May 17, 2020

Infosys Consulting: the future of AI and automation in healthcare

healthcare
Technology
Health technology
Artificial intelligence
Matt High
8 min
Infosys Consulting: the future of AI and automation in healthcare
Global Head for Artificial Intelligence and Automation, John Gikopoulos, on how AI and machine learning is disrupting the healthcare ecosystem.

“The...

Global Head for Artificial Intelligence and Automation, John Gikopoulos, on how AI and machine learning is disrupting the healthcare ecosystem.

“The robots aren’t taking over; it’s a misconception that’s really important to move away from,” says John Gikopolous, Global Head for Artificial Intelligence and Automation at Infosys Consulting. “Unfortunately - or perhaps fortunately - the future isn’t Pepper the Humanoid Robot, and the greater use of AI, machine learning and robotic process automation (RPA) in healthcare doesn’t mean we’ll end up with some vision of Arnold Schwarzenegger’s Terminator diagnosing us or deciding whether we live or die. 

“The reality is that around 80%, and in some cases more, of activities that happen in hospitals are process or operations-driven, rather than focused on actual healthcare,” he continues. “And while you’ll see lots of hype around how AI and other new technologies can change the way that patients interact with doctors and the wider healthcare system, the most significant changes have happened at that automation and simplification level - we’ve come to work with, and rely on these innovations on a day-to-day basis without even realising they exist.”

Philips Healthcare Technology AI

The healthcare industry offers great potential for adopting new technologies such as AI, machine learning and automation, particularly with regards to driving down operational efficiencies, enhancing quality of care and patient engagement, as well as reducing cost of care. Some reports have suggested, for example, the market for AI in healthcare could grow by as much as 40% by 2021 to be worth $6.6bn. For Gikopoulos, the use of machine learning technologies in clinical trials has the potential to improve efficiency by as much as 30%, while AI and automation applications have likely already improved the overall efficiency of the healthcare ecosystem by between 10% and 15%. 

“It’s primarily what these technologies do,” he says. “They simplify, standardise and make processes and operations more effective and efficient. Look at how we interact with healthcare. The process 10 years ago was very manual - you had to book appointments yourself, make a call, physically visit hospitals, physicians or clinics, but today that’s changed across the entire payer to provider value chain. Our life is changing on a daily basis on a level that is unprecedented and which, more interestingly, we don’t fully understand. Everything, irrespective of whether we’re a consumer or a patient, a user or a healthcare professional has been made easier by these technologies across the entire ecosystem.”

AI Healthcare Technology

AI in healthcare: technology transformation

According to Gikopoulos, there are four key areas of disruption related to AI and automation in the healthcare industry: intelligent process automation, standardisation of the way in which patient data is classified, optimising patient treatment, and interfacing with patients. 

“Across these four key levers, the payer and provider sector is differentiated by the margins and the relative financial stability of the players within the respective ecosystem,” he explains. “At the one end is public healthcare, public hospitals and primary care, which has to focus on efficiency and effectiveness and is never going to be the most affluent part of the value chain. At the other extreme are the pharma companies and so on, which have margin levels that most other industries are envious of. Across this value chain is where you see a different emphasis on those four trends, and the achieving of efficiencies and effectiveness that AI technology can bring. It also expedites the adoption of technology.

“On the provider side, for example, there have been vast levels of efficiencies and greater effectiveness achieved through automation and the standardisation of information. Something more evident in the big payer-provider interface is around the way in which individuals, consumers, users and patients interact with the healthcare sector. This is where chatbots, intelligent interactive voice recognition (IVR), telematics, avatars and all of those various remote and or non-remote channels of interaction come in.”

Four key AI in healthcare trends

  1. Intelligent Process Automation

“This is the central one,” says Gikopoulos. “Automation, RPA and IPA should really be running all of the various solutions in the background. From the payer perspective, it’s about collecting information, registering people, allowing one-point access to past information and so on; from the provider side, the entire experience in a hospital should be underpinned by automated processes. For pharma companies, for example, automation is used across all functions, from manufacturing and support, through IT, finance and even the running of clinical trials and R&D.”

