How you view a clinical image-be it head-on or from an angle--could determine whether you find an abnormality.
How you view a clinical image--be it head-on or from an angle-could determine whether you find an abnormality.
"The detectability of an abnormality depends on the angle at which you are viewing the display," said Aldo Badano, visiting scientist with the division of imaging and applied mathematics at the FDA's office of science and engineering laboratories.
At the 2005 RSNA meeting, Badano illustrated how grayscale levels are displayed differently when viewed from various angles. With some radiology viewing stations incorporating two, four, and even six displays, the viewing angle becomes even larger, exacerbating the problem, he said.
"When images are viewed along certain directions the contrast is completely washed out," he said.
Badano and colleagues conducted a human observer study in which 11 trained participants viewed a total of 350 synthetic image pairs. Observers viewed the images at zero, 30-, and 45-degree angles from a 5-megapixel monochrome LCD monitor. Researchers then added Gaussian signals to the images at 4, 8, and 12 gray level signal amplitudes.
At the perpendicular angle (the zero-degree angle) and at a signal amplitude of four gray levels, readers correctly identified the target object 80% of the time. At a 30-degree angle, accuracy dropped to 62% and at 45 degrees, readers were basically guessing, identifying the target about 50 percent of the time, Badano said.
However, by increasing the Gaussian signal to 8 gray levels, accuracy at the 30 degree angle could be improved to nearly 80%. At the 45-degree angle, Gaussian signal levels needed to be increased 3 times the base level to achieve a similar improvement in detection, Badano said.
Not only did performance decrease as the viewing angle moved away from normal, Badano said, but the time needed to make a decision increased. Readers took from twice as long to seven times as long to record their decisions when they viewed the images at angles greater than zero.
Manufacturers can address this problem in a number of ways, Badano said, but at this point radiologists need to know how to better measure and characterize these effects.
MRI-Based AI Radiomics Model Offers 'Robust' Prediction of Perineural Invasion in Prostate Cancer
July 26th 2024A model that combines MRI-based deep learning radiomics and clinical factors demonstrated an 84.8 percent ROC AUC and a 92.6 percent precision-recall AUC for predicting perineural invasion in prostate cancer cases.
Breast MRI Study Examines Common Factors with False Negatives and False Positives
July 24th 2024The absence of ipsilateral breast hypervascularity is three times more likely to be associated with false-negative findings on breast MRI and non-mass enhancement lesions have a 4.5-fold likelihood of being linked to false-positive results, according to new research.
Can Polyenergetic Reconstruction Help Resolve Streak Artifacts in Photon Counting CT?
July 22nd 2024New research looking at photon-counting computed tomography (PCCT) demonstrated significantly reduced variation and tracheal air density attenuation with polyenergetic reconstruction in contrast to monoenergetic reconstruction on chest CT.