Research On The Model Of Preliminary Screening Of Cervical Cell Disease Based On Fuzzy Neural Network

With the increasing rate of the prevalence from cervical cancer between the females,cervical cancer has been one of the most normal cancers which disturbs a number of females in the world.Cervical liquid-based thin-layer cytology(TCT)is considered to be the most reliable method for detecting early cervical precancerous lesions.With the help of image processing technology,fuzzy mathematics and neural network,in this study,a preliminary screening model,whose purpose was to conduct preliminary classification and diagnosis based on the pathological changes of cervical cells to help Pathological diagnosis doctors save time and improve the quality of diagnosis.,based on optimized fuzzy neural network,BP neural network and convolutional neural network was established.The main method in this dissertation is the literature analysis and experimental analysis method.Three-step preprocessing,including segmenting the cell image,deleting the background information purposefully and enhancing the quality of the target area from interest Image,was used as the appropriate image processing technology after scanning the original image.The optimized wavelet transform technology based on the combination of neural network and wavelet transform technology is applied to processing the image of cervical cells.2.The important cell characteristics,including the ratio of nucleus to mass,are extracted from the picture according to the classification requirements of TBS.3.proposes fuzzy inference levels and design fuzzy rules were proposed based on clinical experience.4.Four preliminary screening models were used to train from the extracted features and classify these features into different classification.The root mean square error,accuracy,sensitivity,specificity,accuracy and F value were selected as the evaluation indexes.The experimental results are represented by confusion matrix,and various indexes are obtained after calculation: The accuracy,sensitivity,specificity,precision,F value and negative predictive value of Gaussian adaptive fuzzy neural network model for screening cervical cytopathy were 98.93%,99.6%,98.8%,99.4%,0.995 and 0.992 respectively.The accuracy,sensitivity,specificity,precision,F value and negative predictive value of BP neural network model for screening cervical cytopathy were 96.26%,98.56%,97%,98.46%,0.985 and 0.972 respectively.The accuracy,sensitivity,specificity,precision,F value and negative predictive value of the triangle adaptive fuzzy neural network model for screening cervical cytopathy were 97.3%,98.98%,97.8%,97.8%,0.9893 and 0.9799 respectively.The accuracy,sensitivity,specificity,precision,F value and negative predictive value of convolution neural network model for screening cervical cytopathy were 97.73%,98.99%,98.99%,0.9899 and 0.98 respectively.In the comparison of sensitivity,stability and accuracy of the model,the conclusion of this dissertation is drawn: The preliminary screening model of cervical lesions based on adaptive fuzzy neural network with Gaussian membership degree is the best and reasonable in the classification and prediction of cervical squamous intraepithelial lesions.