有顯著性之特徵值，再經由Logistic Regression(LR)及Support Vector Machine(SVM)分類方法進行建模，並評估合理單光子電腦斷層影像之帕金森氏病分類模型。
結果顯示SVM分類模型於解釋變數中使用偏態值、峰度值及多巴胺轉運體活性體積三個特徵值，分辨正常腦部及帕金森氏病之靈敏度、特異性、陽性預測率、陰性預測率、準確度及Kappa值皆高達99％。分辨正常腦部、帕金森氏病HYS I、II、III 組及帕金森氏病HYS IV、V組之靈敏度、特異性、陽性預測率及準確度均高於80%且Kappa值為68%。證實此模型能有效判定不同嚴重程度之帕金森氏病，期能提供臨床醫師診斷帕金森氏病嚴重程度之參考依據。未來仍須更多無異常影像之案例，以了解分類模型之偽陽性變化情形。

關鍵字

帕金森氏病 ； 活性分佈 ； 紋狀體活性體積

並列摘要

The single photon emission computed tomography (SPECT) is widely applied to image activity of dopamine transporters with 99mTc-TRODAT-1. The ratio of regions of interest (ROI) is often used to quantify the uptake of 99mTc-TRODAT-1 located stratum. However, the improvement of evaluated abnormality of activity of dopamine transporters in stratum has to precisely estimation by SPECT with 99mTc-TRODAT-1.
The retrospective study was adopted to collect 202 cases with SPECT images via 99mTc-TRODAT-1. There were 6 negative cases and 196 Parkinson's disease. The 196 Parkinson's cases was divided into two groups. One group is according to the HYS (Hoehn and Yahr Scale) were I, II and III levels with involved 102 cases. The other group is according to HYS IV and V levels with contained 94 cases. The imaging features were six characteristics related to activity distribution of whole brain and activity volume of striatum by 3D SPECT. The significant features were identified by none parametric statistical testing. The logistic regression (LR) and support vector machine (SVM) methods were modeled by those significant features. Meanwhile, the validation of LR and SVM were assessed by testing and training data.
The features of skewness, kurtosis and dopamine transporter activity volume were used in SVM with sensitivity, specificity, positive predictive value, negative predictive value, accuracy and kappa values are over 99% by SVM model for two groups (normal vs abnormal). The sensitivity, specificity, positive predictive value and accuracy are over 80% by SVM model for three groups (normal, HYS I-II-III, HYS IV-V). The Kappa value classified by SVM for three groups is 0.68 that means substantial agreement with physician judgment.
The features of skewness, kurtosis and dopamine transporter activity volume computed from 3D SPECT with 99mTc-TRODAT-1 are not only significant predictors, but also useful to build SVM model for three groups. The more negative cases are need to figure the variation of false positive rate in future work.

並列關鍵字

99mTc-TRODAT-1 ； Activity Distribution ； Activity Volume of Striatum

參考文獻

[1] Vaughan RA, Foster JD. Mechanisms of dopamine transporter regulation in normal and disease states. Trends in pharmacological sciences. 2013;34.9:489-496.

[2] Merchant BA, Jeffry DM. Insights from molecular dynamics: The binding site of cocaine in the dopamine transporter and permeation pathways of substrates in the leucine and dopamine transporters. Journal of Molecular Graphics and Modelling. 2012;38:1-12.

[3] Kordower JH, et al. Disease duration and the integrity of the nigrostriatal system in Parkinson’s disease. Brain. 2013;136.8:2419-2431.

[4] Leverenz JB, et al. Cognitive impairment and dementia in patients with Parkinson disease. Current topics in medicinal chemistry. 2009;9.10:903.

[5] Chou KL, et al. A recommended scale for cognitive screening in clinical trials of Parkinson's disease. Movement Disorders. 2010;25.15:2501-2507.