The recent development of ratation and translation invariant Synthetic Discriminant Filters (SDFs) vastly increases the image processing ability of optical correlators. Distinguishing features are detectable and recognizable for virtually any object regardless of image orientation or magnification. Extension to medical imagery is a natural application that can vastly increase diagnostic capabilities of a wide range of medical instruments. Numerous advantages can accrue, including: increased sensitivity in feature detection, enormous processing speed, automated image comparisons against much larger databases, and multidimensional image analysis in which detection of multiple features provide greater probability for detection of disease. This project begins the medical exploitation of the technology. Increased processing power will eventually benefit many medical instruments. This initial investigation is limited to a feasibility demonstration for screening of cervical smears. Features amenable to SDF development will be identified. Example SDF filters will be developed demonstrating the capability to distinguish among normal, atypical (inflamed) and dyskaryotic squamous cells. Retroactive data is used, and after screening, images will be used to demonstrate automatic recognition of distinguishing features among cell types relevant to automated cervical screening.
Automatic screening of cervical smears. An optical correlator system and a bank of SDF filters can eventually replace the processing computer and image processing software in existing automated cervical screening systems with a resultant decrease in system cost, improvement in sensitivity and substantial increase in processing speed.
