Overview

Philips N.V. is a Dutch multinational conglomerate corporation headquartered in Amsterdam, formerly one of the largest electronics companies in the world, currently focused in the area of health technology.

Philips Remote Services explores how to make best use of the vast amounts of non-clinical/nonpatient information already being generated by imaging equipment. For example, a typical MRI scanner produces an average of 300,000 log messages every day. Philips is feeding such data into algorithms and models that can interpret it in a way that benefits caregivers and patients.

Supernova Award Category

Data to Decisions

The Problem

Healthcare costs are exploding worldwide, and unplanned downtime of medical imaging systems can have significant impact on health service providers.

Medical imaging systems such as MRI and CT scanners must provide optimal clinical performance and predictable cost of ownership. While clinicians understand the need for maintenance, any system downtime can be costly.

Philips has tens of thousands of connected medical devices around the world that include mission critical MRI machines and CT Scanners. Every imaging system contains many sensors and generates daily large volumes of data: a single MRI system can generate millions of log events each day and hundreds of thousands of data points from its sensors.

Philips’ goal is to move to a zero unplanned equipment downtime by creating predictive analytics models to maximize clinical system availability for medical imaging systems.

The Solution

Rather than waiting for a medical device to fail and dispatch a field-service-engineer to fix it, Philips’ devices are connected to a remote monitoring infrastructure that uses big data to detect and predict issues before they occur, or they have impact on the clinical workflow.

Philips Healthcare embedded predictive analytics in their product and moved from reactive to data-driven, proactive maintenance, utilizing new sources of sensor data along with machine learning models to enable scheduled, predictable and non-intrusive service actions that don’t interrupt regular clinical workflow.

Philips realized the potential of this data for predicting medical device technical issues and built a big data analytics solution by combining its own HealthSuite Digital Platform with the Vertica Analytics Platform. Roughly 500 TB of data is integrated in a big data analytics solution based on Vertica. 

The results

Predictive models are used to generate proactive alerts from data received daily from the installed base of medical devices. The alerts are reviewed by a team of remote monitoring engineers that take care of scheduling preventive maintenance actions worldwide, contributing to decrease unplanned downtime.

The first version of the analytics solution, the predictive models and the new processes were developed in roughly one year of work. The development of the entire solution was fully data driven. Data such as part replacements, contracts, records of maintenance and failures was used to estimate the potential impact of predictive models before realizing them so that the right models could be built, and the resources needed to build them were used optimally. Data was used to create the models and data is used to measure their effectiveness and performance to drive continuous improvements.

Philips now has tens of predictive models covering hundreds of failure modes for many of its connected medical devices. In roughly one out of three cases, Philips approaches the customer proactively rather than waiting for a call to its call center. For connected image-guided therapy devices, Philips’ remote monitoring and proactive maintenance solution has shown to reduce system downtime by 14% allowing 135 hours of additional operational availability per year which translates to the ability to treat more patients.

Metrics

Remote Service cost savings so far:

Transformed the organization’s customer services from being 5% proactive to 30% proactive with predictive maintenance enhancements.

Remote monitoring increased more than 20 times due to improved ingestion and the speed of Vertica’s machine learning.

First-visit fixes or remote fixes improved from approximately 50% to more than 80% due to machine learning providing more accuracy in identifying problems, providing recommended solutions, making assignments based on required skills, and ensuring the proper parts are supplied to the service tech.

Allowed decision-makers to identify training needs for service techs and clients so they could properly use equipment. Client training leads to actions to improve business outcomes and to reduce future issues.

“Remote Service from Philips has enabled us to achieve system availability of more than 99 %. Failure of our cooling water supply, for example, was detected at an early stage and a potential quench of the MR prevented.” Prof. Dr. H. P. Busch, Director of the Center for Radiology, Neuroradiology, Sonography and Nuclear Medicine, Krankenhaus der Barmherzigen Brüder, Trier, Germany

The Technology

Philips’ HealthSuite Digital Platform is integrated with the Vertica Analytics Platform. Roughly 500 TB of data is integrated in a big data analytics solution based on Vertica.

“I find Vertica’s ease of integration and its natural speed to be highly valuable. Vertica’s self-sufficiency makes it possible for us to spend our time and resources on activities that provide business value.” Mauro Barbieri, Philips.

Disruptive Factor

In some ways it’s a strange concept. Imagine a garage calling to say that your car would need some fairly major repairs within the next couple of weeks, even although it was - as far as you could make out - running perfectly. This analogy demonstrates why effective use of the available data is so important. If you can prove to a customer that carrying out regular maintenance and timely replacement means ensured continuity of care and even cost savings - because problems are easier to identify and quicker to address - then they are more likely to embrace this new philosophy.

Philips’ remote services organization is using the output from this goldmine of information to drive a transformation; from reactive to proactive maintenance. They have already developed more than 40 proactive data analytics algorithms which analyze system log files on a daily basis in order to recognize patterns that analyze the need for future equipment maintenance. This really marks a fundamental shift. Currently, equipment maintenance is often carried out when something goes wrong. The question is then; ‘what happened?’ We can now go beyond that, instead asking; ‘why did it happen?’ In fact, through constant monitoring and alert generation we are continually enquiring; ‘what is happening now?’

Shining Moment

Over 20,000 Philips imaging systems are monitored daily to identify patterns that could indicate preventive action is required. 10,000+ cases are proactively handled every year by Philips’ centralized monitoring team to reduce unnecessary downtime.