Week 3: Eddie Wei - Multimodal Endoscope

Figure 1: Multimodal OCT and Confocal Setup Housed in moving cart connected to Endoscope for operations

Endoscope Design

        This week, I finalized the optical setup and handheld endoscopic probe design in SolidWorks CAD for multimodal imaging with Optical Coherence Tomography (OCT) and Confocal Microscopy in both 360° side-view scanning and forward-view orientations, as shown in the setup and endoscope CAD figures above.

Figure 2: Current OCT-Red (Freund et al.)  and Confocal Fluorescence Microscopy-Green (Wang et al.) images in comparison to Histology.

        As a brief review of their demonstrated applications in urology, OCT uses backscattered near-infrared light to generate depth-resolved structural images and can delineate bladder layers up to the muscularis with limited cellular information [1], while confocal fluorescence endoscopy uses fluorescein-induced emission and optical sectioning to distinguish normal urothelial cells from low- and high-grade carcinoma at shallow surface depths with approximately 1 μm resolution [2]. The current design builds on techniques from prior multimodal endoscopic systems, including Lissajous Multi-modal Endomicroscopy with Optical Coherence Tomography and Confocal Fluorescence Microscopy by Lee et al. [3] and An All-fiber-optic Endoscopy Platform for Simultaneous OCT and Fluorescence Imaging by Mavadia et al. [4]. The optical setup will be housed on a cart and connected to the handheld probe through optical fibers and a computer interface (Figure 1). For image generation and processing, the OCT channel detects backscattered light with a spectrometer to generate depth-resolved structural images, while the confocal fluorescence microscopy channel detects pinhole-filtered fluorescence signals with a photomultiplier tube to produce high-resolution near-surface images. Both signals are sent to and processed by the same computer for synchronized multimodal image reconstruction.

Figure 3: Close Up of Solidworks CAD File of Endoscope Design and Components

My proposed endoscopic probe 3D model consists of two major parts, each integrating design features from both publications while addressing limitations of the previous approaches (Figure 3). Part 1 focuses on generating OCT and confocal fluorescence microscopy images from the same targeted tissue region. A key feature adapted from the multimodal Lissajous paper is the double D-shaped fiber configuration, which is used to reduce lateral optical misalignment and support a compact forward-viewing probe. In this design, the single-mode fiber (SMF) for OCT and the double-clad fiber (DCF) for confocal fluorescence microscopy are side-polished and bonded with epoxy to form a double D-shaped fiber structure. As a result, each fiber originally has a 125 µm outer diameter was side-polished by 47 µm, leaving 15.5 µm of remaining radial thickness on the polished side. Because the OCT and confocal beams originate from different cores, simply placing the fibers side by side could cause lateral optical-axis mismatch. This mismatch could cause the two modalities to image laterally offset tissue regions, reducing the accuracy of OCT–confocal comparison.

       Because confocal fluorescence microscopy and OCT operate at widely separated wavelengths, 488 nm and 1300 nm, respectively, a simple lens would not focus both beams at the same axial position. This wavelength-dependent focal mismatch could make it difficult to accurately correlate cellular-scale confocal features with the corresponding OCT structural image. To mitigate this chromatic axial shift, the multimodal probe uses a custom dual-mode objective lens stack designed for both OCT and confocal imaging based on the specifications established of the Lissajous multimodal paper (Li et al.) [3]. The objective is corrected over the 488–550 nm band for confocal fluorescence and the 1250–1350 nm band for OCT, with working distances of 60 µm in water for confocal imaging and 140 µm in water for OCT. The lens stack has a length of 4.69 mm, clear diameters of 750 µm on the image side and 1180 µm on the object side, and a total optical length of 5.20 mm for confocal and 5.28 mm for OCT. This custom objective enables OCT and confocal images to be comparable. For scanning, sinusoidal electrical signals generated by the computer/DAQ are amplified and delivered through four electrode wires connected to the quadrant electrodes of the piezoelectric tube. The PZT tube vibrates in orthogonal directions, moving the attached fiber cantilever and causing the OCT and confocal beams to scan across the image plane. However, the limitation of this section is that it is primarily designed for forward-viewing imaging.

Animation of 360 Degree Scanning

        To image a larger area of the bladder and provide greater flexibility for cystoscope or ureteral applications, Part 2 was added as a detachable 360° scanning module. This module can be easily attached to or removed from Part 1 during a procedure, allowing the probe to switch between forward-viewing localized imaging and circumferential side-viewing imaging depending on the clinical need. When attached to Part 1, Part 2 uses a computer-controlled micromotor to rotate a 45° reflector, redirecting the co-aligned OCT and confocal beams sideways for circumferential imaging. A plastic sheath is also included to protect the probe during procedures and improve sterile compatibility with endoscopic use. Compared with previous designs, the proposed model combines forward-viewing OCT-confocal co-registration[4] with an optional 360° side-viewing scanning capability in a compact 2.5 mm diameter probe[5]. This modular design allows the system to transition between localized forward scanning and broader circumferential scanning while maintaining the ability to acquire OCT and confocal images from closely matched tissue regions.

