Cervical Cancer Detection and Diagnosis Based on Saliency Single Shot MultiBox Detector in Ultrasonic Elastography

Cervical cancer is increasingly threatening the health of women, then early screening and prevention of cervical cancer is very necessary. A traditional saliency cervical cancer detection method in Ultrasound image, assuming that there is only one salient object, is not conductive to practical application. Their effects are dependent on saliency threshold. Object detection model provides a kind of new solutions for multiple salient objects. Shot multiBox detector can accurately detect multi-objects with different scales simultaneously except for small cervical cancer regions. To overcome this drawback, this paper presents a new multi-saliency objects detection model, appending deconvolution module embedded within attention residual module. Experiments show that our proposed diagnosis algorithm achieves higher detection accuracy than comparison algorithms. Also, it improves detection performance for mult-saliency cervical cancer objects with small scales, which greatly improve the diagnosis accuracy of cervical cancer.