A deformed or completely missing part of the skull is not only a health threatening condition but also an aesthetic problem. Fortunately, modern medicine can replace these parts and allow people to live their lives to the fullest. In order for the skull implants to be as accurate as possible and to fit together like puzzle pieces, a team from the Faculty of Information Technology at BUT focused on this issue.

“We deal with cranial defects, which typically occur during head injuries. In such a situation, the brain may begin to swell and a piece of the skull is removed to prevent excessive pressure. After the swelling subsides, the skull can be readjusted, keeping the original bone and putting it back through surgery, but there are a lot of risks. It needs to be stored aseptically so that it does not get infected or otherwise degrade. Therefore, artificial implants are often used. The skull must be symmetrical, it must join seamlessly, and this is not entirely easy with implants,” says Oldřich Kodym, who deals with this challenge at BUT.

In most cases, designers or doctors design implants manually in 3D modelling programmes, which takes many hours. If the missing part is located on only one side of the skull, it is possible to use mirroring. But what if the injury damages the bone on both sides? “You can find a similar skull as a template, which is available in the patient database, but that is problematic as well. The variability of human anatomy is huge, so the adjustment will take a lot of time. Therefore, we try to automate the modelling part using artificial intelligence,” explains the doctoral student.

It is possible to complete a digitally missing piece of the skull thanks to CT scans and a subsequent virtual model, which will be available to doctors in 3D. And this is the first issue that the BUT IT experts have faced. In order to create a model of a bare skull, it is necessary to remove soft tissues from scans and model the thin orbits correctly. Oldřich Kodym wants to leave the demanding and time-consuming manual work to artificial intelligence: “In order for a programme to learn how to work with scans without human intervention, it is necessary to show them independent work on a large amount of data, which is one of the most expensive commodities in the world today. And when it comes to health information, the possibilities are even narrower.” Fortunately, experts from FIT BUT have enough data thanks to contract research with the Brno software company TESCAN 3DIM and cooperation with the implant manufacturer TESCAN Medical. However, it is still necessary for the results to be checked by a doctor or designer, at least for now. They then produce a mould, a biocompatible material called bone cement is poured into the cavity and it is implanted into the patient, for example after being impregnated with antibiotics.

The moment the experts have a model of a virtually reconstructed skull with the injury imprinted, the second part of the process, and that is the modelling of the implant, which the doctors then surgically implant on the patient, closing off the wound. It was the Brno computer scientists who were one of the first to start using artificial intelligence for this step. “It is not easy to make a shape that exactly complements the skull, because it is necessary to have special edges that allow the implant to be surgically implanted and anchored to the skull. We have feedback that the automatic reconstruction of the skull shape is accurate enough to require only a few adjustments to be used for actual implant production. When an implant is off just by a millimetre or two, it is useless. It must be smooth and it must also preserve the aesthetics of the skull,” Oldřich Kodym also addresses the question of how much to preserve the original form, even if it means, for example, modelling an irregular head shape. Seams and irregularities no longer need to be painted on the implant, as they are not essential for proper function.

The programme from computer scientists at BUT should serve as a supplement to the software that is already in commercial use today. Thanks to them, the whole process of cranial implant design could become faster. The researchers also recently teamed up with a team from the Technical University of Graz, when the Brno group succeeded with its project in a scientific competition organised by Austrian technology. “Because we had a similar approach, we finally put our data together and this year we are already helping with the organisation. I think it is always great to work together, because you are not stuck in your own thoughts and get better insight,” concludes Oldřich Kodym.