AI application sees fractures on X-rays

Ariane Esmerian


Software helps ETZ radiologists make the correct diagnosis

Does the patient have a broken bone? To answer this question, approximately one hundred X-rays are taken every day in the Elisabeth-TweeSteden Hospital (ETZ). The Tilburg radiologists are now – as a trial – using AI applications to make the correct diagnosis.

The artificial intelligence application is called BoneView . “We are the first hospital to use this algorithm,” says radiologist Prof. Erik Ranschaert. Before it was put into use at the ETZ, Prof Ranschaert and his colleagues tested the AI ​​application on a selection of 600 recent X-rays. They then had them analyzed by BoneView.


The test result of the BoneView analyzes was positive. Prof Ranschaert: “BoneView turned out to be able to find fractures that the specialists could not see with the naked eye. In the test, BoneView ‘discovered’ previously undetected fractures. Prof Ranschaert: “If you calculate that over a whole year, BoneView can prevent 360 to 370 missed fractures. That is a considerable number.”

However, Prof Ranschaert still calls it a utopia that BoneView can completely take over the assessment from the radiologist. “BoneView missed 3 proven fractures in the test. But overall the results are positive and that means we will be testing BoneView on a larger scale from now on.” Not only the radiologists familiarize themselves with BoneView, but also technologists, emergency physicians, surgeons, and orthopeadics.

How BoneView works

The X-ray is taken and automatically forwarded to a server on which the algorithm is installed. The software will analyze the pictures. That analysis then contains one of the following outcomes: fracture, no fracture or doubt. In the vast majority of assessments, the algorithm is certain of a fracture. In that case, the software shows with a fixed line on the picture where the fracture is located. If there is a suspicion, the fracture will be marked by the software with a dotted line.

Assessment by the doctor

“In such a case of doubt, the specialist still has to check whether or not there is a fracture at this location,” explains Prof Ranschaert. “In any case, the radiologist always has to make the final report, which means that all the pictures are checked. The trial should give radiologists, X-ray technologists, and all other doctors more experience with the reliability of the algorithm.”
The algorithm works fast; BoneView analyses an exam in less than three minutes. By the time the specialist looks at the pictures, BoneView has already done the analysis.

Main goal

The trial in the ETZ will last for three months. Prof Ranschaert emphasizes once again that the AI ​​application does not replace the specialists. “The main goal is to rule out fractures. BoneView offers the specialist an extra pair of eyes that works very precisely around the clock. It provides useful support. The final assessment of the X-rays remains an important task of the specialist, who remains ultimately responsible. Patient safety is guaranteed in all respects.”

During the trial, graduates from Fontys Paramedische Hogeschool Eindhoven investigate the influence of AI on the work process of X-ray lab technicians. The six fourth-year students each have their own research question. Their evaluations together form the graduation project ‘Implementation of AI – pilot BoneView – in the Radiology department’.

End User Group

BoneView is one of the initiatives in the ETZ in the field of AI. The ETZ uses AI as an effective innovation that leads to quality improvement in healthcare. At the same time, AI must also contribute to affordable, sustainable care. An ‘AI end user group ’ is active at ETZ. The main tasks of this group are: supporting healthcare providers who have an idea for an AI application, reviewing all proposed AI projects, building and recording knowledge and expertise in the field of AI, ensuring the requirements of an equipped AI infrastructure, knowledge management regarding AI within and outside the ETZ, inspiring employees and actively involving external partners.

Read article from source