基於深度多工單次多視向光學同調斷層掃描系統的開發

Abstract

隨著科技的快速發展，已有多種口內成像技術應用在牙科臨床上，能有效地輔助牙醫師進行口腔疾病診斷。目前常用的電腦斷層影像(Computed tomography, CT)，具游離輻射，對人體有潛在危險，而光學同調斷層掃描術(Optical Coherence Tomography, OCT)則無此困擾，其非侵入性、高解析度及快速成像的特性，被認為是一種很有發展前途的口內成像技術。它是一種基於低同調干涉儀(Low coherence interferometry, LCI)的技術，透過量測參考端與樣品端的干涉訊號來獲取生物組織的斷層影像。然而，由於牙體硬組織的光學特性，影像中常出現垂直偽影(Vertical line artifact)。這種偽影限制了OCT在實際臨床應用中對病灶特徵的解讀。 基於以上所述，我們希望可以由不同視角對樣本進行掃描，使入射光避開可能影響成像之結構，並藉由折射率校正暨影像拼接演算法，將影像先做校正，再將其進行融合，以補足單一視角下所缺失的影像資訊，進一步提升影像的判讀性。因此，本論文開發了一種基於深度多工技術(Depth-encoded multiplexing)與光學時脈頻率倍增模組(Optical frequency clock doubling circuit module)的新型掃頻源光學同調斷層掃描(Swept-source OCT, SS-OCT)成像系統。該系統無須移動樣本，即可在單次掃描中擷取牙齒樣本在三種不同視角下的OCT影像。 此外，本實驗室團隊開發了一套畸變與折射率校正暨基於迭代最近點演算法(Iterative closest point, ICP)的多視角影像拼接演算法，用以融合多視角的OCT影像。該演算法能有效修正因光學系統設計導致的枕狀畸變，以及因忽略折射率效應而造成的幾何扭曲，再利用ICP演算法，將多視向牙齒OCT影像進行對位並融合，來減少偽影且提高解析度。 未來，我們希望能以此論文為基礎，進一步將系統開發成手持式之微型多視向口內掃描探頭，提供即時的多視向拼接影像，提升其在臨床上的實用性，協助牙醫師能更有效地進行口內疾病診斷。

With the rapid development of technology, various intraoral imaging techniques have been applied in dental clinics to effectively assist dentists in diagnosing oral diseases. Currently, the most commonly used method is Computed Tomography (CT), but its use of ionizing radiation poses potential health risks. In contrast, Optical Coherence Tomography (OCT) does not pose such concerns. OCT is considered a highly promising intraoral imaging modality due to its non-invasive nature, high resolution, and fast imaging capabilities. It is based on a low-coherence interferometer, which obtains tomographic images of biological tissues by measuring the interference signals between the reference arm and the sample arm. However, due to the optical properties of dental hard tissues, vertical line artifacts often appear in OCT images. These artifacts limit the interpretation of lesion characteristics in actual clinical applications. Based on the above, we aim to scan the sample from different angles, allowing the incident light to avoid structures that may affect imaging. Through refractive index correction and image stitching algorithms, we correct and then fuse the images to compensate for the missing information in single-view imaging, thereby enhancing interpretability. Therefore, this study developed a novel swept-source optical coherence tomography (SS-OCT) imaging system based on depth-encoded multiplexing and optical frequency clock doubling. The system can acquire OCT images of a tooth from three different viewing angles in a single scan without moving the sample. In addition, our laboratory team has developed a multi-view image registration algorithm based on distortion and refractive index correction, combined with the Iterative Closest Point (ICP) algorithm. This algorithm can effectively correct the pincushion distortion caused by the optical system design and the geometric distortion caused by the refractive index. After correction, the surface morphology of the tooth is extracted by calculating the maximum image gradient to generate point clouds. The ICP algorithm is then used to compute displacements between these point clouds to align and fuse the multi-view OCT images, reducing artifacts and enhancing resolution. In the future, we hope to further develop the system into a handheld, miniature, multi-view intraoral scanning probe based on this study, providing real-time multi-view stitched images. This would further enhance its clinical practicality and assist dentists in making more effective diagnoses of oral diseases.