The COVID-19 pandemic has fuelled even more dynamic growth in AI tools and their applications in fields as varied as remote health monitoring, suspicious money transfer identification, and preventive alerts ahead of telecommunications failure.
"This is just the beginning of the revolution. When it comes to the applications of machine learning, the only limit is set by the human imagination", says Marcin Trzaskowski, Data Center Director at Comarch, a company that develops and implements advanced AI solutions based on IBM POWER technology.
Artificial intelligence has been revolutionizing business, and not only business, for several years now. How has the COVID-19 pandemic influenced the process? Have any machine learning tools proved particularly useful lately?
Marcin Trzaskowski: Since the pandemic largely restricted social interactions, and brought into sharp relief the problems we have in healthcare, I think I would have to point to e-health solutions. There are a variety of tools that allow medical professionals to help patients more effectively, rapidly, and from a distance.
In the US, a new solution has been implemented to analyse the condition of patients in hospital admission rooms. Based on test results, for instance, these smart systems are able to determine whether the patient requires immediate help or may wait a while; they can even decide on the course of treatment, although, of course, the doctor will always have the final say. This facilitates a vast decrease in the time needed to analyse individual cases.
Does it also reduce the number of errors?
Absolutely, which is particularly important when the healthcare system is as overwhelmed as it is today, during the pandemic. In some countries, systems have already collapsed. In such circumstances, tools capable of performing rapid and automatic analysis can help take the burden off doctors. Such instruments are bound to revolutionize medicine, which is why we are also working on an original solution of our own.
One of the highly advanced and useful tools offered by our company is Comarch CardioVest, which concerns preventive testing, diagnosis and monitoring of patients with cardiac disease. This is particularly important today, when a medical visit is difficult to arrange, and many people, especially the elderly (not only those with cardiac disorders), are afraid to go to traditional face to face consultations.
We provide them with a tool that can monitor their condition around the clock without any need for physical contact with anyone. The device gathers the data which then is send to central system which uses machine learning algorithms to analyse ECG scans and detect cardiac abnormalities, such as tachycardia, bradycardia, ventricular tachycardia, and ventricular and atrial fibrillation. If a system detects an anomaly, it will notify the doctor, and in the event of an emergency, an ambulance will be dispatched right away.
How does Comarch CardioVest work, and what does it demand of the patient?
The device was designed to be easy to use, even by seniors. For a minimum of two weeks, the patients are asked to wear a special vest that monitors their heart function, collects data 24/7, and sends the gathered information to our system. The vest is equipped with flexible diodes that are really easy to attach, which means patients don’t need to give up daily activities such as going to work or shopping. Since the test takes at least two weeks to complete (and may even take up to one month), the amount of data collected is far too great to be handled by a single doctor.
Our system has been trained to analyse ECG scans and detect anomalies such as atrial fibrillation. It is highly accurate at identifying such disorders, but the final diagnosis, of course, will be up to the doctor who verifies the automatic measurements. What the patients get in return is a sense of comfort and security, of being under the watchful eye of specialists, as well as a shorter response time in the event of a life-threatening situation. Just as with other tools that use machine learning, the main objective is to speed up analysis of large volumes of data and automate the decision-making process based on experience in data analysis.
What about the problem of network overload? Many operators have reported that the pandemic has required a boost to data transfer volume and quality. Is this hindering the advance of AI tools?
No, because, artificial intelligence also makes us better placed to respond to such challenges. Comarch has designed a tool intended for huge telecoms, which enables what’s known as preventive maintenance in very complex telecommunications networks. We can use network data for the split-second detection of irregularities, symptoms of overload, and the likelihood of failure. If part of a network is overloaded, traffic can be redirected to other components, the network can be expanded, or failure management protocols can be activated[AC1] . Our solution allows operators to avoid actual failure, [AC2] protecting them from costly repairs and downtime. Analysis allows them to plan for repairs and expansions in advance.
Comarch Integrated Assurance (which is the name of the solution) also guarantees a convenient and efficient way to manage service staff in the field. After all, operators have transmitters in all sorts of places, often in the back of beyond, which makes ongoing observation and maintenance difficult. Our solution monitors the entire network. If irregularities are detected in any element, it will automatically assign the repair task to a specific employee working in the area. Actions can also be prioritized, so staff know which tasks they should attend to first. And again, analysis of huge data volumes will allow us to act faster and be more accurate, as the need for a repair is reported even before any failure occurs. The solution is very innovative and promising, especially in the context of the rise in IoT services.
In terms of the Internet of Things, does that mean I can soon expect a technical support team to visit my house even before my washing machine breaks down?
Indeed, the most advanced ICT network management tools allow automated maintenance and analysis of function and performance of all network devices. The goal is to detect early anomalies and issue advance warning that a given device may stop functioning.
So I do think that the AI revolution may bring us to a point where a technician will call to let us know that our laundry machine may soon break down – rather than us putting in a call when the device has already ceased to function. The question is: how will users react? I suspect that we are in for a huge mental revolution in this sense.
I have the impression that the pandemic has also accelerated this particular change. For instance, a lot of people who would never have thought of it before have started using internet banking.
That’s true. And that’s a clear change we see in financial markets. There has also been a rise in non-standard transactions and operations, especially at the beginning of the pandemic. A lot of people decided to take their money out of the bank and invest in gold or real estate, supposedly to protect their savings against inflation. In order to help financial institutions identify any suspicious transactions in this whirl of activity, we designed AI-based software to detect financial fraud. It’s very difficult for regular bank employees to do so effectively based on reports, because this would require analysis of really huge amounts of data. Comarch Anti-money Laundering allows suspicious transactions to be detected much more efficiently. As with healthcare solutions, it is always the human professionals – in this case, the bankers – who have the final say.
Artificial intelligence may also be used to assess loan risk. Comarch Loan Origination is a tool that makes its estimate based on behavioural data, a client’s transaction record, and even their social media activity. As a result, the loan decision can be issued much faster and it will make a much more accurate estimate of the payment default risk.
I must emphasize that, in all of the cases I mentioned above, the solutions will only be effective once the neural networks have learned the relevant patterns. Thus, the more training data we can analyse in a given time, the sooner the tool will be available to the client.
What is the key factor that helps achieve that increase in efficiency?
Apart from expert knowledge, it is due to dedicated equipment with extremely high computing power. Comarch has long used IBM POWER servers, which have been designed expressly with similar applications in mind. They are very efficient solutions, integrated with Nvidia graphic cards, used in the fastest computers in the world. For us, it is also important that they are scalable, which means that, whenever we need more computing power, we can simply add a couple of devices and immediately enhance the performance of the whole system.
IBM POWER technology allows us to increase the speed of the training process in neural networks, making them four or even five times faster compared to classical solutions based on x86 architecture. For example, we were able to reduce the time needed to process ECG data from a single patient, monitored over a period of two weeks, by a factor of five.
In order that our clients also benefit from the power of IBM POWER servers, we have launched a cloud service known as Comarch PowerCloud. This makes this state of the art technology much more widely available. It can be used without overpaying for unused resources, and without any fear of running out of computing power at a critical moment.
What makes us stand out in the market is also that Comarch is a Polish company with almost 30 years of experience under its belt, a lot of it in international markets. We are big enough to deliver world-class solutions, but still take a customized approach to each individual client. It is also important to mention that Comarch owns the entire infrastructure on which it operates, as well as the Comarch Data Centers that store client data. This considerably increases the security of our solutions, as clients always know exactly where their data are being held even though they are using a cloud-based solution.