Spatio-Temporal Split Learning for Privacy-Preserving Medical Platforms: Case Studies with COVID-19 CT, X-Ray, and Cholesterol Data

Machine learning requires a large volume of sample data, especially when it\nis used in high-accuracy medical applications. However, patient records are one\nof the most sensitive private information that is not usually shared among\ninstitutes. This paper presents spatio-temporal split learning, a distributed\ndeep neural network framework, which is a turning point in allowing\ncollaboration among privacy-sensitive organizations. Our spatio-temporal split\nlearning presents how distributed machine learning can be efficiently conducted\nwith minimal privacy concerns. The proposed split learning consists of a number\nof clients and a centralized server. Each client has only has one hidden layer,\nwhich acts as the privacy-preserving layer, and the centralized server\ncomprises the other hidden layers and the output layer. Since the centralized\nserver does not need to access the training data and trains the deep neural\nnetwork with parameters received from the privacy-preserving layer, privacy of\noriginal data is guaranteed. We have coined the term, spatio-temporal split\nlearning, as multiple clients are spatially distributed to cover diverse\ndatasets from different participants, and we can temporally split the learning\nprocess, detaching the privacy preserving layer from the rest of the learning\nprocess to minimize privacy breaches. This paper shows how we can analyze the\nmedical data whilst ensuring privacy using our proposed multi-site\nspatio-temporal split learning algorithm on Coronavirus Disease-19 (COVID-19)\nchest Computed Tomography (CT) scans, MUsculoskeletal RAdiographs (MURA) X-ray\nimages, and cholesterol levels.\n