Background
Multi-Spectral imaging systems are being developed to reproduce accurate, real world color in digital images for applications such as digital museums and tele-dermatology. By sampling light from an object through multiple, well-controlled spectral bands, multi-spectral systems can produce exact object color and can maintain that color image capture (camera) to display (monitor). Although importance of maintaining exact color is obvious for an art museum, the issue becomes much more complex in microscopic pathology where the "real" color of a tissue section is highly dependent on the interaction between components of the tissue, illumination and staining. Because of this interaction may be possible to extract additional information about tissue components by imaging tissue with sensitive multispectral cameras.
Aim
The long term goal of this project is to determine whether multispectral technology could be used to extract information from tissue that is not available with traditional imaging methods. As part of this process, used multispectral imaging to improve image quality, and used multispectral technology in a prototype Automated Scoring System and Digital Stain Exchange System.
Methods
Initial work was performed on a collection of multispectral imaging systems (including a 16 band multispectral camera) provided by the Natural Vision Project of the Japanese Ministry of Telecommunication and Post Office. To determine the effect of different microscopes, cameras and stains on microscopic images, a database of the spectral properties of a variety of microscope optics, cameras and histologic dyes was created. This database and examination specific slides, were used to build a control color chart appropriate for pathology. The color chart was then used in subsequent work to correct color from traditional digital cameras. On the basis of this preliminary work, software was developed to 1) automatically correct for brightness, filter, and light source, 2) determine percent fibrosis in liver biopsies, and 3) analyze the differential color spectra between Tri-Chrome and H&E staining. This differential spectra is now being used in an attempt to develop a digital staining technique in which H&E stains are digitally transformed to a virtual Tri-Chrome for quantitation of fibrosis in liver biopsies and other specimens.
Results
Using a Multi-Spectral Imaging System, we could correct color problems caused by inappropriate brightness, filters and light sources, automatically determine the percent fibrosis in liver biopsies, and determine the spectral differences in tissue stained with H&E and Tri-Chrome. Attempts to digitally transform an H&E image to a Tri-Chrome are ongoing and will be reported in Banff Conference.
Conclusion
Initial indications are that multispectral technology can be very useful in pathology. We can achieve consistency of image quality over multiple imaging and staining systems, control for image capture errors, automate a number of pathology tasks and possibly develop digital staining. In the future Multi-spectral technologies may be combined with morphologic analysis to create automated scoring systems for each organ and disease.