1.

INTRODUCTION

Traditional endoscopy technology, primarily utilizing broadband white light illumination, has been extensively used in medical practice for many years [1]. Traditional white light endoscopes are particularly challenged in providing high-contrast images of blood vessels. This is attributed to the use of white light as the illumination source. Biological tissues exhibit varied absorption and scattering properties at different wavelengths, leading to a weak contrast of specific biological molecules in tissues. This decrease in contrast can significantly hinder the early detection of diseases such as cancer.

To overcome this limitation, Narrow Band Imaging (NBI) technology has emerged. NBI improves image contrast in pathological tissue by using specific narrow-band light, of which 415nm (blue light) and 540nm (green light) are commonly used [2]. Besides NBI, multi-spectral imaging technology has also been incorporated into the narrow band imaging system. This technology, by analyzing tissue responses at different wavelengths, provides more information about tissue characteristics, further enhancing the capability to identify and classify lesions [3]. However, the design approach integrating filters tends to be costly.

In this study, we developed a new type of NBI multi-spectral system, utilizing low-cost multi-color LED light sources, along with a specially designed handheld probe and upper computer software. This system was evaluated on oral mucosa and pigmented mole.

2.

MATERIALS AND METHODS

2.1

NBI system

The optical setup of our developed NBI system is shown in Figure 1. The narrow-band imaging system in this paper mainly consists of a multi-color LED ring light source, an optical imaging system, and a light source controller. The system primarily utilizes a LED ring light source integrating various colors of light (395nm, 456nm, 525nm, 607nm, 630nm) to illuminate the sample. The optical imaging system is composed of a pair of achromatic lenses (one AC254-050-A-ML and one AC254-035-A-ML, Thorlabs). The light passes through the achromatic lenses and forms an image on the rear camera (VEN-505-36U3M, IMAVISION). In this paper, the ring light source is connected to a digital light source controller (HLS-V4/8-100w, Honglingshi) for switching different LEDs. Both the light source controller and the camera are connected to a PC via USB. The system control and image acquisition are carried out through an upper computer platform built in VisualStudio2015.

Figure. 1

Schematic diagram of the system. L: Lens, DLC: Digital Light Controller, S: Sample. In the system, the multi-color LED ring light source we designed is mounted at the front end of the probe, as indicated by the red dashed box.

Figure 2 shows the design and fabrication of the NBI handheld probe. Figure 2(a) presents the model designed using SOLIDWORKS, while Figure 2(b) shows the actual model. The front view of the ring light source, as indicated within the red box, is marked with a red arrow.

Figure. 2

(a) Probe model, containing components as shown in the system diagram in Fig 1(b) Actual model, with the multicolor LED ring light source indicated by the red box in the lower right.

2.2

Intensity Normalization

The definition of intensity normalization processing is described below:

where a is the sample image captured under narrow-band light illumination, b is the background image captured under the same conditions as a, c is the image captured using a standard whiteboard (Ocean Insight, USA) as the standard reflector, and d is the background image under the same conditions as c. This crucial preprocessing step effectively eliminates variations caused by ambient light and system response. It not only enhances image contrast but also highlights the target in the image, making subsequent semi-quantitative analyses more accurate and reliable [4].

3.

RESULTS AND DISCUSSION

To evaluate the NBI system, we performed imaging on human oral mucosa. We sequentially switched between five types of narrow-band lights for mucosa imaging. In addition, we used a white light LED to simulate a traditional endoscope for broadband imaging of the mucosa at the same location, which was conducted as the control experiment. The results are shown in Figure 3.

Figure 3.

shows images of the human oral mucosa obtained using conventional white light and narrow-band light. For example, image (b) was captured using ultraviolet light with a central wavelength and FWHM of 395nm and 10nm, respectively. (a): in conventional white light, (b-f): in NBI (Narrow Band Imaging) narrow-band light.

Figure 3 presents intensity-normalized images of oral mucosa captured under white light (a) and narrow-band lights (b-f) illumination. In Figure 3(a), different types and regions of blood vessels are marked with red and green arrows. The white light LED, spanning a wavelength range of 450nm-665nm, encompasses the wavelengths in Figures (c-f). White light can display the fine and coarse capillaries on the mucosal surface, but its contrast is lower compared to Figures 3(c) and (d), and the imaging of fine capillaries is also inferior. This is primarily because white light is a broad-band light composed of multiple wavelengths, which may produce more scattering and reflection during imaging, leading to reduced image quality. In contrast, narrow-band light sources, due to their narrow bandwidth, can minimize the interference from the wavelengths beyond the selected wavelength. However, even with narrow-band light, the imaging of fine capillaries in Figure 3(e) does not show a significant improvement, and in Figure 3(f), only the contours of the coarse capillaries are visible, as indicated by the black arrows. This is attributed to the varying absorption and scattering characteristics of hemoglobin at different wavelengths, and its weaker absorption above 580nm. Figure 3(b) utilizes 395nm ultraviolet light, which, compared to Figure 3(a), offers better contrast. Moreover, due to the stronger absorption characteristics of hemoglobin around 400nm and the shallower penetration depth of 395nm, ultraviolet light can reveal more superficial fine capillaries, as indicated by the blue arrows in Figure 3(b). However, for some deeper coarse capillaries, as indicated by the yellow arrows in Figures 3(b) and (c), the ultraviolet light fails to provide a clear visualization. The results indicate that narrow-band imaging can enhance the contrast of traditional white light imaging, and short-wavelength narrow-band can carry more information about shallow capillaries. With increasing wavelength, deeper coarse capillaries become visible. This system demonstrates great potential in delineating the capillary network of the mucosal surface.

In addition, we performed narrow-band imaging on pigmented mole, with the results shown in Figure 4. The image of the pigmented mole sample is presented in Figure 4(a). Under the illumination of light at 395nm, 525nm, and 630nm, we obtained intensity-normalized images of the pigmented mole, as shown in Figures 4(b-d). From these images, it is clearly visible that the nevus area exhibits different levels of brightness under different wavelengths. This phenomenon is due to the varying absorption and reflection characteristics of the pigmented mole in the ultraviolet and visible light spectrum from 395nm to 630nm. As shown in Figure 4(e), the intensity curves of the pixels marked by dashed lines under different wavelengths indicate that the intensity of the mole is the lowest at 395nm. This observation suggests that the pigmented mole has a strong absorption of 395nm light, resulting in a darker image at this wavelength. This variation in intensity is consistent with the spectral characteristics curve of the pigmented mole. By analyzing images under different bands, we can more accurately determine the nature of the pigmented mole, which is of great significance for the early diagnosis of melanoma. These results indicate that this imaging system has a good potential and prospects in clinical applications in dermatology.

Figure 4.

(a) pigmented mole sample. Images of the pigmented mole under different wavelengths: (b) 395nm, (c) 525nm, (d) 630nm. (e) Intensity curves of the pigmented mole at the dashed line under different wavelengths.

4.

SUMMARY

In conclusion, we have developed a low-cost handheld narrow-band multispectral imaging system based on multi-color LEDs, and evaluated it on oral mucosa and pigmented mole. The experimental results show that our NBI system surpasses traditional white light endoscopy systems in visualizing the surface capillaries of mucosa. Additionally, it possesses multi-spectral capabilities, capable of reflecting the spectral characteristics of pathology, and holds a great potential for clinical disease localization and detection.