基于联邦学习的轻量级医疗影像辅助诊断系统设计与实现

Design and Implementation of a Lightweight Medical Image Assisted Diagnosis System Based on Federal Learning

基于深度学习的医疗影像辅助系统在辅助疾病诊疗方面具有较高的应用价值,但在应用过程中仍面临诸多挑战。由于医学数据的独特性,数据孤岛问题导致的数据源不足和标签缺乏制约着人工智能赋能疾病诊疗。联邦学习可以在保护数据安全和隐私的前提下有效地进行数据共享,从而发挥多中心数据的最大价值。文章设计并实现了一个基于联邦学习的轻量级医疗影像辅助诊断系统,使用改进的轻量级RetinaNet算法模型识别医疗图像,构建一个安全、高效、可扩展、轻量级的分布式医疗辅助诊断平台。

The medical image assistant system based on deep learning has high application value in assisting the diagnosis and treatment of diseases,but it still faces many challenges in the application process.Due to the uniqueness of medical data,the lack of data sources and labels caused by the data island problem restricts the diagnosis and treatment of diseases enabled by AI.Federal learning can effectively share data on the premise of protecting data security and privacy,so as to maximize the value of multi-center data.In this paper,we design and implement a lightweight medical image assisted diagnosis system based on federated learning architecture and use an improved lightweight RetinaNet algorithm model for medical image recognition to build a secure,efficient,scalable and lightweight distributed medical assisted diagnosis platform.