Research On Micro-vision Detection Technology For Digital Section Of Exfoliated Cells

Cervical cancer is one of the malignant tumors that seriously affect women health.Timely diagnosis of cervical exfoliated cells is an effective way to prevent and treat cervical cancer.In the process of manual screening,there are problems such as low efficiency,influencing results of doctors' qualifications,and geographical limitations of smear pathological information.With the development of computer technology and medical imaging technology,the means of assisted screening by computer-controlled precision medical equipment has gradually matured.Faced with the actual needs of automated detection equipment,a microscopic vision detection system was built for exfoliated cells based on a motion control platform,industrial camera,optical microscope,and computer.Through this system,an efficient and stable micro-vision auto-focusing algorithm was designed,the motion platform was controlled to find the focal plane of the microscope lens,and then industrial camera was made to take high-quality microscopic images of exfoliated cells.Finally,digital image processing technology was used for image stitching,segmentation,and recognition.In terms of autofocus strategy,time interval was used instead of step size for image sharpness comparison,which reduced the time of motor walking-stop.It was applied to the platform to find the focal plane link of the first time.Compared with a single smear acquisition of 300 fields of view.The time spent focusing on traditional mountain climbing was reduced by 4.97%.In micro-image stitching experiments,five feature point detection operators were compared.A high-precision Forstner feature point detection operator was made to extract feature points.For the problem of redundant feature point detection,the derivative part of Robert was replaced by Gaussian smoothed and Gaussian derivative,filtering the image combined with a matching strategy to find feature points in a fixed area,improving the feature point matching rate by 57.64%,and improving timeefficiency by 86.58%.In image segmentation,through experimental comparison,it was found that the difference between the gray value of the background and the cell in the H channel of HSV color space was the largest.Using this channel,the adaptive double threshold segmentation algorithm was made to remove the cell background.Segmentation of nucleus and cytoplasm with R channels in RGB color space,and counting of overlapping nucleus.In the aspect of image recognition,Experimental comparison shows that the stability of SVM classifier was stronger,and the false-negative result was lower than BP neural network.Therefore,SVM was selected to classify the characteristic parameters of cervical exfoliated cells,and the comprehensive recognition rate was 96.55%.Based on the above technical research,a fully automatic microscopic vision detection system for cervical exfoliated cells was established to achieve functions such as automatic smear filling,autofocus,high-resolution microscopic image acquisition,microscopic image stitching,and cell classification and recognition.Freeing manpower from tedious work,improving efficiency,and facilitating the storage and transmission of smear information.It is the future directions of smart medical development.