Design of a modified endoscope illuminator for spectral imaging of colorectal tissues

Abstract

The gold standard for locating colonic polyps is a white light endoscope in a colonoscopy, however, polyps smaller than 5 mm can be easily missed. Modified procedures such as narrow band imaging have shown only marginal increases in detection rates. Spectral imaging is a potential solution to improve the sensitivity and specificity of colonoscopies by providing the ability to distinguish molecular fluorescence differences in tissues. The goal of this work is to implement a spectral endoscopic light source to acquire spectral image data of colorectal tissues. A beta-version endoscope light source was developed, by retrofitting a white light endoscope light source (Olympus, CLK-4) with 16 narrow band LEDs. This redesigned, beta-prototype uses high-power LEDs with a minimum output of 500 mW to provide sufficient spectral output (0.5 mW) through the endoscope. A mounting apparatus was designed to provide sufficient heat dissipation. Here, we report recent results of our tests to characterize the intensity output through the light source and endoscope to determine the flat spectral output for imaging and intensity losses through the endoscope. We also report preliminary spectral imaging data from transverse pig colon that demonstrates the ability to result in working practical spectral data. Preliminary results of this revised prototype spectral endoscope system demonstrate that there is sufficient power to allow the imaging process to continue and potentially determine spectral differences in cancerous and normal tissue from imaging ex vivo pairs. Future work will focus on building a spectral library for the colorectal region and refining the user interface the system for in vivo use.

1. INTRODUCTION

The gold standard for detection of colorectal polyps is white light endoscopy in a colonoscopy procedure. Modified versions of the white light endoscope have been developed in order to improve specificity and sensitivity such as narrow-band imaging. While narrow band and autofluorescence imaging endoscopy provided improvements in detection sensitivity and specificity in small scale trials, when these techniques were evaluated on larger multi-center scales, the improvements in detection ability were marginal or not present. ^1,2

A potential alternative technology for improving colonoscopy detection ability is by integrating spectral imaging to an endoscopic procedure. The availability of spectral image data provide the ability to characterize different tissue based on molecular reflectance and/or fluorescence in turn improving the sensitivity and specificity in endoscopic procedures.³ We have previously shown that spectral imaging can be used to discriminate lung autofluorescence from exogenous fluorescence markers. More recently, we have shown that a new spectral imaging technique that scans the fluorescence excitation spectrum, has diagnostic potential for discriminating cancerous and normal colorectal tissues.^4–7

The overall goal of this work is to design a LED based endoscopic light source that collects spectral images and data during a colonoscopy using a real time video feed. The prominent goal is to improve on the beta-prototype’s spectral output and to collect initial spectral imaging data for analysis. ⁸

2. METHODS

2.1. System alterations for LED enhancement

A previous optimization analysis of the beta prototype proved the light transmission through the system is a limiting factor for real-time operation.⁸ Based on this prior analysis, we determined that a target illumination intensity, as measured at the end of the endoscope, of 10–20 mW, would be sufficient for real-time spectral imaging. The beta system currently measures approximately 20 μW output through the lightpipe and endoscope. To improve the output we modified the prior system to use high-powered surface mount LEDs with heat sinks were purchased. A new LED-mounting circuit board was designed to allow effective coupling of the LED with the solid light guide illumination delivery system (Figure 1). Printed circuit boards were designed using Pad2Pad software.

Figure 1:

Comparison of LED board designs: original layout with wire nodes (top/left) and Pad2Pad blueprint (top/right) and new layout with high-powered surface mount LEDs (bottom/left) and the corresponding Pad2Pad blueprint (bottom/right)

Once installed the new LEDs were tested for transmission loss through the system. Spectra diametric measurements were performed using a fiber-coupled spectrometer (QE65000, Ocean Optics, Dunedin, FL) equipped with integrating sphere. All measurements were made with respect to calibration by NIST-traceable reference lamp (LS-1-CAL-INT, Ocean Optics) using the parameters: 100 ms, 100 scans to average and 8 boxcar width. Measurements were made of the raw LED output, the lightpipe and the endoscope (CF-P20S, Olympus). The output of the lightpipe and endoscope versus the raw LED output provided the transmission loss results.

