Swedish hospital uses AI to the benefit of patients and staff

Three voices on how AI works

Niclas Feldt is head of Region Skåne’s Department of Digitalisation, IT and MT, Liselott Lading is director of business consultancy at Qinshift and Per Munck af Rosenschöld, is area head of the radiation therapy and radiation physics department at Skåne University Hospital. They represent three different angles on AI.

“AI is nothing new. The concept has existed since the 1950s. When I was at university studying computer and system science around the turn of the millennium, there were courses on AI,” Liselott Lading tells the Nordic Labour Journal.

During this year’s AI Nordic Powwow, organised by Lund University and Skånemotor, she talked about how AI can contribute to more innovation and business, and which AI tools are already available.

Liselott Lading is director of business consultancy at Qinshift, a company that employs around 3,000 people across several countries.

AI policy is not enough

In her job, she meets many customers who are curious about AI. Some in the private sector are also feeling pressured by the competitive situation they believe they might face if they don't jump on the AI bandwagon and do so in the most advantageous way.

“It stems from a sense of urgency, while others are already tired of AI after hearing about it and reading the headlines. But for those of us who have been in the trade for a long time, the media hype is less important. What matters is what happens in the real world, and more and more real stuff is happening,” says Liselott Lading.

It is not enough, however, to develop an AI policy, as some business leaders seem to believe, according to Lading. You need more than that for the technology to function optimally and provide the desired results.

“They believe that by having a policy they have done their job, but in reality they have just addressed one small part of the issue. That is when I bring out the three buckets as I call them, and we can then figure out which of them the company should prioritise. 

The three alternatives put forward by Liselott Lading are:

1. To use external AI, for instance, ChatGPT.
2. To use platforms that the company has already bought and which often contain AI – like the ones offered by Microsoft. 
3. To create AI in-house, and build a company-specific platform.

Efficient AI depends on good data access

Regardless of which alternative the company chooses, the question of responsibility must be solved, points out Liselott Lading.

“AI should be accurate and ethical. It is called responsible AI and involves the various conscious decisions made regarding its use. There is a great difference in the level of responsibility between the marketing department using ChatGPT and using AI as part of a decision-making process.

“For me, it is important that the regulatory framework is in place. You need to know how you want it to work, and that it is about far more than rules. It's about what you want to achieve with your AI. It involves the intentions that should be maintained throughout the entire process, including the actual implementation."

Before you can create AI, you need good access to data. This is where Lading faces the biggest problem among her clients. 

“All AI projects are more or less data projects. So you need to secure the quality and access to the data. There is still an incredible amount of data stuck in silos which cannot communicate. That is often where we have to start our work,” says Liselott Lading.

Another challenge she faces is clients with a short-sighted perspective of AI.

“Many fail to appreciate the long-term effects of AI, which clearly play a monumental role. There are immense possibilities here, including in the healthcare sector,” says Liselott Lading.

Per Munck af Rosenschöld

Cancer patients in Skåne receiving radiotherapy have already experienced a crucial difference because of AI.

“We treat around 5,000 patients a year, nearly everyone with cancer. Half of them receive curative treatment, half receive pain relief,” Per Munck af Rosenschöld, professor of radiation physics and area head of the radiation therapy and radiation physics department at Skåne University Hospital, tells the Nordic Labour Journal.

The radiotherapy department is located at Skåne's University Hospital in Lund, Sus, which also has a branch in Malmö. Both cities are in the South-West of Skåne, around 20 kilometres apart.

Patients who have been diagnosed and referred for radiotherapy primarily come from Skåne, Blekinge, and southern Halland. Radiotherapy is a team effort involving doctors, medical physicists, nurses, assistant nurses, engineers, biomedical analysts, and medical secretaries.

“All the hospital physicians in Skåne are employees here with us at radiation physics as Sus, but they are spread out across all of Region Skåne’s hospitals, so the organisation is a bit like an octopus,” says Per Munck af Rosenschöld.

Implementable methods thanks to AI

Radiotherapy today uses various AI models based on the extensive material documented within Region Skåne, Sweden's southernmost region.

TrueBeam, pictured on the left, is designed to treat cancer wherever it is in the body by delivering precise doses of radiation to tumors. The patient lies on a stretcher under the machine, which can be turned in different directions. Per Munch af Rosenschöld looks into the mylar window that protects the accelerator head against dust and dirt. The radiation comes out of the window when the machine is on, which it is not here. Minna Lerner, hospital physicist, scanned the brain above using three different methods.

“This is input that creates output and implementable methods for radiotherapy.”

One of these methods is sCT, short for synthetic Computed Tomography, commonly known as CT scans.

An MRI scanner, which consists of a very powerful magnet, is used as part of the planning process for radiotherapy. Detailed images of the body's internal structures can be created with radiofrequency radiation.

A CT scan, computed tomography, uses X-rays and a computer to create cross-sectional images of the body. The X-rays are sent through the body from different angles, and the computer combines these images into a three-dimensional picture. 

Both are imaging techniques that visualise the body's internal structures, but they operate in different ways and are used for different purposes. 

Now the department has started creating synthetic CT data with the help of AI, trained on earlier data made up of special MRI sequences and CT scans. Using this AI model and an MRI scan, some of the patients can avoid the need for a CT scan.

“It makes it easier to develop an image because the patient only needs to come in once,” says Per Munk af Rosenschöld, who says the method is being developed as a Vinnova (Sweden’s innovation agency) project in cooperation with a private company in Helsingborg, Spectronic Medical AB.

MRI data is used to map out the patient's tumour and critical organs. Critical organs can include the heart, lungs, spinal cord, and eyes – areas that ideally should not be exposed to radiation during treatment. 

