Hacking and Artificial Intelligence in Radiology: Basic Principles of Data Integrity and Security

Abstract

Category: Radiology Informatics and Artificial Intelligence Modality: Multiple part of sophisticated purchasers and users. This article is intended to provide an introduction to the principles of AI and data security that should be common knowledge among radiologists. AI is the use of a computer to recognize patterns or objects, draw inferences, or perform other functions that were not explicitly described or enumerated in its programming.1,2 It is a broad topic with applications in all areas of human/computer interfaces.3,4 In the work presented here, requirements for maintenance of data integrity and security in AI are described. Data integrity and security is of particular importance in AI because of the “black box” nature of machine interpretation of images. The objective of this article is to provide a rudimentary understanding of security principles in AI for practicing clinical radiologists without discussion of technical details regarding the actual algorithms used. One of the most important goals in this endeavor is an appreciation of the need for vigilance at the user level regarding practices that may lead to security concerns.