Part replacement prediction for Philips digital X-ray machines

Abstract

Proper functioning of medical diagnostic devices is crucial. Valuable digital diagnostic devices like Philips X-ray machines are complex and comprise of many components. When some specific components fail, it can prevent the proper functioning of the machine. If this happens while a patient is being diagnosed, it causes extreme discomfort for the patient. Such situations are risky for machine operating technicians as well. When the X-ray machine stops working, the hospital needs to cancel all appointments which need an X-ray until the machine is fixed. Hence it causes inconvenience to other patients as well. The hospital then registers a service request to Philips, who is a trusted service partner. Then one of the Philips Service Engineers visits the hospital to repair the machine. Most of the time, such repairs involve part replacement for the machine. The Service Engineer makes the judgment based on his/her domain expertise. To make the correct judgment, the Service Engineer sometimes needs to make multiple visits to the hospital, which is costly and inefficient. In the last few years, the connectivity of Philips X-ray machines has increased significantly, and thanks to our Data Science team's deliberate efforts, the event log data from all connected machines are available in a well-organized and easy to retrieve data storage system. This project explores data-driven methods to predict part replacement for various Philips digital X-ray machines and evaluate their usability and scope with inputs from domain experts. We’ve used Decision Trees to represent our data-driven hypothesis to the stakeholders in an intui ve and easy to understand manner. Successful implementation of our methods contributes to the hospital’s smooth functioning and Philips’ proactive service philosophy. And hence the project delivers significant business value.