New Machine Learning Algorithm Elevates X-Ray Imaging to the Next Level. Will It Replace CT?

15/04/20
ALL articles
By Alexander Khmil
MedTech solution consultant at DataArt, expert in 3D Medical Imaging
Share
New Machine Learning Algorithm Elevates X-Ray Imaging to the Next Level. Will It Replace CT?

A new method has been discovered, which makes it possible to reveal 3D information hidden in conventional 2D X-Ray images. This discovery brings us closer to getting diagnostics results comparable to those achieved through the use of computed tomography but through a more accessible and significantly less expensive method.

Introduction

X-Ray imaging was first used in medicine at the very end of the 19th century. It raised diagnostic potential to a significantly higher level. It was a breakthrough in the development of non-invasive medical imaging.

It was more than 70 years until the next fundamental advancement in medical imaging, which was computed tomography. This new technology made it possible to access a new dimension in diagnostics. To put it simply, 2D images were expanded into a fully volumetric 3D model of a particular part of the human body. This created new opportunities in diagnostics.

If we consider these two technologies, we will see that X-ray imaging is more wide-spread and accessible compared with the more expensive and complex CT, as the following graph demonstrates.  

image x-ray
Global market for medical imaging equipment. Revenues 2016 – 2020 ($ millions)

Global market for medical imaging equipment. Revenues 2016 – 2020 ($ millions)

Despite obvious advantages, the conventional X-ray imaging is not capable of providing the 3D data that CT offers. Is this really the case? In fact, X-ray images contain all of the required 3D data but has lost one of its dimensions. The image is a projection of a 3D object onto a 2D plane. By introducing more X-ray images taken from different angles and employing advanced algorithms, we can retrieve the lost dimension.

The Method

That’s how it works. Let’s imagine that we are looking at a hip joint injury. For that we make two X-ray images, one from the front and another from the side. Later we use a process called stereogrammetry, which is similar to the better known photogrammetry (used in computer games, cinema and advanced navigational systems). But the application of this process is not enough to reconstruct medical volumetric data. To make it work, we need to give the process a boost with some special math and machine learning algorithms. As a result, we get a visual representation of the most likely shape of the hip bones. The outcome of such  medical imaging is called a probabilistic volumetric model. It works using extrapolation from incomplete data. It shows us some features of a bone, which is sufficient for a particular type of diagnostic. For example, we can detect the location of small tumors or the severity of a bone fracture.

image x-ray
Probabilistic Volumetric Model of a femur bone head

Probabilistic Volumetric Model of a femur bone head

In fact, it is not as complex as it seems. The algorithm represents what happens in the head of an orthopedist when he looks at the X-ray images and tries to visualize the overall structure of what’s inside. Our method uses ML to perform this complex task and display the results in the most favorable way.

Benefits

The method that we have described serves to extend the diagnostic potential of X-ray imaging. One of the obvious benefits is, of course, the accessibility of more and higher quality diagnostics. Doctors who have access to regular X-ray imaging will achieve the ability to perform surgery using data in a 3D space. At the same time, this technology is expected to provide diagnostic results comparable to those from computed tomography. Considering that X-ray imaging is significantly cheaper than CT, we conclude that diagnostic costs can be significantly reduced. Finally, if we are able to use X-ray instead of CT in particular clinical cases, patients will be less exposed to radiation, which is now one of the highest priorities in radiology.

Will it replace CT?

At first glance, it looks like our method was intended to replace CT with conventional X-ray imaging. However, that is not the case. Due to different approaches to data processing, AI-powered stereogrammetry and conventional CT are used for different kinds of radiological diagnostic procedures. The method, therefore, was not designed to replace CT but to augment X-ray imaging and create a powerful new tool for diagnosticians.

Sign Up for Updates!

Subscribe now to receive industry-related articles and updates

Choose industries of interest
Thank You for Joining!

You will receive regular updates based on your interests. No spam guaranteed

Add another email address
Read more
Sign Up for Updates!
Choose industries of interest
Thank You for Joining!

You will receive regular updates based on your interests. No spam guaranteed

Add another email address
Welcome
We are glad you found us
Please explore our services and find out how we can support your business goals.
Let's Talk