Cloud Based Big Data Platform for Image Analytics

This paper describes a framework for leveraging the cloud platform and big data technologies to perform medical image data analytics even across multiple institutions. The key challenges we have addressed are the security, privacy requirements, scalability of the solution and use of cloud based analytics for the existing proprietary legacy systems with network connectivity in hospitals. The Digital Imaging and Communication in Medicine (DICOM) medical images acquired for clinical purposes are to be shared with radiologists and institutions for interpretation of images to support disease diagnosis, who are miles away from the institutions. Prior to sharing DICOM images with third parties, we need to follow strict guidelines with respect to de-identification of Protected Health Information (PHI) from the medical data to be shared. To de-identify PHI information from images, we have devised a newer generation de-identification techniques based on the big data platform to obscure privacy information from DICOM tags, embedded text in images and face information from CT and MR images. Further to facilitate accurate analysis on medical images, Clinical Decision Support System (CDSS) inference model is developed. In the paper, we proposed few use cases and developed a cloud based framework to distribute/share clinical images with CDSS to radiologists on the move and to the institutions that do not exist in the host organization premises. The developed solution satisfies all the PHI de-identification requirements, supports scalability by efficiently using big data cloud technologies, for faster and reliable processing. We covered obstetrics/gynecology ultrasound application using deep learning approach as an example to illustrate CDSS.