TechNation reached out to industry leaders to find out the latest regarding AI and machine learning for the monthly roundtable article. Sharing their knowledge about AI and machine learning are TINC CMMS co-founder and CEO Luis E. Fernández, Medigate Vice President of Systems Engineering Stephan Goldberg, Asimily Director of Business Development Michael Nielsen and Nuvolo Chief Product Officer Asim Rizvi.
Fernández: Learning from medical equipment experiences and also recording them is of huge importance, since health care establishments can have rotations in their staff. AI/ML can help build predictive models for medical equipment failure and also help newly incorporated clinical engineers learn from previous experiences to reduce medical equipment downtime and increase their availability. Just ask yourself … How many times have we solved this problem? Was this problem solved before? If so, how was it solved? AI/ML can not only answer these questions but also guide the clinical engineering department on how to prevent/anticipate future problems.
Goldberg: HTM professionals can leverage AI and machine learning for several specific applications, most notably predictive analytics for strategic modalities failure and offline time. In turn, this allows for better project management planning, as well as enabling critical functions like spare part procurement and maintenance activity. By utilizing these technologies to achieve critical functions, HTM professionals can help to create a more collaborative, forward-driven health care ecosystem that is based ondata-validated insights and streamlined processes.
Nielsen: AI or machine learning can be used in a variety of ways, including:
Rizvi: HTM has significant challenges across several areas of the business that can benefit tremendously from AI and machine learning. In an environment of disparate legacy client server solutions where data structures are not well architected for conformity, structure and analytics, it can be very difficult to even consolidate the processes and data that drive the engine of HTM. This preverbal gold mine of data is locked behind ancient systems and paper reports. AI and machine learning can transform this landscape by normalizing this data for consolidation to new cloud platforms, providing new insights from unstructured data and helping reduce cybersecurity risk for equipment by identifying vulnerabilities and orchestrated automated action.
Fernández: AI/ML represent the capability of machines and systems to learn, bringing great benefits for numerous sectors. Clinical engineers around the globe are working everyday repairing and practicing preventive maintenance on a diversity of devices. If their operations are not only registered but also introduced as data to be modeled by a system, key information can be produced to showcase clinical engineers and medical equipment performance. It can also enhance decision making by showing which devices are more feasible than others, allowing health care institutions to base their decisions on trends of their current operations.
Goldberg: AI is immensely useful for not only understanding the current state of medical equipment, but predicting their behavior and maintenance needs. For example, calculating the maximum-likelihood-estimate for the time frame that devices will actually break or malfunction (per device model) or using a predictive (online) model to alert about a future failure. By achieving these insights, HTM professionals can then minimize the allocation of redundant PM efforts and eliminate device downtime altogether.
Nielsen: When considering medical equipment maintenance, many aspects need to be looked at, which include deciding how long to use a device, when to service the devices, what level of servicing is required and others. All of this requires looking through a lot of information like age of the device, utilization and many other parameters including any events happening on the device. AI or machine learning can combine all this information and help HTM professionals make decisions on when and how to maintain their equipment.
Rizvi: Maintenance of medical equipment is a complex and scientific process that requires planning and deliberate strategies. The key drivers are compliance, patient outcomes, security, cost reduction and equipment life extension. It is very difficult for HTM teams to deliver results without insights into the service history, asset performance and utilization of equipment. AI and machine learning solve this by providing these insights against traditionally unstructured data to help set more efficient equipment maintenance schedules, address vulnerabilities as they are detected not just for one device but similar devices across the whole asset base as well as base maintenance schedules on utilization in an orchestrated and automated fashion versus traditional costly calendar based schedules. Predictive maintenance can change this landscape but cannot be accomplished without the use of machine learning to arrive at the right maintenance plan for the assets.
Fernández: Since AI/ML produce information regarding usual device and clinical engineering staff operations, by using AI/ML an organization can detect abnormal activity based on their registered data. With this information at hand, it’s much easier and faster to detect cybersecurity threats. Comparing the current analysis that is based on human intelligence, these disciplines and technologies can be more efficient as they’re mostly available on a 24/7 basis, can analyze huge quantities of data, and have the capabilities on running a vast diversity of algorithms to alert of any unusual activity.
Goldberg: AI for cybersecurity – and particularly healthcare IoT cybersecurity – is essential to ensuring visibility and eliminating the knowledge gaps that plague hospital networks. By this I mean the thousands of devices connected to a network , all of which must be discovered, maintained and protected from malicious threats. Notably, AI can generalize device profiling for initial discovery (for devices not seen previously) and before DPI efforts. Once this is achieved, HTM professionals can better create policies and more robust, validated baselines, which allows for enhanced alerts on anomalies and other suspicious behavior.
