Automated screening of Pap smear slides is challenging due the high processing and data transfer requirements placed on the processing engine. These requirements can be significantly reduced by processing the images at the image plane of the camera, and reading out only the relevant data, which results in lower cost and higher performance systems. Image sensors with smarts or computational capability at each pixel can be advantageously used in the application to build extremely compact and low-cost automated screening systems. Morphological filtering algorithms have been shown to be effective at detecting object size and shape, which are distinguishing features in diagnostic microscopy. The goal of this research is to design, simulate and fabricate a CMOS chip with morphological filtering circuits at each pixel, which will allow detecting suspicious cells in a Pap Smear at more than 1000 frames/second. In Phase I, Bosonics will (1) determine the desired imager's technical capabilities for a compact microscope mountable smart camera (2) design candidate morphological filtering architectures and circuits; (3) simulate the morphological algorithms with realistic circuits models; and (4) design and fabricate a 5x5 imager demonstration chip in CMOS.
The chips produced under this program will lower the cost of machine vision systems by providing an integrated detector/processing function as well as increasing performance by reducing the output bandwidth requirements. The detector arrays developed under this program will have wide application in automatic target recognition, machine vision for automated manufacturing, medical diagnostic imaging, and remote sensing and surveillance.
