在过去十年中,数据量、算法和高性能计算的巨大进步已将人工智能(AI)推向了高效用前沿。尤其是在医学影像和数字病理领域,AI的辅助诊疗系统成为了学术与医疗界共同关注的焦点。然而,随着对隐私保护问题的日益重视、政策法规的完善,数据共享成为亟待解决的问题。联邦学习作为一种“数据不动模型动”的范式,为数据的隐私保护与共享提供了一种新思路。本文提出一种联邦学习框架,在确保医疗影像数据本地化及安全性的同时,充分利用多中心肺炎影像数据集,训练高精度的人工智能模型,辅助肺炎影像诊断。通过本地化模型训练和参数服务器的梯度聚合,在不违反数据隐私的前提下,实现了模型的优化与更新。本研究的成果不仅提高了肺炎诊断的准确性和效率,而且扩大了系统的样本量和数据维度,为医疗大数据应用中高精度模型的构建提供了有力支撑,进而有助于提供更丰富和高质量的医疗服务,为社会公共医疗事业的发展贡献重要价值。In the past decade, the tremendous advancements in data volume, algorithms, and high-performance computing have propelled artificial intelligence (AI) to the forefront of efficiency. This is especially true in the field of medical imaging and digital pathology, where AI-assisted diagnostic systems have become a focal point of attention in both the academic and medical communities. However, with increasing attention to privacy protection and the perfection of policies and regulations, data sharing has become an urgent issue to address. Federated learning, as a paradigm of “model moving without data moving”, offers a new approach to the privacy protection and data sharing. This paper proposes a federated learning framework that, while ensuring the localization and security of medical imaging data, fully utilizes multi-center pneumonia imaging datasets to train highly accurate AI models to assist in pneumonia image diagnosis. Through localized model training and gradient aggregation on parameter servers, model optimization and updating are achieved without violating data privacy. The outcomes of this study not only improve the accuracy and efficiency of pneumonia diagnosis, but also expand the sample size and data dimensions of the system, providing powerful support for the construction of high-precision models for medical big data applications. This, in turn, helps to provide richer and higher quality medical services, contributing significant value to the development of public medical services.
近年来,在工程教育专业认证和审核评估的推动下,我国高等教育教学质量稳步提升,特别是计算机类相关专业在课程建设和质量提升方面取得了显著成果。本文提出了一种基于多课程相关性计算和能力评价趋势分析的能力评价联动的课程持续改进机制,旨在将传统的课程内循环方式改进为多课程联动的外循环方式。本文通过皮尔逊公式计算课程间的能力评价相关系数,验证了课程间能力支撑关系的客观性。并通过绘制相关课程能力评价趋势线,观察学生在相关课程中能力评价的变化趋势,并结合实际教学情况进行因果分析,提出了基于多课程能力评价趋势分析的课程持续改进方法。本文的研究为课程持续改进提供了一种新思路,强调了多课程间的联动改进和数据共享的重要性,以期为工程教育专业的课程建设和教学质量提升提供一定支持。In recent years, driven by the accreditation of engineering education and review evaluations, the quality of higher education teaching in China has steadily improved, particularly in computer-related disciplines, where significant achievements have been made in course development and quality enhancement. This paper proposes a continuous course improvement mechanism based on multi-course correlation calculation and competency evaluation trend analysis. The aim is to shift from the traditional intra-course cyclic approach to an inter-course external cycle approach. Using the Pearson formula to calculate the competency evaluation correlation coefficients between courses, the objectivity of the competency support relationships among courses is verified. Additionally, by plotting competency evaluation trend lines for related courses, the changes in students’ competency evaluations across different courses are observed. A causal analysis is then conducted based on actual teaching situations, leading to the proposal of a continuous course improvement method based on multi-course competency evaluation trend analysis. This research provides a new approach to continuous course improvement, emphasizing the importance of inter-course linkages and data sharing, with the aim of supporting course development and improving teaching quality in engineering education programs.
暂无评论