A Two-level Cloud Storage System Based on Asynchronous Message for Medical Image Big Data

Medical imaging has rapidly become a big data problem based on the fact that there are approximately petabyte level of medical images in all the repositories. These data are managed by picture archiving and communication systems in a central high performance disk array that healthcare organizations have to pay an expensive price for storage. It's very important that patient data is collected correctly in a reliable storage media for an indefinite period of time. An economical solution will be leveraging the cost-efficient commercial cloud computing services (i.e. Amazon EC2, Microsoft Azure, Google Cloud Platform, AliCloud, etc.) to host data and conduct required analysis on demand. However, the data should be quickly accessible at any given time in clinical practice that the medical applications can access and render the data to doctors in real time. Therefore, it is impractical for radiologists to wait for a few minutes to download images from public cloud storage. As we all know that the cloud storage system usually is not good at real time scenery applications and the privacy issues should also be considered for medical data in public cloud. This paper proposed a new two-level private cloud storage system based on asynchronous message for medical image big data and the system is developed on Hadoop distributed file system. Two innovative technologies are presented that are message queue based medical image storage and online hot data pooling. The experimental results show that the system has superior real-time performance to support the front-end applications.