The key advantage of RPA and IPA is the skipping of the human interface, according to Gikopoulos. Employing the technologies removes the delays or mistakes that humans make from the equation, thus making the entire value chain more efficient from end to end. 

  1. Standardising data 

“The biggest problem across the entire healthcare value chain is being able to call the same thing the same name at every stage of the process, so underlying or diagnosed conditions of a patient, the effect of different treatments - or a lack of treatment - has had in the past and so on,” Gikopoulos explains. “It drives every decision, from what active agents go into drugs and when to dispense then, through to clinical trials, how long people should be hospitalised and even what impact they may have on the greater public and the healthcare system. 

“And yet, it’s a question that just hasn’t been addressed adequately,” he continues. “Information comes from so many sources in healthcare that standardising that information is essential. Further, once you have that standardisation, then you can apply AI to identify the questions behind that information. 

  1. Machine learning

Machine learning is used at the payer/patient or hospital/patient interface to analyse data, often provided by patients, and to provide informed reactions and decisions on that data. “When you start looking at the outcomes of treating patients in different ways depending on when they came in - what symptoms they have, what were their underlying conditions, demographic or social category, then you define a completely different decision tree compared to the static, traditional one that says the customer journey or the patient journey is greet at the reception, triage, send to doctor, have it diagnosed and then send to be treated,” notes Gikopoulos. 

He also cites the importance of machine learning in clinical trial efficiency, adding that it “could be grossly understated in certain cases”. These include, for example, identifying cause and effect in the use of different agents within drugs along the clinical trial value chain, which can work in a much faster and more targeted way with machine learning. 

  1. Patient interfacing

According to Gikopoulos patient interfacing can “completely change the experience that we all have from communicating with this entire value chain”. Examples he cites include using remote channels to contact patients that need a certain treatment after discharge, or the use of telematics to remote diagnose patients. 

“Interfacing isn’t all about chatbots, avatars or cool looking bots that interact with you,” he says. “To a large extent, it’s more to do with the IoT and making us part of the connected world not just for the mundane and potentially also fun parts of our experience and existence, but also the more crucial areas like healthcare. Imagine, for example, if all our machines, computers, phones and so on, instead of using their enhanced abilities to provide greater gameplay, screens, or sound quality, had actually focused on having the type of sensors that allow temperature or blood pressure to be taken, or a person's retina to be scanned for specific diseases. That kind of interface or interaction could, or might still, completely change everything.”

Healthcare Technology

Catalyst for digital technology

Despite the clear benefits to the deployment of AI and automation technology, the pace of adoption has been slow across several industries including healthcare. Gikopoulos attributes this to a risk averse mindset inherent in all of us, and a strong understanding of the underlying technologies and their ability to improve things in the shorter term. “There’s no way of circumnavigating the technology maturity curve, but you can see why sensitive areas like the healthcare sector would be a later adopter of new innovations. 

“Unfortunately, the severity of the coronavirus pandemic is the kind of catalyst that expedites adoption. If you look at the past, it’s little surprise that the biggest changes in the world order have occurred during war time or periods of significant disruption. It brings a situation where people are willing to spend what is needed to achieve results, there is greater alignment between stakeholders to work towards a specific goal, and the emphasis is on getting things done rapidly.”

Future trends

Gikopoulos sees several trends driving future evolution for AI. Distance learning technology and IoT will dominate, he believes, playing a significant role in how people are trained, how they learn, and are on-boarded to organisations. 

“There is also huge potential in how AI and automation can improve the entire supply chain,” he adds “due to what I call ‘working type’ scenarios. This basically means predicting where a supply chain may fail and proactively taking action to rectify before there are any issues. This is especially relevant in the healthcare sector because of the life criticality element that is evident within the sector. For me, the biggest difference will come from the use of machine learning to standardise patient data and provide insights into questions that we never knew existed. It could be a real revolution”

 

Read the latest edition of Healthcare Global to get more of the latest news and insights across every part of the healthcare industry.

Follow us on LinkedIn and Twitter.

Share article

Jun 15, 2021

Why are healthcare networks so vulnerable to attacks? 

