Cancer Diagnosis through Multispectral Imaging
Early diagnosis of cancer is the key for the management and treatment of malignant lesions. Therefore, cancer diagnosis at the cell level is desired and the optical window ranging from visible to near-infrared presents significant opportunities for achieving this goal. In multispectral imaging, the co-existence of the spatial and spectral dependence of the image data demands synergy of data analysis and multivariate provide an array of tools for studying this type o imaging problems.
We have started a research project for diagnosing dyplasia and malignant melanoma in pigmented skin lesions on the basis of our previous study of skin optics. A preliminary polarimetric multispectral imaging system has been constructed to take multispectral images with a thermo-electrically cooled CCD camera. The images shown below are polarization-subtracted and normalized images from a benign nevus on a volunteer’s arm with monochromatic illumination with the wavelength indicated. A multivariate image analysis of these images indicated that the different region of the skin can be separated by the different concentration of “chemical” components obtained from the analysis. Further improvements in the multispectral imaging system, optical modeling and multivariate analysis are currently in progress and will be reported soon.