|Abstract: ||口腔癌的五年存活率約57%，若能早期發現，則能提升到80%左右。目前診斷口腔癌的方式仍以切片為主，但切片的判斷會有人為經驗上的誤差，因此病理醫生希望能找出量化診斷的儀器來輔助診斷口腔癌。由於顯微超頻譜影像系統能同時提供切片的組織形態資訊以及光譜資訊，近來被使用在in-vitro口腔癌的檢測，但因其掃描結構複雜且高倍率影像品質不佳，影響診斷準確率。因此本研究發展一套嵌入式繼光鏡超頻譜影像系統，並將其結合顯微鏡成為嵌入式繼光鏡顯微超頻譜影像系統，利用繼光鏡的遠心特性，將本來在顯微鏡物平面的掃描機制轉移到像平面，使用繼光鏡來掃描影像，如此掃描解析度不用隨著放大倍率而改變，不僅增加掃描時的穩定性且將掃描結構簡化。並使用20個老鼠切片來做初步的儀器驗證，得到靈敏度88.9%及特異度87.6%。此外，分別使用雞尾酒方法與組織分層法診斷29個及34個人類病理切片，來驗證這套系統的臨床適用性，前者得到靈敏度90%及特異度87.8%，後者得到靈敏度94.12%及特異度91.18%。最後，本研究也將此系統結合光纖內視鏡，做初步的臨床in-vivo實驗。;The 5-years survival rate of oral cancer is approximately 57%; however, if the oral cancer is discovered in the early stage, the 5-years survival rates can be increased to 80%. Biopsy is a conventional approach for diagnosing oral cancer. In addition to allow pathologists to diagnose the stage of cancer accurately, a biopsy also allows them to prescribe the most appropriate treatment. However, the difference in experience and subjectivity between pathologists might affect their diagnostic accuracy. Therefore, the pathologist hopes to find an approach to quantitatively diagnose oral cancer.|
A microscopic hyperspectral imaging system (MHIS), which can simultaneously present the tissue image and the spectral information of each pixel on an image, has been developed to aid physician to quantitatively diagnose oral cancer. However, the MHIS has some hardware issues, such as the unstable scanning mechanism, complex mechanical structure, and inconvenient alignment. Hence, this study develops a novel embedded relay lens microscopic hyperspectral imaging system (ERL-MHIS), which is composed of the proposed ERL-HIS and a microscope, to improve the hardware issues by optical transferring the scanning plane from objective plane to imaging plane.
In this study, the development of the ERL-MHIS is discussed and the capability of the system is firstly demonstrated by diagnosing early stage oral cancer of twenty mice biopsies. The sensitivity and specificity are 88.9% and 87.6%, respectively. Then, twenty-nine oral cancer (fourteen in early stage) patient’s biopsies are diagnosed using cocktail approach, which diagnoses the biopsy according to the patient’s condition. The sensitivity and specificity are 90% and 87.8%, respectively. Finally, the tissue stratification approach, which combines the morphological and spectral information of biopsy to discriminate cancer form normal biopsies, is used to diagnose thirty-four patient’s biopsies. The sensitivity and specificity are 94.12% and 91.18%, respectively. Moreover, we also connect the ERL-HIS with the fiberscope for in-vivo diagnosis.