(Monday, 7/15/2019) 3:45 PM - 4:15 PM
Room: Exhibit Hall | Forum 9
Purpose: Reducing image slice thickness and increment in clinical CT protocols—without changing technique – will yield lower ACR contrast-to-noise-ratio (CNR) measurements, despite generally being preferred by the radiologist. Here, the aim is to determine if phantom and testing methodology changes could better account for clinically relevant partial volume effects and result in CNR measurements that better agree with radiologists’ preferences.
Methods: An anthropomorphic phantom with a 3D-printed insert containing an array of low contrast objects was scanned using acquisition parameters from our routine abdomen protocol, similar to the ACR testing procedure. The objects consisted of groups of 3,6, and 9mm diameter cylinders and spheres at 6 contrast levels, ranging from -2.4 and -51.2HU. Images of different slice thicknesses –1,3,5,7 and 10mm— were acquired and reconstructed at 0.5mm increments. CNR was measured for every slice thickness-object combination. CNR was measured with images centered to the objects and at different positions relative to the object’s center to simulate the arbitrary slice positions possible at different slice increments.
Results: CNR from cylindrical objects monotonically increased with increasing slice thickness, similar to ACR measurements. CNR from spherical objects varied with image thickness and position. For example, 6 and 9mm diameter spheres (-5.3HU) yielded 6% and 17% higher CNR for 5 mm thick images when centered on the object, compared to 3mm images. However for the “worst-case“ scenario of an offset from the object’s center by half the reconstruction increment, data showed that 3mm thick/increment images yielded 29% and 10% higher CNR than 5mm thick/increment images—in agreement with radiologist preference.
Conclusion: Accounting for slice thickness and position variation using different target objects and/or measurement strategy could better reflect low contrast object detection tasks than ACR measurements. Further, these results are germane to strategies accounting for iterative reconstruction-based noise reduction—another limitation of the ACR CNR metric.
Keywords
Taxonomy
Not Applicable / None Entered.
Contact Email
