Research On Detection Algorithm Of Cervical Cancer Cells Based On Improved YOLOX

The early screening of cervical cancer cells is of great significance in reducing the mortality rate of cervical cancer.Currently,the screening of cervical cancer cells is mostly based on manual diagnosis of cervical cell images on Papanicolaou smear slides.With the popularization of deep learning technology,the application of target detection algorithms for cell detection and identification has become more widespread.Although existing cervical cancer cell detection algorithms can improve detection efficiency and accuracy to a certain extent and assist doctors in completing diagnostic work,there are still problems of low detection accuracy,false positives,and missed detections in complex background and heavily overlapping and occluded cervical cell images.In response to these problems,this paper proposes and designs an improved YOLOX cervical cancer cell detection algorithm,which effectively improves the detection accuracy of cervical cancer cells.The main research work of this paper includes the following aspects:1.The CBAM convolutional block attention module is embedded in the output layer of the backbone feature extraction network in the YOLOX algorithm so that it connects to the subsequent feature fusion network.In this way,the feature layer before performing feature fusion will complete the redistribution of the importance of channel and spatial feature information in the CBAM module,focusing on the desired target features to make it work better in the subsequent feature fusion,thus improving the network’s ability to recognize cells in complex backgrounds.2.The feature fusion using PANet structure in the original network is improved to ASFF adaptive spatial feature fusion,which improves the scale invariance of features by adaptively learning weights,effectively filters the conflicting information between features at different levels,and retains only the information similar to the target for fusion to improve the recognition ability of the network for stacked occluded cells.Through the final experiments,it is proves that the improved YOLOX-based cervical cancer cell detection algorithm improves the detection accuracy from 80.04% to 93.43% based on the original YOLOX algorithm.It is expected to further assist doctors in real-time detection and screening of cervical cancer cells,thus improving the current level of cervical cancer cell detection and bringing great help to the early screening and prevention of cervical cancer.