A multispectral image enhancement approach to visualize tissue structures
Winner - Imaging InformaticsPinky A. Bautista ; Tokyo Institute of Technology;
Content:
Visualization of tissue structures necessitates the application of chemical stain. Popular to routine staining is the H&E stain which generally differentiate acidophilic from basophilic structures. Initial investigations on the application of multispectral imaging to pathology has shown that with such imaging technology it is possible to visualize tissue structures that are ordinarily emphasized with special stains from an H&E stained image.
Technology:
In contrast to the conventional RGB imaging, multispectral imaging uses more than 3 spectral filters to capture images such that salient information which can be useful for image analysis is contained in the images. An enhancement technique for stained multispectral images is introduced whereby tissue structures which are not clearly differentiated in H&E stained images can be emphasized
Design:
PC vectors are derived from the transmittance spectra of tissue components which are not desired to be enhanced. Using only m-PC vectors the transmittance spectra of the multispectral image pixels were reconstructed wherein greatest error occurs for those tissue components whose transmittance spectra samples were not included in the set from which the PC vectors were derived. The spectral errors are then used to modulate the original transmittance spectra of the image pixels to produce varying color impression on tissue structures. A change in color from their original H&E stained impression can be specifically observed on tissue structures which acquired high spectral errors.
Results:
Results In our experiment multispectral images captured from the slides of kidney, liver and heart tissue samples were used. The resulting enhanced images were compared to the real MT-stained images with respect to the delineation of collagen fiber from the rest of the tissue components. Initial results show the possibility of differentiating collagen fiber by applying the current enhancement method.
Conclusion:
A multispectral enhancement framework has been introduced whereby the utility of multispectral imaging (MSI) to visualize tissue structures with subtle spectral difference has been shown. In our future work we will consider the enhancement of other tissue structures which are emphasized by stains other than Massons trichrome and also the practical implementation of the current multispectral enhancement procedure.
