Ukranian scientists are developing an artificial intelligence to help triage people with shrapnel injuries sustained during Russia’s invasion, with the aid of UK researchers using 3D printing to replicate the injuries.
at the National University of Radio Electronics in Kharkiv and his colleagues are developing an AI that can analyse CT scans of shrapnel wounds. The idea is that it will work out the type of material in a wound, its location and whether there is an urgent need to remove it.
Smelyakov hopes the tool will speed up treatment at busy and understaffed hospitals across Ukraine and reveal information about wounds that might not be obvious to the naked eye.
Advertisement
Training such an AI system requires a lot of data, however, and a means to check the AI is accurately assessing what the foreign objects lodged inside wounds are.
Smelyakov’s team has been attempting to gather this information by inserting different shrapnel materials into tissue from pig cadavers as an analogue for the human body, but the process is inaccurate and the movement of the animal tissue as it slumps and sags during CT scans reduces the quality of the data collected.
To improve this data collection, Smelyakov’s team is taking part in the , a project that connects universities in Ukraine with others around the world to aid Ukrainian scientists as part of the war effort.
Instead of relying on pig tissue, at the University of Warwick, UK, and his colleagues are designing and 3D printing models of various human body parts.
These models will be printed from different types of plastic to replicate the densities of human tissue, then the researchers will embed pieces of shrapnel in the models and scan the whole set-up using CT scanners. The Ukrainian team will use these scans as a source of known and replicable data to train the AI.
Gibbons says his team will be exploring a range of shrapnel types, from metal of the sort that would come from a bomb or a missile to wood or glass from exploding buildings.
“Ideally, [the AI] will be able to predict with 100 per cent accuracy what the outcome is for that patient, what treatment they need, is it immediately critical, do they need it removed or can it be left in and they can be put on a ward for a while,” says Gibbons. “People live with shrapnel all their life if it’s not in a serious place. They don’t remove it if they don’t have to.”
Smelyakov says his colleagues are seeing up to three new people every day with shrapnel wounds, and that the biggest problems are finding fragments less than 5 millimetres across and safely extracting bits that are close to arteries. He hopes the AI model will help.
at the University of the West of Scotland says that, with the right training data, an AI could make a big difference in Ukraine’s overburdened hospitals.
“These [types of AI model] will help to reduce treatment time and also save lives, because if you have life-threatening injuries, then it’s always beneficial to have a quick response,” says Ramzan. “It’s actually a very viable approach nowadays.”
