Radiation Dose Reduction: Clinical Reader Study
Reading Time: 7 minutes read
Study results demonstrate Smart Noise Cancellation delivers significant improvement in dose efficiency.
Table of contents
Best practice in medical X-ray imaging employs the principle of ALARA – “as low as reasonably achievable” – for radiation dose management. A consequence of this principle is that imaging is performed with a dose just high enough to confidently achieve diagnosis. As a result, images tend to contain noise that reduces clarity and masks anatomical structures, resulting in degraded image quality.
However, a recent clinical reader study demonstrates that Carestream has broken this gridlock. The clinical study results demonstrate Carestream’s Smart Noise Cancellation (SNC) enables customers to lower radiation dose without loss in image quality when compared to Carestream’s standard image processing. This capability is especially important in neonatal and pediatric imaging where imaging at the lowest possible dose is critical. Read on to learn about the reader study.
About the radiation dose reduction study
Carestream executed an observer study that demonstrated that reduced dose images processed with Enhanced Visualization Processing (EVP) Plus and SNC deliver radiographic image quality that is as good or better than corresponding images acquired at nominal dose processed with our EVP Plus alone. (Our EVP Plus software provides better noise control, sharpness, contrast and density while minimizing artifacts.)
Three US-board-certified radiologists with the specialty of diagnostic radiology evaluated 60 pairs of human clinical and cadaveric subjects captured on 5 detector types (GOS and CsI panels from the DRX1 and DRX Plus family of detectors). Various exams and patient sizes were used in the study. Each detector type had a unique SNC convolutional neural network (CNN).
Two types of reduced dose images were included in the study. Independent cadaver acquisitions ranging from 35-60% dose reduction were used for the Plus family of detectors. Clinical reduced dose images were simulated at prescribed dose reduction levels which were dependent on scintillator type (40% for GOS and 50% for CsI) for all detector types. Noise simulation was employed because of the obvious desire to minimize double exposing of subjects.
Each image pair consisted of a nominally exposed image processed with default EVP Plus processing, including traditional noise suppression that is standard with EVP Plus and the reduced dose image processed with Smart Noise Cancellation with the EVP Plus processing modified such that noise suppression was disabled and the sharpness is tuned to be consistent with the sharpness appearance of the default processing.
Images were randomly placed left or right and were presented in randomly ordered worklists with the treatment conditions blinded to the reader. Each radiologist had a different worklist order. The readers were encouraged to pan/zoom and adjust window width/window level to fully explore and compare the images.
Readers first rated the pair for preference using a 5-point relative scale tied to diagnostic confidence. Next the reader rated the left, then right, image for overall diagnostic quality based upon the RadLex scale. After the ratings were completed, ratings were decoded such that positive values indicate favor for SNC processing.
Reduced dose reader study results
Looking at the results, 93% of the preference ratings reflected no preference or preference in favor of the reduced dose images with SNC processing; and 60% of the preference ratings were slightly or strongly in favor of the reduce dose images with SNC. These results suggest overall preference for the reduced dose images with SNC. Preference ratings by image type suggests that the preference response is similar between cadaver type images and clinical images with simulated noise added. Overall, this data suggests that the reduced dose images with SNC are superior.
Additionally, there were no instances of images rated non-diagnostic. Also, greater than 98% of the reduced dose + SNC images were rated Diagnostic” or higher; and greater than 54% were rated “Exemplary”. It is also important to note that most of the nominal dose images were rated “Diagnostic” whereas most of the reduced dose images were rated “Exemplary”. RadLex responses by image type were similar and image type was not a significant factor of the study.
Radiation dose reduction study conclusions
The clinical reader study delivered the following clear conclusions:
- Reduced dose images processed with SNC are clearly preferred when compared to nominal dose images processed with EVP Plus alone.
- SNC on reduced dose images produces diagnostic quality as good as or better than nominal dose images processed with EVP Plus alone.
- SNC is effective on a wide range of exposures and is equally effective across all detector types.
Achieve a new level of clarity in digital radiography
The strong findings of this dose reduction study complement the results of an earlier iso-dose clinical reader study using a similar methodology. That study demonstrated that images processed with SNC have a quality level that is equivalent or better than images processed with EVP Plus alone. Together these two reader studies support that Smart Noise Cancellation has a number of benefits for medical imaging.
- SNC enables customers to lower radiation dose without loss in image quality.
- SNC provides improved diagnostic quality, preservation of fine detail and better contrast-to-noise ratio for images acquired at clinically nominal exposures.*
- SNC allows adjustment for the amount of noise cancellation and exposure to meet the desired image quality. A site can gradually increase the amount of noise cancellation and/or decrease exposure as desired.
Read the study
Want to learn more? Download the Technical Brief on “Smart Noise Cancellation Processing: Providing a New Level of Clarity in Digital Radiography and a Foundation to Reduce Dose” written by Lori Barski (MS), Karin Toepfer (PhD), Jim Sehnert (PhD) and Levon Vogelsang (PhD).
Learn more:
Smart Noise Cancellation
Eclipse Imaging Engine
Understanding and Managing Noise Sources in X-ray Imaging
The ‘Why” Behind Applying AI in Radiology
Ron Muscosky is a Product Line Manager at Carestream Health. He has more than 30 years of experience in healthcare/medical imaging.
*These statements were verified using Carestream detectors in a reader study performed by board certified radiologists comparing pairwise images taken at nominal dose (CsI ISO 400 speed / GOS ISO 320 speed) and reduced dose (CsI ISO 800 speed / GOS ISO 500 speed) with SNC.