Nielsen: Cybersecurity has many aspects to it:
Rizvi: Cybersecurity is a key concern for all HTM teams today. Twenty to 30 percent of the fleet is connected and this number keeps rising every day. However, these OT assets are typically a blind spot for cybersecurity tools. AI solutions identify vulnerabilities and security events from a monitoring perspective but that is not enough. Once a vulnerability is identified against a MAC address, how do we address the vulnerability, which device is it anyway. This is where OT cybersecurity combined with AI is so powerful. Machine learning algorithms monitor and identify, automated orchestration takes it to the next level and identifies the asset and all of its attributes to kick off workflow to address. While this is happening machine learning finds matches in the asset base for all other assets that share the same vulnerability and derive workflow and visibility of threat that was never possible before. This is a transformational in solving traditionally paper and manual methods for addressing cybersecurity concerns.
Fernández: Trend analysis and modeling can play a huge role when building a procurement or maintenance plan for healthcare technology, as most clinical engineering departments rely on limited information that is available on systems that are not designed exclusively for an HTM manager. If medical equipment activity is registered on paper, spreadsheets, ERPs or any CMMS that has not been designed for the clinical engineering department, the organization is leaving big money on the table. Every data input should be used for either identifying a device/activity and/or to produce a metric or indicator that can help the clinical engineering department make a decision. If we have key information at hand about risk, performance, and finance of every piece of equipment, and we can study/compare it over a period of time, we can surely enhance not only patient safety, but also the organization’s feasibility in medical equipment related topics.
Goldberg: Similar to its other applications within a health care ecosystem, AI for equipment planning provides an avenue to better understand and anticipate the best ways to maintain and deploy these devices. Recently, AI has become an increasingly useful tool for device procurement, especially in the age of COVID-19. By understanding and analyzing key usage trends, you can then strategically deploy these devices to not only achieve cost savings, but improve clinical care. For example, hospitals that have a low usage rate for ventilators could conceivably recognize this and move the devices to a sister site within the greater health system, helping to prepare for an influx of patients or otherwise bolster a clinical staff under great demand. AI provides these types of actionable insights that allow us to truly understand and be proactive in our planning, rather than reactive and slow to respond.
Nielsen: HTM professionals need to plan for how many devices they need and in which facility. Several parameters come into play here like utilization, number of devices as well as many of the parameters around maintenance and cybersecurity along with others. AI or machine learning can help with analyzing across all the information and provide the HTM department with information on equipment planning
Rizvi: Health care equipment facility planning is a complex task with so many asset types, models and procedures. The planning has to consider skill sets, manufacturer procedures, tool needs, part needs and technician availability. In progressive cloud solutions, this data is all captured but machine learning can help with automated routing and scheduling while considering all of these factors as rules to the model. Planning is a tedious task that can be automated with the help of AI to reduce errors. Further inclusion of utilization and metric-based predictive maintenance planning takes it to the next level to prevent corrective maintenance scenarios that are extremely costly to the organization and disruptive to the core mission of patient outcomes and patient safety.
Fernández: Before any clinical engineering professional acquires a CMMS, they must first set the questions that need to be answered with the help of this tool in order to benefit the organization’s strategy. A CMMS is much more than just an asset tracking tool, it’s a very powerful tool that can help guide clinical engineering departments into accomplishing their main goal: keeping every piece of technology available, safe, accurate and feasible during its lifecycle. I believe that choosing a CMMS should not be based on the experience of others; the organization needs to assess their current processes, stakeholder expectations and organizational goals so that this can drive their market research. What doesn’t get measured simply won’t ever be improved.
Goldberg: AI is a great algorithmic architecture and works best with unstructured big data. For structured data such as device communications, DPI is essential to extract relevant attributes from the device’s communications and logs, enabling users to better identify, profile and properly analyze the devices’ behavior. It’s important to note that AI is only a piece of the puzzle, not a silver bullet that solves all our problems from a cybersecurity, health care or device perspective. Rather, the technology’s applications to ensure greater device familiarity is essential for contextual decision making and, in turn, better management and clinical care.
Nielsen: Considering the evolving needs and market, an asset tracking system should handle a wide variety of devices and be flexible and configurable. The vendor’s roadmap is also important since technology is rapidly evolving. Other aspects include whether the system integrates with other systems, though for this if the asset tracking system uses APIs, then this should be easily achievable. Finally cost and scope of the asset tracking system should be scalable so that the system does not become prohibitively expensive as the utility of the system is scaled up.
Rizvi: Asset tracking and performance management systems traditionally have been client server based with disparate databases and unstructured data. There is a tremendous lack of flexibility in these legacy systems. A few important aspects that are necessary for taking the HTM process and outcomes to the next level are process and workflow flexibility, secure cloud access to ensure availability not just in one site but everywhere work is done, cutting edge automation to reduce tedious planning, mobile experiences to reduce paper and data entry errors as well as focus on cybersecurity to prepare for the next age of assets that are connected. Without the above considerations, a database with a UI only will never be able to meet the challenges of the new and demanding remit of HTM.
*By entering your email address, you agree to receive emails regarding TechNation Magazine, Webinars, and Exclusive Promos.
© 2020, TechNation Magazine. Site designed by MD Publishing, Inc.