Zoltán Perkó

With a proven track record in algorithm development, deep learning, and multidisciplinary mathematical modelling, rising star Zoltán Perkó basically has the world at his feet. But rather than working for a techno-giant or some high-frequency trading company, he much prefers to apply his skills to the medical field. ‘The nice thing about the medical field is that it has an actual positive impact on society. For someone who works on computational methods, I find that very rewarding.’

Truth be told, Zoltán started his career at TU Delft researching computational methods for the sensitivity and uncertainty analysis of nuclear reactors. When it turned out that TU Delft would soon have its own proton therapy centre at its doorstep (HollandPTC) and that some of the methods he had developed could perhaps also be applied to proton therapy, he was quick to arrange a two-year postdoc position at Massachusetts General Hospital in Boston – one of the leading proton therapy institutes in the world.

Proton therapy

“Traditional” radiotherapy for the treatment of cancer uses X-rays to irradiate the tumour. The advantage of using protons is that it allows much better targeting of the tumour, with less side effects to the normal tissues. ‘The downside is that you have a much higher sensitivity of your treatment to any uncertainties, such as anatomical changes, patient breathing or the tumour shrinking,’ Zoltán says. ‘You need to incorporate these uncertainties in the treatment planning procedure, but this makes planning much more computationally expensive.’ By adapting the approach he used for the uncertainty analysis of nuclear reactors, in combination with some clever coding, Zoltán’s method can assess thousands of treatment error scenarios for a patient in a matter of seconds, providing clinicians crucial information regarding plan quality. ‘It is a huge step forward and we are working on implementing these methods at HollandPTC.’

On the fly and out of the box

Zoltán is also applying his algorithmic skills to shorten the time of creating a proton treatment plan – now taking a few days typically – to only a few seconds. It will allow radiation treatments to be adapted ‘on the fly’, during patient treatment. He also has a few grant ideas up his sleeve for future research. ‘One will be on the optimal use of proton therapy, given that it is an expensive commodity,’ he says. ‘Combining it with traditional X-ray therapy may allow more patients to receive an optimal treatment.’ As a completely out of the box idea, he wants to apply machine learning techniques to help make policy decisions on how to apportion health investments most effectively to various cancer treatment technologies. As Zoltán puts it: ‘The first proposal helps at the patient level, the second at the society level.’

An open mind

He could have stayed in Boston but decided to take an opportunity in the Netherlands. ‘TU Delft, with the proton centre on-site, is in a unique position,’ Zoltán says. ‘You have access to experts in imaging physics, radiation detectors, mathematics, computer science, and so on. And Erasmus MC is close by, with a world-class radiotherapy department and experts in biology and immunology.’ What is best is that clinicians and medical physicists at HollandPTC are very open to his new and unconventional ideas. ‘They do not necessarily understand all the math I apply, but as long as there is evidence that it works, they are willing to let go of what may have been the standard for fifty years.’ Then again, it might have been the cookies and cakes Zoltán likes to bake and bring along that helped convince the clinical staff of his ideas.