Detection of Filamentous Microorganisms in Fluorescence Microscopy Images.

We present a robust, precise image binarization technique for automatically detecting filamentous microorganisms from digital fluorescence microscopy scans, with application to finding the pseudohyphae that are fungal pathogens responsible for Candida vaginitis. This method employs a hybrid constant false positive rate processor that integrates cell average and order statistic detectors, with linear windows at multiple orientation angles. The hypothesis test rule incorporates elongation enhancement and region of interest masking. Our approach achieves the adaptivity to local noise and all possible object orientations. The designed processor is evaluated theoretically and experimentally using clinical images. Successful detection results are demonstrated.