A Novel Hyperspectral Endoscopic Imaging Method Based on Broad- and Overlapped-Band Filters

Abstract

In medical domain, hyperspectral imaging (HSI) [1] offers a hybrid modality of optical diagnostics that obtains spectral information and renders the information in image form, thus it has great potential for noninvasive disease diagnosis and surgical guidance [2]. Masood [3] explored a series of research problems for classification of hyperspectral (HS) colon biopsy images, and concluded that HSI has enough discriminatory power to distinguish normal and malignant biopsy tissues. However, HS illumination is always with low intensity and signal-to-noise (SNR) ratio is low for HS images. The reason is that sensors of endoscopic systems, such as charged-couple-device (CCD), work by converting photons into electrons which are then stored in each pixel. The number of electrons stored in each pixel well is proportional to the number of photons that struck that pixel, although in actual pixel well, electrons are not only generated when receiving photons, but also due to physical processes within CCD itself, such as thermal noise which generates additional electrons dependent to temperature [4,5]. At the same time, dispersive device such as monochromators (such as prism and grating), and optical filters (including interference filters and tunable filters) are commonly used for HS imaging. Using these traditional filtering approaches, only a low fractions of photons are transmitted into the sensor. Moreover, due to the requirements of safety and keeping the working time per frame as short as possible for in-vivo and real-time application in medical domain, common methods which could improve SNR, such as using cooled sensor or high exposure time, are not applicable for HS endoscopy systems. To overcome this drawback, we proposed this novel HS imaging method based on broad- and overlapped-band filters [6].