Smart Tomo Engine — future X-ray tomography

Send Request

Smart Tomo Engine — sees what others imagine


The X-ray tomography method allows you to see the morphological object structure (a three-dimensional internal image of an object) without destroying it. The whole object or only the necessary (targeted) part is placed in a CT scanner, then X-ray projections are taken at various angles and are transmitted for reconstruction to form a three-dimensional digital image of the scanned volume.


Smart Tomo Engine performs three-dimensional tomographic reconstruction of objects of various nature using a set of X-ray projections. Our solution generates a high-quality image which helps a doctor easily make the right diagnosis and the technologist build a high-quality assessment. In these types of processes a human operator can easily detect and filter out parasitic distortions. The image reconstruction produced by Smart Tomo Engine can be used without additional processing in industrial object recognition systems, simulation modelling of filtration processes in core salvages and drill samples, or can be transferred for physical reproduction to a 3D printer or a molding machine, where distortions are unacceptable.


Smart Tomo Engine is a product that can be adapted to existing CT scanners and synchrotron beamlines to replace the standard software and can be integrated at the stage of developing fundamentally new CT scanners.


Unique reconstruction algorithms in Smart Tomo Engine significantly reduce the computation time without increasing the required computing power. We created the groundbreaking Hough Filtered Back Projection (HFBP) algorithm which achieves unsurpassed tomographic reconstruction speed. The Brady algorithm for fast calculation of the Hough transform is used for back-projection, and the Derish method is used to speed up linear filtering. As a result, the computational complexity of the tomographic image reconstruction is reduced to O (N^2) multiplication operations. A previous leader, the classical high-performance Filtered Back Projection (FBP) algorithm required O (N^2 log N) multiplications to reconstruct an image of size N x N. And the usage of AI algorithms makes it possible to obtain high-quality reconstruction at low radiation dose.


To ensure the best security and privacy, the software Smart Tomo Engine does NOT transfer client’s data to third-party servers, does NOT require internet access, the recognition process is performed in the local RAM memory of the computing device or workstation.


The software Smart Tomo Engine is delivered as a standalone SDK (software developer kit) containing all the necessary API documentation tools (application programmable interface) such as data loading, tomographic reconstruction and saving the results. The documentation is also provided with examples. The SDK is easy to integrate into the software in Objective-C, Swift, C, C++, Java, C#, PHP and Python for a wide range of systems: Windows, Linux-based, macOS, and more.

Send Request

Key features Smart Tomo Engine

We created unique algorithms which avoid reconstruction artifacts (structural distortions that make it difficult to interpret the image correctly). Smart Tomo Engine manages artifacts that occur in the following cases: equipment imperfection or loss of calibration; probing radiation with a broadband spectrum; in addition to the target structure of low optical density, the object contains highly absorbing inclusions; the object was in motion; object does not fit in the detector’s field of view, etc.

Support for non-standard hardware
Smart Tomo Engine is the first tomography software supporting hardware platforms with ARM processors natively (without emulation tools). Experimental evaluation was carried out on a single-board computer Odroid-N2 based on a big.Little processor with quad-core Cortex-A73 and dual-core Cortex-A53 with 4 Gb RAM. The input of Smart Tomo Engine is formed by 432 images with a size of 432х432 each. The output of Smart Tomo Engine is a three-dimensional image with a size of 432x432x432. The algorithm execution time in single-thread mode was 49 seconds; in multi-thread mode was 21 seconds.

If the images are too noisy or distorted to analyze the internal object structure using classical reconstruction methods, Smart Tomo Engine will likely improve the quality of images without any equipment replacement, by using regularized algorithms, novel AI algorithms, methods of analyzing absorption outside the field of view, and methods for the analysis of imperfect spectrum of probing radiation.

Patented algorithms
Smart Tomo Engine includes a large number of patented algorithms that are aimed at solving specific problems in tomography. For example, the HFBP algorithm provides an unrivaled speed of tomographic reconstruction. The FOVEA algorithm allows obtaining high-quality reconstructions, even if the object does not fit into the detector’s field of view. The BH algorithm automatically takes into account and allows to suppress artifacts caused by polychromatic radiation.

Easy to integrate
Our solution is compatible with various input data formats, including the medical DICOM file and network standards, image file formats and scientific HDF5 containers. Modular I/O system Smart Tomo Engine allows easy integration with hardware and information systems that use non-standard data formats.

This software is developed for the reconstruction of X-ray tomographic images in real time with a wide array of applications: from detecting manufactured object defects, to medical diagnostics. Smart Tomo Engine can work with the data obtained using the latest nano CT scanners, medical and industrial scanners or synchrotron centers. Smart Tomo Engine supports CT scanners with all standard scanning schemes: parallel, layer-by-layer, fan, cone, spiral, etc.

Send Request

Please fill out the form to get more information about the products,pricing and trial SDK for Android, iOS, Linux, Windows.

    Our customers


    Tessi use Smart Engines AI-based solutions into its business process services


    SignD Unified Onboarding Platform uses Smart Engines AI-solution to OCR credit cards


    Rosbank has implemented an artificial intelligence solution for the paperwork

    Oman Arab Bank

    Smart Engines helps to implement Digital User Onboarding at Oman Arab Bank

    Send Request

    Please fill out the form to get more information about the products,pricing and trial SDK for Android, iOS, Linux, Windows.