Encrypted login | home

Program Information

Relationship Between Pixel Noise and Task-Specific Low Contrast Detectability for Various Patient Sizes in Abdomen CT

J Nute

J Nute*, Y Zhou , A Scott , C Lee , Cedars-Sinai Medical Center, Los Angeles, CA


TH-CD-207B-8 (Thursday, August 4, 2016) 10:00 AM - 12:00 PM Room: 207B

Purpose: To establish a relationship between pixel noise and the statistically defined minimum detectable contrast for various phantom sizes in order to provide a means for dose modulation driven by the task-specific low contrast detectability.

Methods: Seven anthropomorphically-shaped abdomen phantoms (TE01-TE08, CIRS) were employed to mimic a wide range of patient sizes. Each phantom was scanned on a Siemens 64-slice mCT using a range of CTDIvol values (1-38mGy) with an average of 19 dose levels per phantom. Pixel noise was determined from the pixel standard deviations in the uniform phantom background of the images. The low contrast detectability was characterized in the following manner: adjacent slices were subtracted and regions representing uniform phantom background were partitioned to matrices with elements matching the size of the lesion (1.8–10 mm). For each dose level and lesion size, the mean pixel value from each matrix element was computed and the distribution was tested against Gaussian. The minimum detectable contrast (MDC) was defined by 3.29 multiplied by the standard deviation of the aforementioned distribution. We proposed that MDC is proportional to pixel noise and inversely proportional to the lesion size and that the proportionality factor is a function of the phantom size. This model was used to fit the MDC data and the dependency of the proportionality factor on the phantom size was examined.

Results: Pixel noise distributions were found to be Gaussian (p<0.05) in over 99% of cases. The proportionality factor of our model was found to be stable as the phantom size changes (mean = 0.301, coefficient of variation ranges from -2.42% to 5.22%).

Conclusion: The relationship between pixel noise and task-specific MDC has been characterized for a range of patient sizes. This relationship is the first step in establishing dose modulation driven by patient- and task-specific low contrast detectability.

Contact Email: