利用影像分析及顯示技術於顱顏部腫瘤之診斷及手術輔助

Abstract

長久以來， 醫學影像一直是顱顏部位診斷所倚重的診斷工具。近年來，隨著科技之進步，此類影像已經從傳統之黑白平面影像，進步到多色之立體模式。隨著影像之數位化，各式軟體及資訊科技之進步，現今醫學影像之範疇，早已超脫傳統成是解剖及病理構造之範疇，進一步而能藉由影像分析等之進階技術，在術前診斷，手術計畫，甚至術後追蹤做出貢獻。 目前此類影像分析之研究，在顱內腫瘤有許多之研究成果，但在顱顏外部腫瘤之影像分析，卻少見相關文獻。這是因為此區域大規模且具侵犯性之良性腫瘤或增生，在臨床上相對少見，且不管是在位置上或質地上，都較顱內腫瘤更為多變且歧異，因此研究較難有定論，在臨床治療上也更具挑戰性。 因此本研究之主題，為嘗試利用進階影像分析技術中之影像擷取，做為為顱顏區域良性腫瘤診斷之工具及手術輔助。本論文基本上分為兩部分，第一部分為介紹一種利用模糊分類演算法為基礎之多階段影像擷取技術，以用來量測顱內部之腫瘤。第二部分則為嘗試活用此種技術來解決臨床上更具挑戰性之題材，即顱顏外部良性腫瘤之擷取。鑑於此區域病灶及環境之多樣化，因此需要更靈活之策略運用，才能獲致理想之影像擷取結果。我們並利用適當之影像顯示，來表現影像分析之成果，以利臨床之應用。最後並以數個病例來呈示此研究在臨床醫療之應用及價值。我們認為這些科技能有效的幫助術前診斷，輔助顱顏外科之手術，在臨床上做出貢獻。其中發展出之介面又能和將來之高科技技術，例如導航系統，醫學模型建置等契合，因此發展之潛力不可限量。

For years the medical image study has been a reliable and important tool for pre-operative diagnosis in the craniofacial surgery domain. With the improvement of the imaging instrument, and the progression of the computer hardware, now these imaging technique had went beyond simple visualization, the emerging of more complicated and advanced image analysis technique had extend the utility of image diagnosis. To meet the specific and complex requirements of these biomedical image analyses, many researchers had devoted themselves in the development of the analyzing algorithm. Among them, algorithm used for isolating the meaningful component or pathology from the medical image, called image segmentation, had been studied intensively in recent years. However, most of the researches were focused on the neoplasm detection over intra-cranial space, and studies regarding the extensive neoplasm and hyperplasia over the extra-cranial and facial area were few. This may due to the its small case number and more variable clinical presentation, combined with more complicated anatomy make conclusive result more difficult in this area. So, in this research, will try to develop a feasible algorithm and also a strategy to sucessfully isolated the neoplasm over the craniofacial area. The research comprised mainly two parts, at the first part we will introduce a multi-stage algorithm based on Fuzzy-c-mean technique for image segmentation of the intra-cranial tumor. On second part of our study, we’ll extend the use of this algorithm to more challenging extra-cranial lesion, that is, the benign neoplasm and hyperplasia over the craniofacial region. Due to more variable of the tumor location and the heterogeneous character of the tumor images, a more flexible and freely-used strategy is needed here for optimizing the result of image analysis for individual case. We’ll also introduce a visualization method that will properly demonstrate the results of these image analyses. Finally we will present few clinical cases in order to the contribution of our research to clinical practice. We think these techniques could effectively help us in pre-operation diagnosis, surgical planning and post-operative follow-up. This technique could be easily interfaced with other Hi-tech instruments, and the potential for further development is promising.