Exponential technologies applied to image diagnosis

– With the new cloud management systems available, what perspectives on innovation are emerging in the field of digital clinical imaging?

In our area of expertise there has been talk of cloud solutions for some time now, and from a technological point of view they are used in many fields. Certainly, they are starting to take hold in our field too.

The concept of the cloud as seen by the Infrastructure Council is extremely interesting. We view it relatively, although we don’t see it as being suitable for end users. Nor do we see the viability of the concept of platform as a service in cloud terminology either.

We have started to see the introduction of software as a service in our field. Compared with the way in which we worked some years ago with applications installed on computers this is where the possibility of using cloud applications and paying for a service that offers the latest technology becomes really useful.

– Do you think this could result in new digital diagnostic imaging processes based on services that are not currently implemented?

Being able to use cloud software because it is more accessible already allows us to use analytical and image processing techniques that we could implement differently but which would be more difficult to use. Therefore, more advanced diagnostic services will become more powerful, as will quantifying the information contained in the images.

Another concept that is gaining momentum, known as teleradiology or telehemedicine until now, is cloud-based medical services, which consist of second opinions, quantification and data analytics. I think this will really take hold.

– Artificial Intelligence applied to the processing of information obtained from digital clinical imaging is currently a benchmark trend. To what extent do you think it could become a reality?

For the last two or three years, we have been experiencing a boom. Now, everything is AI. In fact, the CT and magnetic resonance imaging machines we currently use have employed AI in image processing for the last ten years.

It is worth noting the difference between Narrow AI and Deep Learning. We have a great deal of experience in Narrow AI, which consists of using algorithms, computerised systems to help interpret and to detect images, and to aid diagnosis.  What will really make a difference is applying more complex algorithms within the area of machine learning where machines can do things that we are unable to do. This will take a bit longer because the clinical circumstances need to be validated. We need to define the problem and the solution, and this will also need to be validated, but I am sure it will take effect. Although not as much as people expect.

– What are the challenges faced by digital clinical imaging in the coming years?

Using software and AI solutions has, and will have, a foreseeable impact both on service provision and the medical profession in general. Not only on those people who work with imaging but on everyone else too.

Even though the service demand is so high, in our case, I doubt that AI tools will end up replacing us. We will become more productive and we will need to change the way we work, but they won’t replace us.

Other more relevant aspects include patients having direct access to treatment decisions, consultation and test results. This is having an impact on the way in which we deliver the service. At the same time, the information aimed at patients has, and will have, a big impact. To explain more about what our work consists of, we currently conduct magnetic resonance imaging, CT, scans and we compile precise reports on specific techniques using these images. What is increasingly valuable is being able to use this information in conjunction with other data, from the lab or pathological anatomy for instance. Therefore, the concept of comprehensive reports is gaining potential.

– Could you give us some examples?

Two problems that affect our department, the health service and medical professionals in general are: the time we take to do things and variability in the way we do them. One doctor can do things very differently from another, often without any consequences for the patient but, in theory, it can result in service inefficiency.

Using AI techniques helps us to reduce time and variability so that we can standardise processes and quantify results. Therefore, we focus on quantification rather than art, which is also good but its over-reliance can jeopardise efficiency.

A case in point is stroke sufferers, when time is crucial. We always say that “time is brain”. The time we take to diagnose a patient until we start to treat that patient is vitally important. With AI we can obtain data much faster to aid diagnosis, which will reduce variability and enable us to make better recommendations and to provide more accurate treatment.

– At the TICSalut Foundation, we are promoting an Innovation Observatory in Digital Clinical Imaging. How do you think this observatory can help an organisation like yours?

It can help us manage the explosion of solutions and experiences. Colleagues from a vast range of medical backgrounds and from all around the world are promoting new solutions and scenarios for applying analytical image quantification techniques based on AI. It is impossible for professionals to manage this deluge of information, especially when we want to analyse its clinical application and reliability or to exchange experiences with other professionals who are trying new solutions.

I think that having an Innovation Observatory like the one being promoted by the TIC Salut Social Foundation is not only relevant but is also crucial so that our work may continue and we can lead the application of these technologies.