Multiple wavelength microscopy image analysis and its applications in the study of prostate specific acid phosphatase in prostate cancer

A new color segmentation algorithm, multiple wavelength decomposition, was developed to accurately decompose multiple stain components in biomedical specimens. By using images captured at N wavelengths, N components with different colors can be separated. Realizations of the algorithm for two and three wavelengths have been successfully applied to actual pathological specimens. We have demonstrated that the developed methods are accurate. The selection of optimum wavelengths for decomposing the two and three stain components used in the study is discussed.; In order to implement this multiple wavelength algorithm and improve the efficiency of the multiple wavelength acquisition in a conventional microscope, an automatic microscope imaging system was developed. This system was configured in a conventional transmitted light microscope and a PC 486 computer. The stage, light power, and filter wheel were controlled by the PC through an RS232 port. The image was digitized by a monochromatic high resolution digital camera and an image frame board with resolution of 1024 x 1280 x 10bit. In order to evaluate a large group of slides in a batch mode, the PC based imaging system was connected to a SUN computer network through ETHERNET.; The developed image analysis system also facilitates the multiple tasks of image analysis, which is very important in quantitative histology and cytology. A batch mode image process of a large group of slides was also developed. The sequential steps involved in this batch mode process were image acquisition performed on the PC based imaging system, image saving to the SUN system through ETHERNET, and image evaluation performed on the SUN workstation.; In order to test the hypothesis that the level of immunohistological Prostate Specific Acid Phosphatase (PSAP) staining correlates with survival, we applied the dual wavelength algorithm to analyze doubly labeled specimens of prostate cancer from three RTOG (Radiation Therapy of Oncology Group) protocols. In these study specimens, the PSAP was labeled with the peroxidase-antiperoxidase complex technique using diaminobenzidine as a substrate (PAP-DAB) while the nuclei were stained with hematoxylin. These two stain components, which could not be separated adequately by the conventional thresholding method, were completely separated by the dual wavelength algorithm. Therefore, the measure of one stain was independent of the other. In order to limit the variations resulting from section thickness and stain conditions, the degree of PAP-DAB was evaluated by comparing measures obtained from malignant gland cells with those of control gland cells in the same slide. Our preliminary results have shown that the cytoplasm intensity and extent of PAP-DAB correlate with survival. These measures of PAP-DAB are independent of other grading systems in prostate cancer: Gleason, M. D. Anderson, and clinical stage.