Smart Noise Cancellation: A Groundbreaking Advance in X-ray Image Quality

Reading Time: 7 minutes read

Carestream applies Artificial Intelligence to improve medical image quality.

In some ways, the path to improving X-ray image quality seems pretty straightforward. There are three fundamental components of image quality: noise, sharpness, and contrast. The complexity comes from the fact that the sharpness and the noise are intertwined. Traditional noise reduction introduces blurring–which degrades image sharpness and might remove important anatomical information. Conversely, the more an image was sharpened, the more noise was enhanced. “Denoising an image” is a challenge that medical imaging scientists have been grappling with for some time.

Carestream is pleased to share that our team of imaging scientists have successfully broken this tight, interdependent gridlock and have “separated” image noise from sharpness. We call this groundbreaking technology “Smart Noise Cancellation” or SNC.

How does this advance X-ray image quality? SNC significantly reduces image noise while retaining fine spatial detail [i] –there is no degradation of anatomical sharpness. When SNC is applied, it produces images that are significantly clearer than with standard processing. It also provides better contrast-to-noise ratio for images acquired at a broad range of exposures. And when combined with our SmartGrid software, it promises benefits in gridless imaging where the removal of scatter typically leads to an increase in noise appearance.

We believe this technical breakthrough will be especially beneficial for the high-contrast detail necessary for MSK imaging. For example, in a bone exam it is important to visualize the trabecular bone pattern help rule out a fracture. With its ability to separate anatomical detail from noise, SNC can remove noise while maintaining the fine trabecular structure, easing the interpretation of the exam. 

Smart Noise Cancellation is an optional feature of our ImageView Software, powered by Eclipse, the intelligent platform that serves as the backbone of Carestream’s image processing. The synergy of SNC along with Eclipse results in image quality that is truly remarkable. 

Reader study demonstrates that SNC improves image quality

Objective measurements and subjective ratings demonstrate that SNC processing can reduce noise while simultaneously retaining fine spatial detail. Objective testing demonstrated that SNC processing enables a 2X to 4X noise reduction in flat areas, preserves high frequency sharpness, and improves contrast detail. More specifically:

• A 2x to 4x noise reduction in flat areas is attainable;
• High contrast sharpness is preserved;
• A 10% to 20% improvement in contrast-detail scores on the CDRAD 2.0 phantom is attainable.

Additionally, in a blinded Clinical Reader Study utilizing board certified radiologists:

• 89.5% of all study ratings showed a slight to strong preference for SNC processed images.
• 64% of the diagnostic quality ratings improved based on the RadLex image quality rating scale.
  • 56% of the diagnostic quality ratings improved from “limited” or “diagnostic” to “exemplary”.

In summary, the subjective assessment by board certified radiologists yields a strong signal that Eclipse with Smart Noise Cancellation significantly improves image quality and is strongly preferred.

We understand that the desired level of noise is subjective (e.g. some radiologists expect to see a certain degree of noise in images which assures them that the patient was not over exposed). For this reason, SNC allows a facility full control over the amount of noise cancellation and exposure to minimize dose while meeting their desired image quality.

The Noise Reduction Percentage feature is available to the user on the Image Processing Preference Editor and enables the key operator to set the amount of noise that is removed from 100% (the full noise field) to 50% (half magnitude of the noise field).

Another strong benefit of Smart Noise Cancellation is that it provides the ability to deliver images that are consistently of high quality. This gives imaging facilities more tolerance of acquisition variables – like the skill level of the radiographer – that can possibly negatively impact image quality.

Applying Artificial Intelligence (AI) to improve medical image quality

How did Carestream’s team of imaging scientists achieve this groundbreaking advance in X-ray imaging? The short answer is through a great deal of persistence and by leveraging Artificial Intelligence. Carestream is a leader in using AI for noise cancellation with X-ray images.

Smart Noise Cancellation is the first step in the image processing chain after receiving the raw images from the detector before any other image enhancements. It uses a deep convolutional neural network [ii] (CNN) that is trained to predict a noise-field from an input image. The CNN was trained using low noise/high noise image pairs of clinical patient, cadaver, and anthropomorphic phantom images representative of general radiography.

Let’s break this down into less-technical language. We presented the AI-based CNN a series of the same diagnostic image – one with high noise, the other with low noise. We created the series of high noise anatomical images by injecting noise into it. Our “noise injection technology” is what I like to refer to as “our secret sauce”. Through deep learning on a massive number of images, the CNN learned the levels of noise, the desired level of noise, and then how to separate the noise to produce images that are significantly clearer than with standard processing. 

It was really exciting for our team to achieve this long-pursued target level of noise reduction that delivers improved image quality, increases contrast-to-noise, and ultimately produces easier-to-read radiographs that will not only aid diagnosis, but could help alleviate physician fatigue.

SNC enables imaging professionals to better optimize radiation dose

SNC processing is another important step forward in the direction of ALARA – “as low as reasonably achievable”. The guiding principle of ALARA is that imaging is performed with a dose just high enough to confidently achieve diagnosis.[iii]  SNC enables medical imaging professionals to lower radiation dose without loss in image quality when compared to our standard image processing.  This is especially important in neonatal and pediatric imaging where imaging at the lowest possible dose is critical. The objective measurements present reasonable evidence that dose reduction is possible, potentially up to 2X for detectors using CsI scintillators. A dose reduction study is forthcoming to further explore the capabilities of this new and exciting technology.

Improving image quality is a daily pursuit for Carestream’s team of imaging scientists. We are excited to have overcome this difficult challenge of image denoising, and we are gratified to know that our Smart Noise Cancellation imaging software can help our customers in healthcare take care of their patients.

Carestream imaging scientists who contributed to the development of Smart Nosie Cancellation are Karin Toepfer (Ph. D.), Lori Barski (MS), Jim Sehnert (Ph. D.) and Levon Vogelsang (Ph. D.).

Jim Sehnert (Ph. D.) is an imaging scientists at Carestream Health. He has more than 25 years of experience in X-ray imaging and holds more than 20 patents.

Go Deeper:

• AI Features in Radiology to Adopt Today
• Managing Noise Sources in X-ray Imaging
• Learn About SmartGrid Software

References:

[i] SNC Technical Paper
[ii] Rikiya Yamashita, Mizuho Nishio, Richard Kinh Gian Do, Kaori Togashi, “Convolutional Neural Networks: An Overview and Application in Radiology“, 2018
[iii] Everything Rad; “Managing Noise Sources in X-ray Imaging“