Next Steps for Immersion

     Given the very positive feedback from my clinical mentor, this endoscope is planned to be actively developed during my studies in Ithaca and/or potentially continue beyond graduation for collaboration and commercializing. For the remainder of immersion, the next step is to develop a machine learning model that uses OCT and confocal images from the same tissue region to classify tumor areas as benign or malignant. This AI-assisted diagnostic capability could improve real-time diagnosis, guide more targeted biopsies, and potentially reduce unnecessary tissue sampling during surgery or cystoscopy. As a start, preliminary confocal and OCT imaging data will be collected from tissue samples extracted during my clinical mentor’s operations, using available imaging platforms such as histology scanners and Imalux OCT probes before the samples are sent to the pathology team. These data will be used to train the machine learning model and validate its performance in classifying tumors from both OCT and confocal datasets.

Operating and Robotic Surgery Motivation

      On June 17, I observed one robotic surgery case and two operating room cases that highlighted the importance of real-time tissue assessment during urologic surgery. The robotic surgery case involved a tumor within a bladder diverticulum. The patient underwent cystoscopy followed by robotic removal of the diverticular portion of the bladder. Because the diverticulum extended toward the left pelvic area and there was extensive lymph node involvement, the surgical team had to divide a nearby nerve during dissection. This case showed how tumor location and surrounding anatomy can make surgical planning more complex. In the first operating room case, the surgical team removed an entire testicle associated with an enlarged hydrocele and a cystic tumor mass, which was sent to pathology for further evaluation. In the second operating room case, there was a mass near the superior aspect of the right testicle/scrotal region. A portion of the mass was sent for intraoperative frozen section analysis to determine whether it was benign or malignant. The team waited approximately 50 minutes for the pathology result, because the classification would determine whether the testicle needed to be removed. Once the frozen section suggested the mass was benign, the surgeons preserved the testicle and removed only the spermatic cord/scrotal mass. This case stood out because the delay in tissue classification directly affected the extent of surgery and how much tissue could be preserved.

       On June 18, I observed two robotic surgery cases. The first involved removal of the entire prostate for pathological evaluation. The second involved removal of a tumor located beneath or near the right kidney. These cases further emphasized how urologic surgery depends on accurate tumor localization, assessment of nearby anatomy, and determining the appropriate extent of tissue removal. Across both days, these experiences reinforced the clinical need for real-time imaging tools that can reduce uncertainty during surgery. This connected strongly to my immersion project, which focuses on multimodal OCT and confocal fluorescence endoscopy with AI-assisted interpretation. Even with experienced surgeons and intraoperative pathology support, real-time tissue classification can remain a major bottleneck. The 50-minute wait for benign classification in the testicular/scrotal mass case was especially meaningful because the result directly affected whether the testicle would be preserved or removed. By combining structural information from OCT with cellular-level information from confocal imaging, this approach could potentially support faster intraoperative guidance and help surgeons make more confident decisions about when to biopsy, preserve tissue, or remove additional tissue. While it would not replace pathology, it could serve as an additional real-time tool to support tissue-preserving surgical decisions.

References

[1] J. E. Freund, M. Buijs, C. D. Savci-Heijink, D. M. de Bruin, J. J. M. C. H. de la Rosette, T. G. van Leeuwen, and M. P. Laguna, “Optical coherence tomography in urologic oncology: A comprehensive review,” SN Comprehensive Clinical Medicine, vol. 1, pp. 67–84, 2019 

[2] J. Wang, M. Yang, L. Yang, Y. Zhang, J. Yuan, Q. Liu, X. Hou, and L. Fu, “A confocal endoscope for cellular imaging,” Engineering, vol. 1, no. 3, pp. 351–360, 2015 

[3] S. P. Chen and J. C. Liao, “Confocal laser endomicroscopy of bladder and upper tract urothelial carcinoma: A new era of optical diagnosis?” Current Urology Reports, vol. 15, no. 9, Article 437, 2014 

[4] M. H. Lee, J. Im, G. Cho, Y. Chang, and C. Song, “Lissajous multi-modal endomicroscopy with optical coherence tomography and confocal fluorescence microscopy,” Sensors and Actuators A: Physical, vol. 401, Art. no. 117547, 2026 

[5] J. Mavadia, J. Xi, Y. Chen, and X. Li, “An all-fiber-optic endoscopy platform for simultaneous OCT and fluorescence imaging,” Biomedical Optics Express, vol. 3, no. 11, pp. 2851–2859, 2012