Multimodal endoscopy for colorectal cancer detection by optical coherence tomography and near-infrared fluorescence imaging

: While colonoscopy is the gold standard for diagnosis and classification of colorectal cancer (CRC), its sensitivity and specificity are operator-dependent and are especially poor for small and flat lesions. Contemporary imaging modalities, such as optical coherence tomography (OCT) and near-infrared (NIR) fluorescence, have been investigated to visualize microvasculature and morphological changes for detecting early stage CRC in the gastrointestinal (GI) tract. In our study, we developed a multimodal endoscopic system with simultaneous co-registered OCT and NIR fluorescence imaging. By introducing a contrast agent into the vascular network, NIR fluorescence is able to highlight the cancer-suspected area based on significant change of tumor vascular density and morphology caused by angiogenesis. With the addition of co-registered OCT images to reveal subsurface tissue layer architecture, the suspected regions can be further investigated by the altered light scattering resulting from the morphological abnormality. Using this multimodal imaging system, an in vivo animal study was performed using a F344-Apc Pirc Uwm rat, in which the layered architecture and microvasculature of the colorectal wall at different time points were demonstrated. The co-registered OCT and NIR fluorescence images allowed the identification and differentiation of normal colon, hyperplastic polyp, adenomatous polyp, and adenocarcinoma. This multimodal imaging strategy using a single imaging probe has demonstrated the enhanced capability of identification and classification of CRC compared to using any of these technologies alone, thus has the potential to provide a new clinical tool to advance gastroenterology practice.


Introduction
Colorectal cancer (CRC) is the third most common type of cancer, consisting of about 10% of all cancer cases globally [1]. In 2012 alone, approximately 1.4 million new cases of, and almost 0.7 million deaths from, CRC were reported [2]. Colonoscopy (i.e., white light endoscopy) is the gold standard for CRC diagnostics as it provides visualization of abnormal tissue growth on the mucous membrane in the colon or rectum, known as colorectal polyps. In addition to CRC screening, physicians also utilize colonoscopy to excise small polyps, and biopsy larger polyps or tumors for further diagnosis, in a minimally invasive manner. However, several disadvantages of colonoscopy exist. While being the standard imaging technique, colonoscopy provides only surface morphology of the rectal wall and cannot resolve the abnormal layer architecture and subsurface microvasculature which are highly associated with CRC. Hence, cancer staging often requires biopsy, which not only is more invasive and time-consuming but also has diminishing accuracy limited by the sample number (four and six biopsy specimens yield accuracies of 68% and 78%, respectively) [3,4]. The size of the polyp also affects the accuracy of CRC diagnostics. The miss rate increases significantly in smaller sized polyps which is estimated to be as high as 26% for small polyps (< 5 mm) [5]. An imaging system that provides a high sensitivity and specificity for differentiating all kinds of polyps is therefore necessary.
To address the limitations of conventional colonoscopy, many endoscopic imaging modalities, such as endoscopic ultrasound (US), optical coherence tomography (OCT), Doppler OCT, photoacoustic (PA), and near-infrared (NIR) fluorescence imaging [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20], have been applied in GI tract to visualize the layered architecture or subsurface vascular network for better CRC staging and management. Each modality has its own strengths and limitations. To improve the diagnostics accuracy, many groups focus on developing multimodality imaging systems (such as combined OCT/US, PA/US, and OCT/NIR fluorescence). Among them, combining OCT and NIR fluorescence allows cross-sectional visualization of the tissue morphology and vasculature with high spatial resolution and sensitivity, providing a powerful tool to monitor the hallmarks of CRC (i.e., morphological abnormality and angiogenesis). Furthermore, integrated OCT/NIR fluorescence imaging decreases the procedure cost and time as only one imaging probe is used to acquire both OCT and NIR fluorescence data in one session. Several multimodal OCT/NIR fluorescence systems [9,18,[21][22][23][24] have been reported, demonstrating their capability of visualizing tissue morphology and molecular composition simultaneously. However, a successful clinical adaptation for CRC diagnostic imaging requires a combination of miniaturized endoscopic packing, FDA approved contrast agent, high-speed and high-resolution 3-dimensional (3D) imaging, and animal model validation. Currently, none of the reported systems has achieved all of the above prerequisites.
In this study, we combined OCT and NIR fluorescence into one imaging system and packaged the imaging probe into a miniature endoscope. OCT and NIR fluorescence imaging were performed simultaneously with high speed and high resolution. Using an FDA approved contrast agent ICG, we validated the multimodal system with a diseased rat model by demonstrating the capability to visualize the change in vascular density and morphology through NIR fluorescence and study the tissue layer architecture via co-registered OCT images. This allowed for real-time identification of suspect lesions with microscopic detail in vivo, leading to better determination of GI tract disorders. Through the animal study, our preliminary results from a rat model demonstrated the potential of the multimodality imaging system for detection of CRC and differentiation of cancer stages.

Contrast agent
Indocyanine green (ICG) is a fluorescent dye used in medical imaging as an indicator substance and was approved by the FDA in 1959. ICG binds 98% to plasma proteins and accumulates in the tumor region more than normal tissue after intravenous injection due to the "enhanced permeability and retention" effect caused by angiogenesis [25,26]. For this reason, ICG is widely used for angiogram and tumor border delineation [27,28]. Because of tumor vasculature is more dense and highly tortuous, the morphology and density of tumor vasculature can be used as biomarker for tumor detection. In this study, ICG was used to perform angiogram and identify the suspected tumor region based on the altered density and morphology of vasculature. ICG has a short half-life (~3 minutes) and is cleared from circulation exclusively by the liver to bile juice quickly. Therefore, the imaging will be performed right after the intravenous injection and finished in 3 minutes.

