腸胃道黏膜微血管透視內視鏡影像處理晶片設計之研究(I)

Abstract

本計畫目標在於開發設計能透視腸胃道黏膜微血管影像的處理晶片。內視鏡影像技術近年來因為較低侵入程度的特性被廣泛使用來取代MRI, CT, PET等腸胃道病灶檢驗技術。為了能不經切片檢查而直接觀察黏膜微血管來發現早期腸胃道癌症，內視鏡技術已發展出利用較短波長光線成像技術來使微血管現形。因為特定的短波長光線被血紅素吸收而造成反射光線的減弱，促成較細血管的成像。這個特定的短波長光線的頻寬為60nm方可產生高對比的影像，故稱為窄波段影像。目前的系統仰賴精密機械來控制濾光片以切換可見光與窄波段影像。機械的控制往往因時間差造成臨床醫師使用上的不便與取像的不易。實務上，臨床醫師通常分開擷取全可見光影像與窄波段影像。如此一來，內視鏡需要較長時間的操作與反覆的抽動。這樣對病患會造成很大的不適感，視內視鏡手術為畏途。因此，本計畫將試圖利用晶片實現電子式的影像合成與強化方式取代傳統精密機械的控制方式。我們將分成兩個策略方向進行以配合臨床醫師不同的需求。第一個方向是將可見光影像與窄波段影像結合同時擷取、各自成像，影像視需求重疊或分開。這樣可使臨床醫師同時對照兩種影像降低診斷的錯誤率。第二個方向是將自動化窄波段影像擷取時機，減少臨床經驗對診斷的影響，同時協助提醒臨床醫師不易察覺的微小病灶。

This project targets on the development of narrow-band gastrointestinal image processing chips. Because the gastrointestinal endoscopy is less invasive, it has been widely applies for detection and treatment of gastrointestinal cancers, as replacement of MRI, CT, and PET examinations. In order for physicians to identify early-stage aberrant crypt foci, the state-of-art endoscopy has employed specific-wavelength lights to unveil image of capillary vessels. Because the center wavelengths of specific-wavelength lights are within the hemoglobin absorption bands, the capillary vessels are identified on the image. The bandwidth of these specific-wavelength lights are normally narrow, say 60 nm, for generating sensible contrast on the image. The image formed by these lights is called the narrow-band image (NBI). The off-the-shelf NBI system performs mode switching and image filtering between white-light image (WLI) and NBI by mechanic controller. The mechanic controller often makes the image acquisition and identification appreciably difficult. Hence, clinicians usually separately capture WLI and NBI clips. This results in long operation time and repeatedly forward-and-backward camera moving which make patients very painful and unacceptable to gastrointestinal endoscopy. Therefore, this project intends to replace the mechanic control approach with electronic image processing. We plan to implement the image reconstruction and enhancement in silicon chips. There are two strategic goals per demands of clinicians in this project. First, the proposed circuits will simultaneously acquire the WLI and NBI, and separately reconstruct them. Two images might be overlapped or pulled apart on the screen. On the other hand, we will automate the turn-on of NBI by aberrant crypt foci detection. This will reduce the dependence of correct diagnosis to clinic experience and assist clinicians to find hard-detectable lesions.