Cybersecurity
IoT
healthcarenetwork
cyberattacks
5 min
Elisa Costante from Forescout Technologies gives us the lowdown on how vulnerabilities in the healthcare sector happen, and how to secure them

Forescout Research Labs has published a study on the vulnerabilities impacting the healthcare industry’s connected devices. The research division of Forescout Technologies has published the report as part of its Project Memoria, and it reveals that healthcare organisations are affected five times more by TCP/IP vulnerabilities than any other sector. 

Elisa Costante, a software engineer and Forescout's Vice President of Research, explains why this is and how to prevent it. 

What is Project Memoria? 
Project Memoria aims to improve the security of TCP/IP stacks and understand what  the main security issues are. TCP/IP stacks are a very core component of every network device, whether it's an iPhone connected to the internet, or a robot controlling  the process of manufacturing. If they're connected to the internet they need to have a piece of software controlling communication. 

There are several variants of this software and we're analysing them to understand if they have security bugs or vulnerabilities that if misused by attackers, could lead to disruption of the device itself, and to the network at large. Our goal is to make the industry aware of the problem, and engage with stakeholders as well as the customers. 

Why is healthcare particularly vulnerable? 
This is what the data is telling us. We have a device cloud, which is like a data lake of device information. This device cloud has a lot of information about the devices, like who the vendor is, what the role of the network is, and which vertical this is. We are able to leverage this information, and join it with the intelligence we have from Project Memoria to understand which devices are vulnerable. 

We found that in healthcare there was a huge spike in the number of devices that are vulnerable - as much as  five times more than in other verticals. The reason seems to be because of the number of devices, and because of the intrinsic difficulty of addressing the problem. 

The problem surrounding TCP/IP stacks is that there is not one single vendor that is vulnerable; on average, a healthcare organisation has 12 vendors that are vulnerable. 
Let's say that on average we have 500 devices per healthcare organisation.  Then you need to contact 12 vendors for each of these. These vendors then need to issue a patch to secure the device, and this patch cannot just be automatically delivered and installed in 500 devices. You have to be realistic and think about whether each of the devices  is critical, for example if it goes down will it turn the lighting system off, or stop the MRI machine from working. 

Patches are very complex to deploy. On top of that, the patch needed might not even be available.  That's why we want to understand this problem better  so we can provide solutions. 

How much of the responsibility of keeping a device secure lies with the vendor? 
There are responsibilities that lie with all the different stakeholders, and one of these is  the vendor. There might be multiple vendors involved, which makes it very complex  from a management perspective. 

For instance the device at the end of the chain, which might be an MRI, contains a board that has a connectivity module, and this has one of the stacks that is vulnerable, which could have four different vendors. 

If the vendor responsible for the TCP/IP stack releases a patch, this patch has to go down the chain. We identified chains with a length of six vendors, so you can imagine how complex this is. Some vendors have good hygiene security and some don't because they don't know how to deal with it - they need training. 

This is a new issue related to the software bill of materials, which is being tabled for legislation at the moment to create policies regarding the complexity of the supply chain. We need to shed light on this issue so that legislators can put these policies in place to help with security.  

What can healthcare providers do themselves to stay secure? 
Visibility is important; they need to know what they have in their network. In the case of vulnerable devices they should find out if there's a patch available. If there isn't, because it's an old device for example, but it's still critical to the system, they may want to isolate it so it only communicates with the devices it really needs to. 

Interestingly enough, our research found that most of the healthcare organisations we analysed had a flat network, which means they don't have isolated devices. For instance, a drugs dispensing machine, which you typically find in pharmacies,  is connected to a building automation light system, which is connected to a switch. This is also connected to an IoT sensor device. Why would you have all of them together in the same place? 
The first step is having this information, which often comes as a surprise. Then you can take action; you can segment a network, and if you can't do that you can control the network's access by isolating devices that are risky.

How can Forescout help healthcare organisations? 
Forescout is uniquely positioned to help. We provide visibility end-to-end, which means having a full inventory of devices that includes quite granular detail, so they can know what the operating system is, who the vendor is and so on. Then we enable them to do network segmentation. 

This enables organisations to write policies around how to secure their networks, for example if a device is vulnerable specify which connected devices must be isolated, or which device it must communicate with exclusively. 

Share article