A Machine Learning Utility for Detection of Potential Protected Health Information Images

Abstract

Often, dental x-rays and digital images which may contain sensitive patient information are saved to network and local directories increasing the vulnerability as well as the legal liability of the institutions that process them. To reduce the risk, various approaches are employed by IT staff to detect such images and ensure their proper handling. Searching for images manually is a tedious and time-consuming task; rather automated tools would be preferred. The goal of this project was to investigate how machine learning can be used for automated image recognition as a tool in dental image detection. This paper presents the design, implementation and testing of a user-friendly tool to analyze directories and subdirectories identified by a user to identify and handle Protected Health Information (PHI). The tool provides an interface for a user to select various options for file handling and image search parameters that are then used with a convolutional neural network (CNN) to identify dental radiographs in an arbitrary list of files. Experimental results show that the tool accurately detects potential PHI-related files, leading the way for a practitioner ready deployment.