They design an algorithm that helps find aneurysms in the brain

American scientists have developed an algorithm that can indicate aneurysms in the brain. The model calls HeadX and marks the possible place of the aneurysm in the brain image.

American scientists have developed an algorithm that can indicate aneurysms in the brain. The model based on the work of the convolutional neural network marks the possible place of aneurysm in the brain image obtained by CT-angiography, which greatly simplifies the diagnosis for radiologists. The study was published in JAMA network open.

The presence of aneurysm (dilation of the blood vessel) in the brain is a very dangerous condition: its rupture can lead to hemorrhage that can lead to various neurological disorders, even death. The most effective way to prevent such consequences is early diagnosis and subsequent treatment to prevent rupture.

Computer tomography angiography is now used as one of the main diagnostic methods because it allows accurate visualization of blood vessels and an badessment of the nature of the blood flow from an image to an image. three dimensions. However, aneurysms can be very small and difficult to examine.

The new algorithm

Now, a team of researchers led by Allison Park of Stanford University has decided to improve the diagnosis of aneurysm in computed tomography angiography using automatic methods. They developed an algorithm called HeadXNet based on the badysis of three-dimensional images using convolutional neural networks. To train it, the researchers took pictures and the results of 611 diagnoses: during the explorations used, the aneurysms and their absence were diagnosed. The resulting model highlighted the likely location of the aneurysm in one of the sections of the red image.

After that, the resulting model was tested on 115 non-partitioned images and presented to 8 qualified radiologists with standard untagged results from CT angiography. The accuracy of the diagnostics with HeadXNet increased significantly (p = 0.01) compared to the usual diagnoses. At the same time, there has been no significant change in the time devoted by the specialists to the badysis of images. (They design a system that converts thoughts into speech)

Despite promising results and great accuracy, the authors of the paper point out that it is still not possible to use the new algorithm as a unique diagnostic method. Any such automatic method should be accompanied by additional evaluation by an experienced radiologist.

Most automatic methods of diagnosing diseases are being developed precisely to simplify the work of the medical staff and not to seize the much-feared job at a given moment.