數位光源在醫學頻譜影像之研究

Abstract

本計畫結合數位頻譜影像技術與內視鏡光學系統，利用新式數位光源於不同時序照射特定的白光頻譜，增加取樣速度，希冀達到即時頻譜醫學影像功能，幫助醫師做臨床診斷。以內視鏡應用於早期消化道檢測為例，傳統的醫學取像只透過紅綠藍三個通道，僅可提供特定色彩影像，然而初期癌細胞仍侷限於腸壁組織的黏膜層(原位癌)、或是部份黏膜下層(T1癌瘤)，尚未有遠端器官擴散，不易透過肉眼做自然色辨認，因此透過頻譜影像的協助，可以在早期病徵出現時即發現並開始治療。本研究利用新型的數位發光二極體光源系統，可以調整任意波段的光源頻譜，取代使用彩色濾光片或傳統掃瞄機制，經由主成分分析等矩陣式數位計算，可快速獲得生醫樣品之頻譜影像，配合醫師的診斷，以期對於早期的腫瘤組織異常或病變狀況做正確的治療與判斷。

This project is aimed to develop an endoscopic multispectral system based on a digital light emitting diode (LED) illumination. Proposed architerture is able to increase the interpretation quality of various modalities in medical applications. Taking detection of early malignancy (malignant polyp) in the bowel lumen as the example, we will implement a platform composed of a spectrally tunable LED light sources, multi-channel camera acqusition system and corresponding color calibration algorithm. Rather than using converntional scanning optimechincs or color filter scheme, we provide a multiplexing sequence of spectral illumination so as to minimize the number of acquired images and acqisition time. Based on principal component analysis and pseudo inverse algorithm, we estimate the the spectral reflectance of the captured medical image (eg. mucosal lesions of upper and lower gastrointestinal tract, microaneurysms of retinal blood vessel image). A proof-of-concept prototype will be demonstrated. In addition, color discrimination will be investigated to find the other early symptom via illumination manipulation.