Cervical cell classification based on attention mechanism and multi-resolution feature fusion

Liquid-based thin-layer cell smears are very important for the early screening and prevention of cervical cancer, and computer-aided diagnosis can reduce the workload of pathologists. The cell classification method based on deep learning can process data efficiently. However, most classification methods are based on a single resolution for recognition. When the resolution is low, the processing speed of the whole slide image is faster, but lack of picture details, which makes the identification inaccurate. When the resolution is high, it takes more time to process the whole slide image, but with more image detail. To this end, we propose a model based on Attention Mechanism and Multi-resolution Feature Fusion Module (AMFM), which combine the advantages of various resolutions to classify cervical cells. Experiments show that the accuracy is increased by 3.93% and the AUC is improved by 0.022 on the four-classification task of the cervical cell compared to the model based on a single resolution.