Research And Implementation Of Segmentation Algorithm Of Cervical LCT Image

Cervical cancer only next to breast cancer becomes the second most commonmalignancy among the women in the worldwide. However, the process of cervical lesionsneeds a long time, earlier detection and treatment can reduce the mortality rate of cervicalcancer. Therefore, routine examination of cervical cancer plays an important role inprotecting the women’s health.Currently, liquid-based cytology testing technology (LCT) has been the most widelyused screening technology because of quick production and high quality. It will produce a lotof smears during cervical screening. Inefficient manual reading of smears is unable to meetthe heavy work. Coupled with standardized TBS diagnostic criteria and advanced computerimage processing technology, developing computer-aided diagnosis system is a dominanttrend. It is meaningful to use machine to help doctors diagnose instead of manualfilm-reading.According to the characteristics of the cervical image, a new segmentation method isproposed based on the research of predecessors. Firstly, threshold method of the histogram isused to determine the whole cell region. Next, extract the approximate area of the nucleus asa marker, then use marker controlled watershed segmentation method to extract accuratenuclei. For overlapping cells, we only focus on the overlapping portion, which is segmentedby gradient watershed algorithm. Similarity criteria are used as a measure of standards tomerge the results of segmentation. Finally, overlapping cells are separated. The algorithmproposed in this paper divides each cell into two parts, which are cytoplasm and nucleus.Experimental results show that the algorithm can achieve good results.