• Používáte starou verzi internetového prohlížeče. Doporučujeme aktualizaci na nejnovější verzi. Některé funkce portálu nemusí být podporovány.

• Pravděpodobně máte vypnutý JavaScript. Některé funkce portálu nebudou funkční.

A head or neck cancer diagnosis is bad in itself, unfortunately even radiation treatment is not a certain cure, as it does not work for a large proportion of patients. Subsequent surgery after failure of radiotherapy is much more difficult for the surgeon, because wounds are difficult to heal in a person who has already been irradiated. But what if it were possible to predict the effectiveness of radiotherapy before it started? The software that helps doctors and scientists do this was developed at the BUT in cooperation with the CAS and MU.

“DNA has two strands and when both break, the molecule breaks, which is one of the most serious types of damage that can be inflicted on this vital molecule. Ionising radiation is very effective in creating these double-strand DNA breaks, which is why it is used in radiotherapy, where it can effectively kill cancer cells. We can mark and image these DNA breaks under a microscope,” explains Martin Falk, who is the head of the Department of Cell Biology and Radiobiology at the Institute of Biophysics of the Czech Academy of Sciences. He also adds that cells have a variety of mechanisms and proteins involved that, when cells are irradiated, “rush” to the fracture areas to try to repair the damage as quickly as possible. Scientists can image these proteins using a fluorescent dye.

Coloured foci (speckles) of glowing proteins marking DNA breaks that can be observed under a microscope are the main protagonists of the research conducted at the BUT by Tomáš Vičar from the Department of Biomedical Engineering, Faculty of Electrical Engineering and Communication. His task was to create a programme that could automatically find these places, calculate them, and determine as many shape and other parameters as possible. “We tried standard methods, but traditional algorithms failed due to the high variability of the studied foci both over time and between different cell types or even individual samples. In addition, we wanted to work with 3D cell images, which is much more complicated than the 2D situation. Artificial intelligence methods are currently very popular for image processing. Instead of telling the program that it should count a focus if the intensity of the displayed DNA break is greater than a certain value, our algorithm learns this on its own,” Vičar emphasises the importance of artificial neural networks.

Although the team had to feed the software with training data for a year and a half, it is now functional and fully autonomous. In addition to the two proteins studied so far, it even learned to analyse foci of other proteins by itself. Martin Falk of the CAS admits that humans are still more reliable in some respects: “When a person looks at the same images of damaged cells for 15 years, they are probably still more accurate than a piece of software. But when large amounts of data need to be analysed, manual analysis is virtually impossible, and in practice, the speed of obtaining results is particularly important to us, which is huge compared to humans.”

And how much time will doctors save using the programme? “We examine one sample manually for several days to a week, a machine can do it in about an hour. Moreover, it does not take people’s time and the calculation can run at night,” says Tomáš Vičar about one of the solution’s biggest advantages.

The team from the CAS, Masaryk University and the BUT is investigating both the purposeful irradiation in the treatment of tumours and the harmful effects of accidental irradiation on human health. This could happen to astronauts, for example, who are exposed to completely different types of radiation in space than on Earth. However, scientists are also working with catastrophic scenarios such as a nuclear power plant accident. In such a case, blood would be drawn on the spot from persons affected by radiation, processed and microscopic images of the blood cells analysed by computer. Very soon they would find out who needed priority care because they were hit the hardest by the radiation.

However, determining the effectiveness of radiotherapy in treating cancer is a more likely use. The software could examine both pre- and post-treatment patient samples. If the programme shows that the blood sample does not respond to radiation, doctors would not start radiotherapy and would not burden the patient with a treatment that would be useless in that patient’s case. On the contrary, they might start a more effective treatment strategy right away.

Experts have the opportunity to work with samples of real patients thanks to the inclusion of several institutions, which is ensured by Jaromír Gumulec from the Masaryk University Faculty of Medicine: “We have been monitoring patients for about 6 years and therefore know how they responded to radiotherapy. Our question then is simple – does what we predicted using our software match how patients responded to the actual treatment?”