Rapid Acquisition And Quality Evaluation Methods For Cytopathological Cell Image

With the advancement and development of science and technology,traditional diagnostic methods that rely on pathologists under the microscope face a variety of problems,such as being time-consuming and labour-intensive and more subjectively influenced by doctors,which has led to the large-scale development of automated pathological diagnosis.It is vital to obtain clear images of the sample with a good evaluation process by means of a rapid pathology image acquisition device.There are still problems with the use of rapid pathology image acquisition equipment in automated pathology diagnosis.Firstly,current pathology rapid acquisition equipment captures images at a very high resolution,which leads to inefficient calculation of the evaluation function of the image clarity function and affects the speed of rapid acquisition.Secondly,a large number of samples are scanned every day and the substandard quality of the cytopathology images produced by the scans can seriously affect the efficiency of automated pathology diagnosis and even lead to misdiagnosis and missed diagnoses.Although many methods exist for evaluating image quality,they do not take into account the criteria used by cytopathologists to evaluate sample images and are therefore not applicable to the field of automated pathology diagnosis.In order to address these issues,this paper investigates both improving the efficiency of the rapid pathology image acquisition equipment and evaluating the quality of the scanned cytopathology images.The details of the study are as follows.1.A focus window selection method based on grey-scale projection is proposed.The method uses the grey-scale projection method and the adaptive mean interval finding algorithm to determine the focusing window.This method can effectively reduce the computation of sharpness evaluation function,and can improve the effect of sharpness evaluation function to enhance the focusing efficiency.In terms of focusing method,a focus prediction method based on thin-slab sample interpolation is proposed.This method predicts the focus plane by focusing on a few fields of view,which effectively reduces the number of focuses and greatly improves the speed of pathology image acquisition.2.A sharpness determination method based on the residual network is proposed.In order to solve the problem that the original pooling strategy of the residual network leads to the loss of clear and blurred information at the edge of the image,inspired by the traditional sharpness function,the maximum-minimum pooling is added to the network to enhance the network’s ability to perceive clear and blurred information of the image;at the same time,in order to further improve the classification accuracy of the network,the channel attention mechanism is added to the network,and the addition of the channel attention mechanism improves the network classification performance.The method solves the problem that the traditional sharpness evaluation function can only evaluate the sharpness of images of the same content with reference,and the method can be used in the field of image sharpness evaluation,especially in the field of cytopathological image sharpness evaluation.3.A method for evaluating the quality of cytopathological images based on the classification and reporting of diagnostic vaginal cytology（The Bethesda System,TBS）criteria is proposed.In order to overcome the lack of methods for evaluating the quality of cytopathological images.The method is based on the TBS diagnostic criteria and proposes a variety of evaluation criteria for evaluating the quality of cytopathological images.The image quality evaluation is completed by extracting features related to cytopathological image quality and inputting them into a regression model.Experiments show that the focus window selection method based on greyscale projection proposed in this paper can accurately add windows to cytopathological images,effectively reducing the computational effort of the sharpness evaluation function and improving the computational efficiency of the sharpness evaluation function.The focus prediction method based on thin-slab sample interpolation proposed in this paper can accurately predict the focus plane,thus improving the efficiency of fast acquisition.The sharpness determination method based on residual network proposed in this paper can classify cytopathological images with different focal planes in a clear and fuzzy way and evaluate the image sharpness without reference.The TBS-based cytopathology image evaluation method proposed in this paper can effectively evaluate the quality of cytopathology images and improve the efficiency of automated pathology diagnosis.