開發可用於多尺度光學同調斷層顯微術之三維影像重合對位演算法

Development of volumetric image registration algorithm for multiscale optical coherence microscopy (OCM) images

劉懷仁(Huai-Jen Liu)

指導教授 ： 李翔傑

國立臺灣大學/電機資訊學院/光電工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202200200

摘要

在此研究論文中，我們使用一套實驗室自主開發的近紅外光II-b波段的掃頻式光學同調斷層顯微術（SS-OCM）系統。本套系統所使用的光源中心波長為1688奈米，光源輸出頻寬則為175奈米，波長掃描頻率則為90 kHz。透過使用一中心波長位於長波長的掃頻光源可以在散射介質中提供更加的成像深度，但其系統的成像解析度則會有所減少。因此可進一步將光學同調斷層掃描術系統及共焦顯微鏡系統做結合，以提升系統本身的橫向解析度。所開發的系統在使用5倍物鏡下的橫向成像解析度約24.8微米，軸向解析度約17.4微米，整體的光學干涉訊號偵測靈敏度則為96.3 dB。
在放置組織樣本卡匣的設計上為了避免近紅外光在掃描樣本時光線垂直反射回系統造成系統所使用的光電偵測器產生訊號飽和，因此將卡匣以斜面的方式來設計。但考量到卡匣若是以斜面呈現，會造成在光點在樣本中的聚焦位置會隨不同的橫向掃描位置進行變化，進而影響最後所欲呈現的正視(en-face)OCM影像強度以及對比度。因此，我們針對所使用特製化設計樣本卡匣取得廣域OCM影像，透過拉平OCM影像表面位置方式配合使用一可自動化調整聚焦位置之縱向電動平移台，進而提升廣域OCM影像品質。
另外，樣本端為了增加OCM影像的橫向解析度，因此再加入10倍及20倍高數值孔徑的物鏡。雖透過使用高數值孔徑的物鏡可以有效提升影像的解析度，但卻導致成像範圍縮小。為了要同時保有高解析度以及廣域的影像，於此論文中，我們開發出一套馬賽克拼接演算法來對有限視野的影像進行後處理的拼接，且在不降低解析度的情況下對影像進行亮度調整以提升拼接後之廣域OCM影像品質。
為了證實所開發演算法的穩定性和效能，我們利用前述系統針對在台大醫院牙科部所取得之口腔粘膜組織樣本進行多尺度OCM成像，並且利用該演算法進行廣域OCM影像拼接。透過檢視取得之組織樣本廣域OCM影像，可以證實該演算法於自動化拼接產生具有優化成像品質之廣域OCM的可行性。在未來的研究中，我們將致力於縮短在影像處理上的時間，以及在影像拼接時雜訊的處理方式，來呈現更加完善的高解度、廣域的SS-OCM影像。

關鍵字

光學同調斷層顯微術 ； 廣域 ； 影像處理 ； 馬賽克拼接 ； 定位 ； 亮度調整

並列摘要

In this thesis work, we used an in-house developed swept-source optical coherence microscopy (SS-OCM) system operating at the near-infrared II-b wavelength regime to perform volumetric imaging. The light source used in the system exhibits a central wavelength of 1688 nm, an optical bandwidth of 175 nm, and an A-scan rate of 90 kHz. Using the long-wavelength swept source, although it can provide a deeper imaging depth in the scattering medium during imaging, the imaging resolution is degraded. Therefore, by combining the optical coherence tomography (OCT) system with the confocal microscopy, the lateral resolution of the developed OCM system can be further improved. Using a 5X objective in the sample arm, the developed OCM system supported OCM imaging with a lateral resolution of 24.8 μm, an axial resolution of 17.4 μm, and detection sensitivity of 96.3 dB.
In the design of tissue sample cassette, in order to prevent the near-infrared light from being reflected vertically back to the system leading to the saturation of the balanced detector used in the system while imaging the samples, the top surface of sample cassette is designed with an inclined plane. However, given the above cassette design with an inclined top surface plane, it will compromise the intensity and contrast of the en face OCM image due to the varying focus position within the sample at different lateral imaging locations. Therefore, we have developed surface flattening algorithm and incorporating a z-axis motorized stage allowing automatic control of the focusing position within the sample, and have applied the above approach on the acquired wide-field OCM images to further improve the quality of the OCM images.
In addition, in order to improve the lateral resolution of the OCM images, we have introduced two high-numerical aperture (NA) objectives into the sample arm, including 10X and 20X objectives. Although the image resolution is effectively improved with the high-NA objectives, the imaging field is reduced. In order to provide high-resolution and wide-field OCM images of the samples, in this thesis, we developed a mosaic-stitching algorithm to compositing multiple individually-acquired OCM image to generate wide-field OCM images in post processing. The developed algorithm also supports automatic adjustment of the brightness of the OCM images without affecting the image resolution to improve the quality of the composited wide-field OCM images.
In order to verify the robustness and effectiveness of the developed algorithms, we first performed wide-field OCM imaging of the freshly-resected oral tissue sample collected from the Department of Dentistry at the National Taiwan University Hospital (NTUH) using the OCM system aforementioned. Then, we applied the developed algorithms on the acquired OCM image to composited wide-field OCM images. By examining wide-field OCM images of the imaged tissue samples, we have demonstrated that the developed algorithm exhibits the feasibility of automatically generating wide-field OCM images with optimized imaging quality. In future research, we will focus on decreasing the computation time for image processing as well as for suppressing noise during the imaging stitching or compositing step, promising to develop a more optimized high-resolution and wide-field OCM imaging system.

並列關鍵字

optical coherence microscopy ； wide-field ； imaging processing ； mosaic stitching ； registration ； brightness adjustment

參考文獻

[1] K. Özduman, E. Yıldız, A. Dinçer, A. Sav, and M. N. Pamir, "Using intraoperative dynamic contrast-enhanced T1-weighted MRI to identify residual tumor in glioblastoma surgery," Journal of neurosurgery, vol. 120, no. 1, pp. 60-66, 2014.

[2] E. Uhl et al., "Intraoperative computed tomography with integrated navigation system in a multidisciplinary operating suite," Operative Neurosurgery, vol. 64, no. suppl_5, pp. ons231-ons239, 2009.

[3] S. Dasarathy, J. Dasarathy, A. Khiyami, R. Joseph, R. Lopez, and A. J. McCullough, "Validity of real time ultrasound in the diagnosis of hepatic steatosis: a prospective study," Journal of hepatology, vol. 51, no. 6, pp. 1061-1067, 2009.

[4] J. Fujimoto et al., "New Technology for High‐Speed and High‐Resolution Optical Coherence Tomography a," Annals of the New York Academy of Sciences, vol. 838, no. 1, pp. 95-107, 1998.

[5] S. A. Boppart, J. Herrmann, C. Pitris, D. L. Stamper, M. E. Brezinski, and J. G. Fujimoto, "High-resolution optical coherence tomography-guided laser ablation of surgical tissue," Journal of Surgical Research, vol. 82, no. 2, pp. 275-284, 1999.