Artificial Intelligence

“AI-supported software gives us more confidence in cancer care”

In Zagreb, a specialist hospital with 100 employees gives cancer patients from all over the region a chance of survival. AI solutions provide valuable support.

Meike Feder
Published on November 27, 2023

“The most important part of our job is to find the lesion. The biggest mistake would be to not find all the lesions in the first place,” explains Ivo Pedišić, MD, chief radiologist at Radiochirurgia Zagreb, a hospital that specializes in cancer diagnostics and tumor treatment. Ivo Pedišić is looking at a patient’s clinical images displayed on a large screen. His colleagues are with him, also familiarizing themselves with the case. As the tumor board, they spend up to 3.5 hours a day discussing how they can best help their cancer patients. 

“In the tumor board, we decide if we need further diagnostic images or tests to decide on a patient’s individual therapy plans,” explains Domagoj Kosmina, chief physicist for Radiochirurgia Zagreb. “In our hospital, it’s standard practice to combine CT and MR images to really see the patient’s current status and prepare the best possible therapy.” 

The search for cancer is the primary task for the experts. “Our AI-supported software helps us to find lesions, for example in the lung or prostate. This gives us more confidence about whether a lesion is present or not,” says Ivo Pedišić.

A tumor board is a panel of experts who study the patient and their diagnostic findings. The board brings together all the competencies of an oncology center: Experienced physicians discuss the course of the disease, the X-rays, and the examination results. They then produce a therapy proposal for the patient. As well as oncologists and radiologists, the board might also include surgeons from other specialties.

Top row, from left: Jelena Hajredini, Hrvoje Kaučić, MD, Ivo Pedišić, MD Bottom row, from left: Domagoj Kosmina, Vanda Leipold, Ivica Mažuranić

“In the past, I sometimes went home in the evening, and sitting on my couch, I asked myself: Have I missed cancer in this or that case?” says Ivo Pedišić. “Now, I know the computer also didn’t see anything, so there most likely is nothing.” 

His colleague, Professor Ivica Mažuranić, a radiologist and pulmonologist, agrees: “For me, the quality of findings is the most important advantage of AI-Rad Companion. It’s a big time-saving to use it, too. But for me personally, this is less important. It’s the quality.” Radiochirurgia uses AI-Rad Companion Chest CT, Prostate MR and Organs RT.

The clinical images on the large, wall-mounted screen show several tumors that almost completely fill the patient’s left lung. After studying the images, the board looks at the physician’s letter and reviews the patient’s lab results and clinical history. “We’re 100 percent digital. Everyone can always see all test results, clinical images, and appointments for a patient,” says Domagoj Kosmina, explaining the hospital’s cancer information system. Radiochirurgia Zagreb has been successfully using Varian's ARIA oncology information system for many years.

Some of the hospital’s patients have already been diagnosed with cancer at other institutions and have been referred for radiosurgery. “We have a really good cooperation with the public hospitals,” says Hrvoje Kaučić, MD, a radiation oncologist and head of radiosurgery and radiotherapy. This is because Radiochirurgia Zagreb is the only hospital in Croatia that performs radiosurgery. It has been providing this service since it was founded in 2017.

“The word radiosurgery is a combination of radio, meaning radiation to the tumor, and surgery, as it is precisely focused on only the tumor,” says Hrvoje Kaučić. “It’s kind of an ablation with radiation. Compared to radiotherapy with multiple sessions, radiosurgery is usually delivered in a single sitting with a very high dose of radiation.” 

In the tumor board, the experts discuss their recommendations for the individual patients. The outcome of the meeting is a proposal for treatment. This might be radiosurgery, radiotherapy, traditional surgery, chemotherapy, or a combination of different approaches. “We look at every patient individually to decide the best treatment option,” explains Domagoj Kosmina. 

This is also the case for the lung cancer patient, whose images the board members were just looking at. His individual therapy recommendation is radiosurgery: “He was lucky to have this option as, usually, the entire left lung would have had to be removed. With the precise approach of radiosurgery, we had the chance to preserve his lung.”

Now it’s time to start planning the radiosurgery in detail. Here, too, the hospital uses an AI-supported solution: Autocontouring, which involves drawing a colored outline around the organs surrounding  the tumor. Typically, this is done manually, and the specialists have to draw each outline themselves. It’s therefore a very time-consuming (but essential) task in treatment planning. “Autocontouring makes our job easier and faster. We leverage the power of the computer to do the routine stuff,” explains Vanda Leipold, a physicist and expert on the AI-Rad Companion Organs RT software solution. 

“We use supporting AI solutions to prepare the clinical images as standard and are thereby saving 80 percent of the time needed in the past. Our doctors always check the findings and have to correct some, but it’s a huge support. We have a larger number of treatment plans per day,” says Vanda Leipold, adding: “The decisions are taken by a doctor. But this AI support means that—in my estimation—one of our doctors can do the same work as three doctors who aren’t using this solution.” This also means that more patients can start treatment sooner.

Every year, Radiochirurgia Zagreb performs 2,000 radiosurgeries and 600 traditional cancer surgeries. These are in addition to 22,000 accompanying diagnostic imaging examinations. All this happens in a hospital with fewer than 100 full-time employees. “Compared to classical surgery, the diagnostic imaging is much more important for radiosurgery. Since we don’t open up the body, we don't have eyes on the lesion that we’ re treating. Clinical images are how we look inside the body,” says Ivo Pedišić.

At the same time, the high radiation dose means it’s particularly important to position the patient correctly. However, that doesn't make those responsible nervous, says radiation therapist Jelena Hajredini: “I’m confident regarding the positioning for the radiosurgery, because there’ s a lot of technical support in the system.”

JRadiation therapist Jelena Hajredini standing in front of a Varian Edge linac, explains the procedure to a patient.

In her experience, patients are also very grateful for the new solutions: “They’re sometimes afraid that I have to leave the control room, and they tell me that the computer is then watching out for them,” she says. Jelena Hajredini is relaxed and approachable and enjoys devoting her attention to people who are in a difficult situation in their lives. She explains how examinations or treatments are carried out and says: “The patients are the most important part of my job. I love being here for them.” 

She sees some of them over a long period of time. The patient with the lung cancer had to undergo radiosurgery multiple times in a year. A total of 23 tumors were completely removed. He has been cancer-free for a year now, and his lungs are fully functional again.

Domagoj Kosmina says Radiochirurgia Zagreb sees its strategy confirmed by the large number of positive cases it has achieved. “I think what makes us so successful is that we’re embracing new solutions and are innovative,” he adds.

The team at Radiochirurgia Zagreb has been treating locally advanced pancreatic cancer with radiosurgery since 2017 using Calypso extracranial tracking from Varian and AlignRT from VisionRT. They undertook a study to evaluate the potential positive impact of local dose escalation during stereotactic ablative radiotherapy using intrafractional fiducial-based motion management on clinical outcomes. This approach resulted in very favorable one-year local tumor control of 100 percent and median overall survival of 24 months, with no grade > 2 toxicities.  The results were published in the journal Cancers in 2022. 


By Meike Feder

Meike Feder is an editor at Siemens Healthineers. She focuses on stories around patient care.