摘 要: 周围型面瘫作为一种多为急性发作的临床常见病,治疗上需要准确掌握患者发病状态,以便判断治疗的最佳时机。现有的面瘫治疗方法多依赖于医生对患者临床症状的直观判断,而本文通过研究当前基于深度学习的异常检测方法,并将异常检测生成对抗网络(GANomaly)作用于面瘫图像上,实现正常人脸和面瘫人脸的识别分类,为医生诊断提供辅助工具,可有效提高诊断效率,弥补现有方法的不足。训练后的GANomaly网络可有效分辨出面瘫图像的异常。实验结果表明,基于深度学习的GANomaly网络可有效实现面瘫的诊断识别。 |
关键词: 深度学习;面瘫识别;异常检测;GANomaly |
中图分类号: TP39
文献标识码: A
|
|
Application Research of Facial Paralysis Recognition based on GANomaly Network |
WANG Qi1, CHEN Gong1,2, HU Wenxin1, HU Jia1
|
( 1. School of Artificial Intelligence and Information Technology, Nanjing University of Chinese Medicine, Nanjing 210046, China ; 2. Jiangsu Province Hospital of Chinese Medicine, Nanjing 210029, China )
20190616@njucm.edu.cn; wqdlrb@163.com; 20200998@njucm.edu.cn; 20200997@njucm.edu.cn
|
Abstract: Peripheral facial paralysis is a common clinical disease that is mostly acute. It is necessary to accurately grasp the patient's onset status in order to determine the best time for treatment. Existing facial paralysis treatments mostly rely on doctors' intuitive judgments of patients' clinical symptoms. This paper proposes to study current deep learning-based anomaly detection methods, and apply generative adversarial networks for anomaly detection (GANomaly) to facial paralysis images, so that recognition and classification of normal faces and facial paralysis ones are realized. It provides auxiliary tools for doctors to diagnose, which effectively improve the efficiency of diagnosis and make up for the shortcomings of existing methods. The trained GANomaly network can effectively distinguish the anomalies of facial paralysis images. The experimental results show that the GANomaly network based on deep learning can effectively realize the diagnosis and recognition of facial paralysis. |
Keywords: deep learning; facial paralysis recognition; anomaly detection; GANomaly |