Daytime Project

Philips MRI

Smart Maintenance for MRI

Introduction to MRI

MRI (Magnetic Resonance Imaging) devices are used to create images of the inside body for diagnostic purposes. Referring physicians will base their diagnostic conclusion on the diagnostic images of the patient made by the MRI device.


Figure 1. Philips MRI system

A session where the patient is placed inside the MRI system is called an examination. A typical examination duration takes between 10 and 30 minutes. During the examination multiple scans are performed. Typically 5 to 10 scans per examination are performed. Each scan results in multiple images which can be experienced as slices through the body. Each scan has a different purpose. Scans differ to balance between time and resolution of the resulting images. As an example: the first scan of an examination is a Scout scan which is a fast low resolution scan to view a large part of the body enabling planning of the next high resolution scans on the right anatomy or clinical question. Scans also differ in technique to balance between robustness for patient motion during the scan and resolution or sharpness of the images. Techniques also differ in what needs to be shown depending on what the referring physician wants to see, e.g. the grey or white matter of the brain, flow of fluids, tissues like nerves and cancer cells. Scans are controlled by scan parameters which can be modified by the operator to optimize the scan for the diagnostic question. Hundreds of parameters are available to modify and optimize the scan. The operator typically modifies a few parameters to adapt the scan to the patient, obtaining the best image quality.


Figure 2. Examples of resulting MRI images using different scan techniques. From left to right: imaging blood vessels, imaging brain, imaging brain trauma, fMRI (functional MRI).

Current Status and Use Case Challenges

Maintenance Process

Uptime is a critical KPI for MRI systems. Systems which run 24hrs / 7 days in the week are no exception anymore. Planned and remote predictive service are key elements in achieving those needs.

Planned Maintenance

The activities performed during the maintenance are defined per configuration. They vary from re-adjusting to proactive cleaning filters.

Predictive Maintenance

Majority of systems are connected and machine data is uploaded to a Philips Cloud. The machine data is used to monitor the systems under service contract. Alerts are created towards the service organization in case of degradation and predictive failures based on analysis of the available data.

Corrective Maintenance

In case of failures, a customer calls the helpdesk. The helpdesk will try to remotely diagnose the problem using uploaded machine data in combination with remotely triggered tests, executed locally at the MRI system.

HelpDesk Engineer

Answers the call and delegates the investigation by submitting a Service Work Order (SWO). A SWO is a data record in which actions are consolidated between the moment of reporting the issue and solving the issue.

Current Status and Use Case Challenges

Models to support predictive service are built on top of collected machine data. Typical machine data consists of:

Configuration Data

Identifies the unique composition of the MRI system, with identifiers for all hardware elements and versions of the hardware


One central logfile across the complete MRI system contains in an unstructured way the workflow of the system including error events reported by hardware.

Sensor Data

Dedicated files per sensor property contain the sampled values of sensors

Test Data

Test results of tests executed at idle time and test results of tests triggered by Service Engineers.

The Use Case for MRI systems in the context of Daytime is built around 5 challenges:

- To which extent can behavioral information automatically be retrieved from logfiles for predictive maintenance? -> Anomalous Logline Detection Tool

- Can enhanced AI techniques be used to monitor sensor data for predictive maintenance? -> Yanomaly Diagnostics Tool

- To which extent can textual data from earlier corrective maintenance be used to validate proactive service models? -> Multi Agent Reinforcement Tool

- What are the parameters of age-based planned maintenance? -> Age-Based Maintenance Tool

- How can Augmented Reality applications accelerate the work of a Field Service Engineer during corrective maintenance? -> Augmented Reality Maintenance Guide

Contact Information: