Studying cytology of the uterine cervix on Win.'98/NT platform using color image processing and mathematical morphology of shapes: a cancerous cell detection and scoring system

Cervical cytopathologists are mainly interested in automatically detecting abnormal cells (dysplasia), CIS (carcinoma in situ) and SCC (squamous cell carcinoma) and also their scoring. This paper presents an interactive algorithm based on color image processing, a combination of morphological algorithms based on shape analysis and finally the connected component analysis for scoring. This system runs on Windows '98/NT PC-400 MHz (Pentium-II) platform, using Visual C++ (6.0) environment in three different architectures: single document interface, multiple document interface and dialog based applications. The system takes around 1.5 seconds per image of 512/spl times/484 square pixels using a high speed threading architecture. The system's accuracy is 95%.