2333 - _{Translation Accuracy of Scanning Protocols across Various CT and PET/CT Scanner Models and Manufacturers Using Quantitative Methods}

H. C. Kuo¹, S. B. Lim², U. Mahmood¹, C. Della Biancia¹, P. Tamas¹, M. Cebisch¹, S. L. Berry¹, X. Li¹, and L. I. Cervino²; ¹Memorial Sloan Kettering Cancer Center, New York, NY, ²Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY

Purpose/Objective(s): To evaluate the quantitative effectiveness of translating a master protocol from one scanner to a different scanner, even of a distinct model and manufacturer. Materials/

Methods: A master protocol was used in a previous phantom study to match kV, slice thickness, pitch, reconstruction kernel, iterative reconstruction level, and the strength of mAs modulation using automatic tube control between three scanner types from two different vendors. Here image quality per exam was assessed by calculating the global noise index (GNI), which represents the mode of noise histogram per clinical CT image. The patient size was determined by computing the water equivalent diameter (wED) from the patient’s CT scan. A total of 589 patient exams were acquired over four scanners: Two scanners (P₁ & P<sub>2) of the same model from vendor P and two scanners, a PET/CT (S_p) and a CT (S_c), from vendor S. The exams were categorized based on anatomical site: brain, Head & Neck (H&N), breast, lung, abdomen, pelvis, and spine. Protocol matching was evaluated using GNI as the dependent variable at three levels: (1) between the same model and manufacturer (P₁ & P<sub>2); (2) between the same manufacturer but different models (S_p & S_c); and (3) among different manufacturers or models. Student t-tests tested levels (1) and (2) and one-way ANOVA tested level (3). A p-value<0.05 was defined as statistically significant. wED was included as a covariate in an ANCOVA test when a significant difference was found in the t-test.
Results: There was no significant variation in the average GNI (GNI_m) within each subgroup of the exam for P<sub>1 & P<sub>2. However, when comparing S_p and S_c from the same manufacturer, there are noticeable differences in the brain and abdomen subgroups. The mean wED of the abdomen subgroup at the S_c exam (30.5 cm) was 3.4 cm larger than the S_p exam (27.1 cm). A subsequent ANCOVA suggested that there was no significant difference. Additionally, the one-way ANOVA test conducted for (3) revealed that, except for the H&N subgroup, GNI_m values were 1-2 HU higher in the S_p & S_c exams than those conducted with the P scanners. Despite this, the GNI of CT images obtained from scanners of different manufacturers and models remains close to 10±2 HU.
Conclusion: There is no distinction observed between scanners of the same model, nor is there any differentiation by manufacturer. Therefore, a master protocol can be properly translated with an acceptable variation among scanners originating from different manufacturers. Abstract 2333 – Table 1. Summary of GNI_m scanned using various scanners along with the p-value calculated post statistical analysis. “*” indicates that the p-value was calculated using ANCOVA

	GNIm (HU)				p-value
Subgroup	P1	P2	Sp	Sc	P1: P<sub>2	Sp: Sc	P: Sp: Sc
brain	6.2	6.2	8.4	7.4	0.89	<0.01	<0.01
H&N	8.2	9.2	8.8	8.3	0.31	0.38	0.69
breast	7.1	7.3	-	9.3	0.17	-	<0.01
lung	8.8	9.0	10.5	11.0	0.40	0.17	<0.01
abdomen	9.4	9.4	10.9	13.9	0.99	0.01/0.07*	<0.01
pelvis	10.0	10.0	10.9	11.5	0.92	0.18	<0.01