2.2. Video quality analysis

The resulting intensity output of the system was sufficient for a first pass video-rate test of the endoscope system. Camera acquisition and spectral illumination were synchronized using NIS Elements software (Nikon Instruments). Several parameters were adjusted to optimize image quality and acquisition speed: binning, shutter speed and gain also, were LED intensity and wavelength selection also optimized. Using these parameters, we were able to optimize/compromise a group of settings allowing near video-rate image acquisition. Two different settings were determined given the output is still lower than the current goal, high framerate with moderate quality and low framerate with high quality.

2.3. Ex vivo beta testing

Resected transverse and descending colon were obtained post-mortem form unrelated ongoing studies at the University of South Alabama College of Medicine Vivarium. All manipulations in the unrelated studies were approved by the University of South Alabama Institutional Animal Care and Use Committee. The colon section was flushed and tied off at one end. Then various imaging tests using the high framerate settings were shot and still images using the high picture quality settings were performed and regions of vascular interest.

3. RESULTS

3.1. High-powered LED integration

High powered LEDs improved the output intensity compared to the prior alpha prototype (Figure 2). However, there is still a 95% transmission loss through the light pipe and a 99% transmission loss through the lightpipe and endoscope. Hence, a key goal for further optimization of the system is to minimize the transmission loss through the lightpipe.

Figure 2:

Original LEDs over the reference voltage range for visible (A) and IR regions (B), new high-powered LEDs over the generic reference voltage range for visible (C) and IR regions (D). The reference voltage is the controller for the illumination intensity. It works in reverse as a safety feature, the lower the voltage the higher the intensity

3.2. Framerate optimization

By compromising between image signal–to-noise ratio (SNR), number of wavelength bands acquired, and resolution, we were able to achieve a spectral imaging frame rate of ~8.5 fps. Using alternative settings for higher SNR images (in a still shot), we were able to achieve a frame rate of 2 fps. The continuing goal is to improve the signal strength while obtaining it at 20–25 fps.

3.3. Tissue sample reflectance data

Images acquired from ex vivo pig colon demonstrated the ability to visualize general mucosal structure, the reflectance of the remaining fluid and darker clusters of tissue (Figure 3).

Figure 3:

Overlayed, false-colored images of ex vivo pig colon region using two sets of wavelengths: 420–670 nm (A,C,E) and 620–940 nm (B,D,F), demonstrating visuals of mucosal structures and reflectance

The image data shown in Figure 3 portray a false-colored representation of all spectral channels merged into a single RGB composite image. To better visualize spectral image data, we also visualized image data by the individual wavelength channels (Figure 4).

Figure 4:

Images of each wavelength 420 nm (A), 470 nm (B), 525 nm (C), 590 nm (D), 620 nm (E), 670 nm (F), 680 nm (G), 750 nm (H), 810 nm (I), 850 nm (J), 940 nm (K) and the combined image (L)

Here, we show that a novel excitation scanning spectral imaging system allowed endoscope image acquisition with a frame rate of 2 ~ 9 fps, and produced images of sufficient SNR for visualizing colonic mucosa.

4. FUTURE WORK

The goal to reach 10–20 mW in the light source is directly correlated to the end goal of reaching a spectral imaging speed that will allow video-rate acquisition. The limiting factor of the system is the solid light guide (lightpipe) and the key step next step is to optimize the optical output by minimizing transmission loss. The options under investigation are a liquid light guide or a change in geometry and/or materials of construction. There is a light array using a liquid light guide, in another project within the lab, which can be incorporated into the endoscope if initial testing proves noteworthy. Changing geometry will require computer models (Inventor, SolidWorks) to change angles of branches and connection points and test in light tracing software (TracePro). Multiple iterations will be tested to determine if it is worth the investment. Afterwards, the next step is testing the video-rate imaging acquisition for reflectance and excitation scanning.