Enabling Better Medical Image Classification Through Secure Collaboration

Abstract

Privacy is of growing concern in today's day and age. Protecting the privacy of health data is of paramount importance. With the rapid advancement in imaging technology, analysis of medical images is now one of the most dynamic fields of study today. Image analysis is performed for a variety of purposes, ranging from image enhancement to image segmentation. It can easily be seen that having access to more information makes the analysis results more accurate. For example, supervised classification based image segmentation requires good and plentiful training data. We wish to utilize the training data at different locations to obtain more accurate image segmentation while still protecting the privacy of individual patients. Work in the field of secure multi-party computation (SMC) in cryptography shows how to compute functions securely and quantifies what it means to be secure. Applying SMC protocols in image processing is a challenging problem. This paper looks at how some of this work can be leveraged to perform privacy-preserving image analysis and classification.