Animal model
Male rats from the F344-Apc Pirc Uwm strain were used for this study. The rat model carries a knockout allele in the gatekeeper gene adenomatous polyposis coli (Apc) and has been designated the polyposis in the rat colon (Pirc) kindred, which can develop spontaneous intestinal lesions starting at the 45 days of age which then eventually expand to the colon and rectum [29]. This model has been widely used in longitudinal analysis of colorectal tumor biology, chemoprevention, microbiome effects, chemotherapy, and early detection.

Multimodal OCT and NIR fluorescence imaging system
Figure 1(a) shows the schematic of the integrated multimodal OCT and NIR fluorescence imaging system. A wavelength division multiplexer (WDM) and a double clad fiber (DCF) are utilized to fully integrate OCT and NIR fluorescence together. Figure 1(b) shows a micromotor-based imaging probe, which is able to perform stable and fast 3D scanning. The system setup and probe design have been described previously [30]. In comparison to our previous design, the outer diameter of the imaging probe has been reduced to 1.6 mm to readily access the colon. In addition, an upgraded micromotor and correction algorithm were applied to provide a stable cross-sectional OCT/NIR fluorescence images. All the software was written entirely in C + + utilizing the graphics processing unit, which allows fast data processing and real-time display during the rat imaging.

Imaging protocol
The male rats with a certified report were purchased from Rat Resource & Research Center. The rats were kept on a normal diet upon arrival and were imaged at weeks 1 (after 1 week of acclimatization), 4, and 8 to track the development of the colorectal tumor. To anesthetize the rat for the imaging procedure, the rat was first placed in a hermetically sealed plexiglass chamber for general anesthesia induction and was then removed from the chamber for an intraperitoneal injection of a ketamine-xylazine mixture (87 mg/kg and 10mg/kg, respectively). After the rat was anesthetized, an enema was performed to remove fecal matter from the rectum. Then, a plastic tube was inserted to inflate the rectum. After gently warming the rat tail, ICG (1.5mg/kg) was then administrated intravenously through either of the lateral tail veins, followed by immediate imaging in consideration of the fast-hepatic uptake and the short half-life less than 3 m were acquired hematoxylin a were reviewe University of

Dual mo
Using our mu NIR fluoresce intensity is d calibration m fluorescence fluorescence extracted from

Histologi
Histology of t modality syst histopatholog adenocarcinom Healthy r structures and 6(b)] which i fluorescence brightness cor Adenomat thickening in intensity in th observed. In secretion of th is also increas which the tub

Conclusion
While OCT and NIR fluorescence imaging each can individually provide clinically valuable information to supplement conventional white-light colonoscopy, combining these two imaging modalities further enhances the practicality of this dual-modality system. Colonoscopy allows viewing of the colorectal surface morphology and is intuitive for the surgeon, but because of lower sensitivity and specificity it cannot access information from subsurface tissue layers. With the multimodal imaging system, NIR fluorescence imaging can be used to identify the suspect lesions rapidly, and OCT endoscopy helps visualize the microanatomy of the subsurface layer structures with microscopic detail for further diagnosis, allowing better staging and diagnosis of GI tract disorders.
With histological confirmation, our results demonstrated the capability of the multimodal imaging system to identify and differentiate healthy tissue, hyperplastic polyps, adenomatous polyps, and adenocarcinoma. In NIR fluorescence images, the contrast is from vascular network. The vascular density gradually increases, and vascular morphology becomes more tortuous with disease progression. In OCT images, tissue layered architecture alters as the disease progresses. We observed that different types of polyps exhibit unique patterns, which can be used for differentiating polyps. In healthy colorectal wall, well-defined layers can be visualized: the intensity of each layer is uniform, and boundaries are well-demarcated. In hyperplastic polyps, mucosal thickening is observed, and the OCT signal intensity of the thickened mucosa is similar to the neighboring healthy tissue. In adenomatous polyps, mucosal thickening is also visualized, but the OCT signal intensity is reduced in contrast with the surrounding normal tissue due to lowered light scattering. For the adenocarcinoma, boundaries are blurred and intensity of OCT images are non-uniform (e.g., the uneven dark areas in Fig. 9).
The endoscopic multimodal imaging is minimally invasive and can be performed in real time. With its small size (probe outer diameter < 3 mm), the multimodal imaging probe can access the colon through the accessory channel of a commercial endoscope. This approach can be easily integrated into clinical practice as surgeons can still rely on intuitive colonoscopy guidance but also gain additional subsurface information provided by the multimodal imaging system. Such integration can help differentiate polyps observed by colonoscopy to provide a better diagnostic yield.
Future studies will focus on determining sensitivity and specificity of the proposed multimodal imaging system with respect to colorectal cancer. A larger sample size will be proposed for quantitative analysis. Additionally, it is also possible to utilize autofluorescence rather than NIR fluorescence, which can eliminate the need of a contrast agent, further reducing the invasiveness of our multimodal approach.