標題: 利用大鼠神經影像預測嗎啡耐藥性
Prediction of morphine tolerance using spinal cord images of central nervous system in rat
作者: 王冠維
Wang, Kuan-Wei
何信瑩
Ho, Shinn-Ying
分子醫學與生物工程研究所
關鍵字: 嗎啡;耐藥性;影像;量化;分類;預測;morphine;tolerance;image;quantify;classification;prediction
公開日期: 2011
摘要: 嗎啡為臨床上最有效的麻醉性止痛劑,然而長時間使用嗎啡會導致耐藥性的產生,其特徵為使鎮痛效果減低,讓嗎啡的臨床治療效果受到限制。許多文獻指出,嗎啡耐藥性與中樞神經系統的細胞表現有關,這些細胞多位於脊髓(spinal cord)中,包含星狀膠細胞(astrocyte)、微小膠細胞(microglia),以及神經細胞(neuron)。到目前為止已有許多研究致力於解決嗎啡耐藥性的問題,其研究方法為試圖開發嗎啡副作用的抑制劑使耐藥性降低,或是開發取代嗎啡的新藥。 上述解決嗎啡耐藥性的相關研究一般透過免疫化學的西方墨點法(Western blot) 來量化細胞表現,其中有少數研究同時觀察細胞免疫螢光影像,然而量化免疫螢光影像的準則通常是人工或電腦輔助的半自動方法,因此難免有不公平性。本篇研究提出了有別於以往的人工量化影像方法,以實驗室開發的影像特徵萃取軟體量化上述三種細胞的免疫螢光影像,影像分為實驗組與對照組,分別是長期施打嗎啡及無施打嗎啡。量化出的特徵共有兩千多個,先以顯著性(p-value)篩選能區分兩組的特徵,接著以智慧型雙目標基因演算法(IBCGA),從剩餘的大量特徵中選出有鑑別兩組能力的特徵集,並經由未被電腦訓練的獨立測試影像的分類結果證實選出特徵的區分能力,並與文獻長使用的特徵做比較。為了選出一組最佳特徵進行分析,進一步用循序向後選擇法將智慧型雙目標基因演算法選中次數前三十名的特徵做篩選,最後選出最佳穩定特徵集,並加以分析。 從IBCGA的獨立測試影像的分類結果顯示,星型膠細胞、微小膠細胞及神經細胞的預測準確率各別為96.67%、78.13%、74.07%,相較於文獻長用的特徵為53.33%、59.38%及55.56%,顯示IBCGA有較高的準確率。同時,分析最佳穩定特徵集,發現有些特徵與文獻描述特徵有關,例如其中之一是長期施打嗎啡使空泡化的神經細胞變多,此現象屬於文獻所提及細胞凋亡的特徵,證實我們選出的特徵具有重要意義,可進一步提供科學家們作為嗎啡耐藥性研究的影像變化觀察指標。
Long-term morphine treatment leads to tolerance, which attenuates analgesia effect and hampers clinical utilization for moderate to severe pain. μ-opioid receptors are well recognized to localize on pre- and post-synaptic nociceptive neurons, but they are also present on astrocytes and microglia, which have been shown to play a role in the development of morphine tolerance Recently, studies have devoted to reveal the mechanism of opioid receptors and tolerance by observing immunostain images of spinal cord cell in order for developing tolerance inhibitors. However, most of these studies consider one common feature for discriminating morphine-tolerant cell images from normal cell images and neither of them tries to propose novel feature. In the present study, we propose a novel feature set for each cell type including astrocyte, microglia, and neuron to robustly discriminate morphine-tolerant cell images from normal cell images. We first use ANOVA and inheritable bi-objective genetic algorithm (IBCGA) combine support vector machine to select optimized feature set from large number of image features. The comparison of classification accuracy between old features from literatures and new features from our algorithm shows that our new features have higher accuracy. Moreover, we analyze the new features and find some phenomenon related to morphine tolerance, which may be useful to the research field of morphine tolerance.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079929511
http://hdl.handle.net/11536/49980
顯示於類別:畢業論文