Smart Engines: AI method of CT diagnostics

The main articles are:

• Computed tomography
• Artificial Intelligence in Medicine
• Artificial intelligence in radiology

2024: Introduction of the AI method of CT diagnostics, which reduces the radiation load on patients

Smart Engines scientists on December 20, 2024 presented technology that significantly reduces the radiation load on humans during computed tomography. The invention is based on the idea of ​ ​ monitoring the scanning process with specialized AI interrupting radiation as soon as all the necessary data is collected. This makes the procedure safer for patients and does not reduce the accuracy of the diagnosis. The United States Patent and Trademark Office issued a patent for the technology on December 17, 2024.

Computed tomography (CT) is used in medicine to diagnose diseases of internal organs without surgery. The tomograph first receives X-rays of the organ from different angles, and then converts them into a digital 3D image with special algorithms. The doctor then examines the sections of this image to examine an internal organ not obscured by the surrounding tissues. The quality of the digital image depends on the number of initial images, as well as on the body weight of the patient, the presence of prostheses and other factors. As of December 2024, a single protocol is used when shooting - the number of images does not depend on the individual characteristics of a person, which can lead to excessive radiation load.

For example, a standard lung CT procedure requires about 10,000 X-rays. Researchers at Smart Engines have set themselves the task of understanding how much it is possible to reduce this number. To do this, they developed a neural network that tracks the quality of the analyzed image directly during the collection of projections. The image is reconstructed from incomplete data. After receiving each partial reconstruction, the AI analyzes the quality of the 3D image, and if there is enough information to make a diagnosis, the procedure is completed ahead of schedule.

As a priority, the development was tested to diagnose pneumonia, one of the most dangerous diseases of the respiratory system. According to researchers and the World Health Organization (WHO), the disease affects about 450 million people each year, of which 4 million die. Among children under 5 years of age, 15% of all deaths are also caused by pneumonia.

The technology was tested on the materials of the open base COVID-CTset, which contains clinical data of 377 patients: both healthy and diagnosed with viral pneumonia. The result of the study showed that the patented method, like the procedure carried out according to the standard protocol, allows achieving high diagnostic accuracy, but using fewer X-rays. So, on average, the approach reduces the radiation load by 15%, but is especially effective in detecting pathology. With an early diagnosis, radiation may be reduced by 25%. At the same time, when examining a healthy person, the procedure minimizes the need for re-examination, although it does not significantly reduce the dose of radiation.

Unlike traditional protocols, the proposed technique adapts to the features of each patient, and this allows more efficient use of the tomographic apparatus and reduce the radiation load on a person. In addition, real-time monitoring minimizes the risks of having to re-scan, "said Vladimir Arlazarov, CEO of Smart Engines, Doctor of Technical Sciences.

The solution opens up new perspectives in medical diagnosis and improves patient safety. The use of AI technologies in combination with the latest developments in the management of the scanning process makes the diagnosis of serious diseases such as pneumonia, cancer, aortic aneurysms, coronary vascular pathologies and others more accurate and effective.