Previously, all of this mapping was done manually by doctors, but these AI models have now been developed in collaboration with RaySearch AB in Stockholm, explains Per Munck af Rosenschöld. 

RaySearch Laboratories AB is a medical technology company that develops innovative software solutions to improve cancer treatment. In May of this year, it was approved for use with TrueBeam. The company assists cancer clinics by collecting, structuring, and analyzing data through various software programs. Photo: RaySearch.

“The gaming industry has actually been helpful since the high demand has driven the development of computers and graphics cards at reasonable cost. These can be used to handle image data and optimise radiation treatments."

The third method is GE AIR, short for General Electric's Adaptive Iterative Reconstruction. This is a technology which improves image quality while reducing the scanning time the patient is exposed to during an MRI scan.

“This model reduces image noise which means patients don’t have to stay still for so long while we scan with our MRI camera,” says Munck af Rosenschöld.

Mixed skills benefit the patient

The department needs a team with diverse skills and professions in order to offer radiation therapy, including doctors, nurses, radiology nurses, engineers, lab-oriented biomedical analysts, chemists, pharmacists, assistant nurses, and medical physicists, a profession that Per Munck af Rosenschöld himself belongs to. 

“We are a technical part of the hospital and it is an advantage to have engineers and hospital technicians who are used to adopting new technology. In many organisations, scientists are far removed from the patients, but not here with us,” says Munck af Rosenschöld.

As a researcher, he supervises four PhD students and four postdocs who work on improving the precision of radiotherapy and reducing side effects. Several of the projects utilise deep learning/AI models. 

Parallel with his work at Sus, Per Munck af Rosenschöld is also keeping in touch with his Danish contacts.

“I worked in Denmark for several years and have many friends and colleagues there, especially at Rigshospitalet in Copenhagen, in Herlev and in Aarhus.” 

Long-term strategy based on maximum benefit

The Nordic Labour Journal visited Region Skåne’s Department of Digitalisation, IT and MT too, where there is also a great need for different skills when the AI strategy is being formulated.

“The most exciting developments are taking place at the intersection of different competencies. AI is so big and complex that it is extremely valuable to gather various perspectives when we draft our strategy. The technological solutions must work in practice,” says Niclas Feldt,  Head of development at Digitalisation IT and MT.

In addition to staff from HR, lawyers, technicians, and economists, experts in data protection and privacy have also been involved in creating the new AI strategy for Region Skåne. It is now complete and awaiting approval from authorities higher up.

“We focus on projects that are crucial to the region, not on anyone’s pet project. In the short term, there are many stand-alone concrete issues, but in the longer term, we need to put together a learning plan. The strategy will be in place for several years and focuses on utility. The emphasis is on practical, everyday AI,” he says.

Johan Åhlén, head of AI, adds a rhetorical question:

“Where do we have the most to gain?

“We must look after taxpayers’ money. That is why we would like to apply AI to areas where it can have a beneficial economic effect as quickly as possible."

Rather than reinventing the wheel, Region Skåne wants to make use of AI solutions others have already had success with. One example being looked at is from the British National Health Service, which has managed to reduce the number of missed appointments by 30 per cent through the use of AI. 

“It involves targeted measures using AI, ranging from extra reminders to offering patients free transport in certain cases. Summing it all up, there are huge gains to be had by preventing patients from missing appointments,” says Johan Åhlén. 

Skåne's University Hospital in Lund, Sus, where the radiotherapy department is located. Photo: Region Skåne © Bengt Flemark

Another example is being able to inform patients about the expected waiting times at hospital emergency departments. For those with less acute conditions, this can mean avoiding unnecessary time spent in the waiting room.

Then, of course, there are examples in Sweden like the radiotherapy department at Skåne University Hospital.

Patient perspective for all

As chief medical officer at the hospital, Ulrika Pahlm works with patient safety and operational security. She sees the potential AI has to help employees avoid repetitive and tedious tasks and to assist with complex analyses.

“We already know why we should be doing this at this time. Now we need to understand the benefits for the entire chain and how we can use AI to help deliver better care for patients.” 

Ulrika Pahlm believes the arrival of AI is perfectly timed and points out that those who make decisions first will succeed sooner – and that is why the time to invest in AI is now. 

“In healthcare, we have more work than we can handle. We need to increase accessibility. But as we know, all change is difficult. We have been doing things the same way for generations. Now we need to think differently and act accordingly,” says Ulrika Pahlm.

Niclas Feldt agrees:

“New technology can create unease and fear, and many believe AI is here to take their jobs. We want to show how people and AI together can harness the benefits."

It is crucial to communicate this to employees and patients, believes Ulrika Pahlm, who often says this about change management:

“We do it for the patients and the patients are us – our friends, relatives, neighbours and colleagues.”

Strong safety measures a hindrance

A stumbling block for AI development in the region is that in Sweden, data sharing between the 21 regions is not permitted. 

“The privacy aspects at the national level offer no guidelines, so each municipality and region spends time and resources on interpreting legislation and often arrives at different conclusions. 

“So far, we cannot share data between the regions. Here, we have incredibly strong safety measures. Finland and Denmark have come much further. We could have too, if only the strong safety measures did not hold us back,” says Niclas Feldt.

The rules do not, however, stop Region Skåne from establishing collaboration for AI development, and here AI Sweden plays a crucial role. 

“They are the spider in the web,” says Johan Åhlén. 

New skills are needed

One precondition for efficient AI is digital maturity among employees, and the Department of Digitalisation, IT and MT at Region Skåne has recently conducted a digital maturity assessment in collaboration with Ängelholm Hospital.

“We need to look at what we need to work with in order to offer the right skills development,” says Niclas Feldt.

The department will also recruit new personnel with AI skills and has set money aside for this.

“We are an attractive employer and expect to receive many applications, both internal and external, for these services,” he says.