Abstract
Band selection is a common approach to reduce the data dimensionality of hyperspectral imagery. It extracts several bands of importance in some sense by taking advantage of high spectral correlation. In medical imaging, narrow-band imaging (NBI) is an imaging technique for endoscopic diagnostic medical tests, where light of specific blue and green wavelengths is used to enhance the detail of certain aspects of the surface of the mucosa. A special filter is electronically activated by a switch in the endoscope leading to the use of ambient light of wavelengths of 415 nm (blue) and 540 nm (green). Because the peak light absorption of hemoglobin occurs at these wavelengths, blood vessels will appear very dark, allowing for their improved visibility and in the improved identification of other surface structures. NBI when compared with the white-light imaging (WLI) have proven to have better precision when combined with computer-aided diagnosis (CAD, Intespec C, Hitspectra Intelligent Technology Co., Kaohsiung, Taiwan) in detecting cancerous images. NBI endoscopes are specialized equipment that may not be widely available in all healthcare settings. By leveraging existing WLI endoscopic systems and developing algorithms to simulate NBI imaging, healthcare facilities can achieve similar di-agnostic capabilities without the need for additional costly equipment. Therefore, in this study, algorithm known as the SAVE (spectrum-aided visual enhancer) has been developed which can simulate NBI from the WLI images through an intelligent band-selective hyperspectral imaging for Olympus endoscope. The results suggested that the SAVE-NBI images had a better precision and F1-score than the WLI images.