多中心临床科研服务

对于医学研究，越来越多的项目依赖大量数据的积累，单个医疗机构的数据量不足以支撑，因此研究开展往往需要多家医疗机构将各自的医学数据联合起来，并且由多位行业专家进行科研协作。然而，由于数据孤岛的问题、传统数据脱敏的局限性带来的隐私问题、数据监管问题，导致医疗数据在多方之间无法进行共享和互通，成为当前医学研究面临的最大挑战。在这样的背景下，医渡云推出了基于多方安全计算的多中心临床科研服务，通过在多机构内部署安全计算服务，来构建多中心研究网络，通过同态加密、秘密分享、混淆电路等多方安全计算技术来保证机构间无私有数据传输。在保障强隐私保护约束的前提下，达成跨多机构联合统计建模的需求。 医渡云自研了安全计算引擎YiduManda，采用多方安全计算、联邦学习、联盟区块链等技术保障数据的安全高效分享，应用于临床研究feasibility调研、大样本量队列管理、疾病预测模型、基因组数据分析等多种应用场景，帮助科研人员提升多中心研究的效率，通过优化通信帮助控制多方安全计算通信开销的二次增长。同时，还可以通过多通道实现更细粒度和支持优先级的调度。客户也可以根据需要选择不同的底层安全算法插件，并获得卓越的安全性保障。

For medical research, an increasing number of projects rely on the accumulation of large-scale datasets. The volume of data held by a single medical institution is insufficient to support such research endeavors, so conducting such research often requires multiple medical institutions to pool their respective medical data and collaborate on scientific work with a number of industry experts. However, issues including data silos, privacy risks stemming from the limitations of traditional data de-identification, and data regulatory challenges have prevented medical data from being shared and interoperated across multiple parties, which has emerged as the most significant challenge facing contemporary medical research. Against this background, Yidu Cloud has launched multi-center clinical research services based on Secure Multi-Party Computation (SMPC). By deploying secure computing services across multiple institutions, it constructs a multi-center research network, and leverages SMPC technologies such as homomorphic encryption, secret sharing, and garbled circuits to ensure that no private data is transmitted between institutions. Under the constraints of strict privacy protection, it meets the demand for cross-institution joint statistical modeling. Yidu Cloud has independently developed the secure computing engine YiduManda, which adopts technologies including SMPC, Federated Learning (FL), and consortium blockchain to enable secure and efficient data sharing. This engine is applied in various scenarios such as clinical research feasibility surveys, large-scale cohort management, disease prediction models, and genomic data analysis, helping researchers improve the efficiency of multi-center research. It curbs the quadratic growth of communication overhead in SMPC through communication optimization, and also supports fine-grained and priority-aware scheduling via multi-channel mechanisms. Additionally, customers can select different underlying secure algorithm plugins according to their needs, and obtain outstanding security guarantees.

搜集汇总

背景与挑战

背景概述

该数据集提供多中心临床科研服务，利用同态加密等安全计算技术实现医疗机构间的隐私保护数据协作，支持联合统计建模和疾病研究。服务由医渡云提供，涵盖临床研究、疾病预测等多种医疗健康应用场景。