The Research On Some Key Techniques On Screening Of Liquid Based Cervical Cytology

The method of cervical cytology is one of the most important methods to examine the cervical precancerous lesions. And cell morphology measure is the key technology, and the accuracy and stability of measure results has great influence on the screening results.Based on the differences between normal and lesion cells in cell morphology, ac cording to TBS diagnostic system, this article introduces four characteristic measurem ents and applies them to the early screening system of cervical cancer, ensuring the judgment accuracy of screening system.(1) Due to settling TLT and image collecting method, the microscopic image of cervical cells will exist phenomena of cytoadherence and uneven illumination.Conventional segmentation method cannot segment cervical cells directly and accurately. The method in the article is taking advantage of the sensibility of cell pap staining to red and blue color, combining color space conversion with OSTU segmentation to realize the integral segmentation of cell and the separation of cell nucleus from cytoplasm. According to the differences between adherent cell nucleus and single cell morphology, use morphological method to segment adherent cellular image into the single cell.(2) The core of cell morphology detection is to adopt which accurate and effective characteristic measure method to assess whether lesions occur in the cell or not in the liquid-based cervical cell morphology detection process, the core of cell morphology detection is to adopt which accurate and effective characteristic measure method to assess whether lesions occur in the cell or not. Building four characteristic Measure methods, the cell nucleus eccentricity, the round degree of cell nucleus,nucleus nucleo-cytoplasmic ratio and nucleus optional density, of which includes the improvement of target eccentricity and roundness description method, allowing description method to distinguish normal cells from abnormal cells better. These four methods have been proved to have advantages of fine repeatability and high screening accuracy rate.(3) The article applies four characteristic measure methods to the cervical cancer automatic screening system, using metric results to assess whether lesions occur in the cell or not. If finds suspicious lesion cells, use image annotation to locate suspicious cells, letting pathologists analyze and confirm.The article uses methods like image preprocessing, image segmentation,characteristic extraction to treat the microscopic image of cervical cells, thus builds four measure methods based on characteristic. At last, combine the measurement results and experience of pathologists, realizing better detection effect of cervical cell and reaching planned research purposes.