We report an improved method to visualize lipid distribution in axial and lateral direction within arterial vessel walls by
spectroscopic spectral-domain Optical Coherence Tomography (OCT) at 1.7μm wavelength for identification of lipidrich
plaque that is suspected to cause coronary events. In our previous method, an extended InGaAs-based line camera
detects an OCT interferometric spectrum from 1607 to 1766 nm, which is then divided into twenty subbands, and A-scan
OCT profile is calculated for each subband, resulting in a tomographic spectrum. This tomographic spectrum is
decomposed into lipid spectrum having an attenuation peak at 1730 nm and non-lipid spectrum independent of
wavelength, and the weight of each spectrum, that is, lipid and non-lipid score is calculated. In this paper, we present an
improved algorithm, in which we have combined the lipid score and the non-lipid score to derive a corrected lipid score.
We have found that the corrected lipid score is better than the raw lipid score in that the former is more robust against
false positive occurring due to abrupt change in reflectivity at vessel surface. In addition, we have optimized spatial
smoothing filter and reduced false positive and false negative due to detection noise and speckle. We have verified this
improved algorithm by the use of measuring data of normal porcine coronary artery and lard as a model of lipid-rich
plaque and confirmed that both the sensitivity and the specificity of lard are 92%.