Smart Engines: Controlled tomographic reconstruction technology

The main articles are:

• Computed tomography
• Artificial Intelligence in Medicine

2025: Developing ultrafast algorithms for the next generation of AI-driven CT scans

Specialists Smart Engines have developed a family of high-performance ones. algorithms computed tomography Compared to those previously known, they are significantly faster at reconstructing 3D tomographic images. The company announced this on January 23, 2025. The development of scientists will create a last-generation tomograph for medicine high-tech industries and scientific research. The Bureau patents trade and Stamps USA on December 17, 2024 issued a patent for a general method of building such. algorithms

Computed tomography (CT) is a common method of medical diagnosis and quality assessment in production. First, the tomograph receives X-rays of the object from different angles, and then converts them into a digital 3D image with special algorithms. After that, the doctor examines the sections of the image to consider the internal organ, not blocked by the surrounding tissues. In industrial production, a specialist studies the visualization of the internal structure of an object - for example, an engine or a microcircuit - to find possible defects.

Meanwhile, traditional methods of reconstructing tomographic images are a complex computational process, and the rapid increase in the resolution requirements of computer tomographs involves processing an impressive amount of data. There are two ways to solve the problem: either spend more money on heavy-duty processors, or optimize the process of converting data from a tomograph to an image.

The group of the fastest known tomographic reconstruction algorithms involves two stages: the so-called filtering and reverse projection. Reverse projection is most computationally costly and can take up to 99% of the reconstruction time. The optimization proposed by Smart Engines scientists is based on dynamic programming ideas, and reduces the number of operations required for reverse projection by hundreds of times. As a result, the "bottleneck" for the first time was the filtration stage. To accelerate it, the Derish method was used, based on the replacement of direct convolutions with specially optimized recurrent filters. As a result, scientists managed to significantly surpass all previously known methods in performance and increase the efficiency of processing large amounts of data.

In medicine, the discovery will significantly speed up the diagnosis using CT, which is extremely important in emergency situations, when when fighting for a person's life, the bill goes for minutes. To achieve the greatest effect, the new accelerated algorithms work in conjunction with another patented development of Smart Engines - a neural network that monitors the quality of the analyzed image directly during the collection of projections. 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.

Also, this method significantly reduces the processing power requirements of processors in tomographs. Thanks to this, it is possible to reduce the cost of equipment, which means that as many hospitals as possible can be provided with tomographs, including in the regions. At enterprises, this method of tomography will be especially useful for quality control. Studying scanned data on an industrial scale during flaw detection of parts will become much faster and cheaper.

Our team was able to develop algorithms that solve several problems at once: increasing the speed, accuracy and efficiency of tomographic reconstruction. The technology has great potential for implementation in various, and industries economies obtaining a foreign patent confirms its worldwide importance. This demonstrates a high level of competence of Russian scientists in the field of image processing, - said the CEO of Smart Engines, Doctor of Technical Sciences. Vladimir Arlazarov

2023: Obtaining a U.S. patent

On July 6, 2023, Smart Engines announced that it had patented experimental technology in computed tomography in the United States. The development allows you to reduce the dose of radiation when diagnosing patients with COVID and other diseases. Smart Engines intends to integrate the technology into its products and promote them both in Russia and abroad.

The image illustrates the dependence of the quality of the CT image of the lungs on the number of projections: the more projections, the better the image

As reported, tomographic examination using the controlled tomographic reconstruction technology described in US 11,663,757 of May 30, 2023 is fundamentally different from standard CT procedures. If in the classical protocol the number of images and the time of their registration is clearly regulated and cannot be changed after the start of the measurement, then Smart Engines scientists propose to consider tomographic reconstruction as a process with a tracked result.

Artificial intelligence analyzes the results of reconstruction in real time after obtaining the next series of projections and, when the necessary quality of the reconstruction result is achieved, decides to stop scanning. The image illustrates the dependence of the quality of the CT image of the lungs on the number of projections: the more projections, the better the image.

Let's imagine that 360 projections are recorded from the object during tomographic measurement using a standard protocol. We suggest registering five projections first, then, if the image quality is not good enough, another five and so on. The question arises: why register additional projections and spend time on this, if the image quality is already sufficient. told Konstantin Bulatov, one of the authors of the invention, senior researcher programmer - Smart Engines Candidate of Technical Sciences

Smart Engines technology solves several problems at once. On the one hand, it allows you to reduce the dose load - this is important in the context of medicine.

The technology can be applied in any field of medicine where tomographs are used. For example, to build a model in prosthetics, to determine brain neoplasms or injuries, to tomograph the lungs. This technology is especially promising in the process of observing the dynamics of treatment, when a patient is prescribed several CT procedures in a row - with COVID-19 or with bone fusion after a fracture.

{{quote 'author
= told Marina Chukalina, head of the tomography department of Smart Engines, candidate of physical and mathematical sciences' The problem of optimizing the dose when observing the dynamics of the course of coronavirus in a patient or the dynamics of recovery after a previous illness is extremely acute. Since the CT examination in this case is carried out repeatedly and the dose load is summed up, then, of course, I would like to the time of each measurement from the measurement chain the patient received a lower dose.}}

Smart Engines technology solves another important task - it significantly reduces the time of the tomographic experiment. This is important for industrial tomography, since, for example, in nanotomography, the time of the experiment - studying one industrial object - can last tens of hours for July 2023.

{{quote 'author
= Marina Chukalina explained' If one dimension lasts so long - therefore, it is expensive and it is not always advisable to carry it out, because you can find a cheaper way to analyze the internal structure. Application our technology saves time and therefore reduces the cost of a single nanotomographic measurement.}}

The method of controlled tomographic reconstruction patented in the USA was developed by scientists, candidate of technical sciences Konstantin Bulatov, Director General of Smart Engines Candidate of Technical Sciences Vladimir Arlazarov, Head of the Tomography Department Smart Engines Candidate of Physical and Mathematical Sciences Marina Chukalina, technical director, candidate of physical and mathematical sciences Dmitry Nikolaev and candidate of technical sciences Alexei Buzmakov back in 2020 while working on the first version of the Smart Tomo Engine software product.

The SDK can be installed on operated laboratory tomographs to replace standard software or in other hardware and software tomography complexes.

As of July 2023, the Smart Tomo Engine includes three components - a collection of projections and their transfer to the reconstructor, the reconstructor itself and the technology for analyzing results. We are working to include a control unit in the Smart Tomo Engine. It is required to conduct classical experiments in tomography using patented technology. added Vladimir Arlazarov

Smart Engines, which has been developing software for tomographs for more than four years and has invested over 200 million rubles in this area, intends to integrate the technology into its products and promote them both in Russia and abroad.

As of July 2023, the market for CT scans in the world is growing: annually it increases by an average of 7.10%. It was valued at US $41.7 billion in 2021 and is expected to reach US $77.3 billion by 2030. The leader among the regions is North America, which accounts for over 30% of global revenue.

The CT market is expected to grow rapidly due to an increase in the number of projects in medicine and the development of a new market segment - industrial tomography. Vladimir Arlazarov, CEO of Smart Engines Ph.D.

This US patent was the first CT invention obtained by Smart Engines scientists in the States.