| Biological studies of intercellular communication are more effective if observations are made in the same domain as the cells themselves, i.e. in three dimensions. The confocal microscope, along with techniques such as fluorescence resonance energy transfer and blob analysis, aid in this aspect. An approach to automated real-time analysis of biological cell images using a confocal microscope is proposed with a paradigm of hardware-based architecture. For real-time applications, field programmable gate arrays, along with high level programming languages such as Handel-C, make a good alternative to expensive application-specific hardware and powerful workstations. Basic segmentation techniques, such as thresholding, were modified to fit the hardware constraints. Prior to cell segmentation, the multiband images are corrected for nonlinear tilt distortion and warp-induced distortions, following which they are registered so as to facilitate proper analysis. Blob analysis is then performed, through which a feature vector is constructed for each cell. The feature vector contains various cell features such as area, major-axis length, minor-axis length, orientation, first-order moments, and second-order moments. An architecture is proposed that will enable easy portability to FPGA. A detailed analysis of timing is performed, through which an approximate clock rate of 22 clocks/pixel is estimated to be achieved after the actual hardware coding. |
