Comparison of Automated Methods to Measure Patient Size for Dose-Monitoring in Computed Tomography
O Christianson1*, D Frush2, E Samei3, (1) ,Durham, NC, (2) Duke University Medical Center, ,,(3) Duke University Medical Center, Durham, NCWE-A-218-8 Wednesday 8:00:00 AM - 9:55:00 AM Room: 218
Purpose: There has been a growing movement to monitor the size-specific dose estimate (SSDE) for patients undergoing computed tomography (CT) examinations. Information on the patient size, however, is not easily accessible in the DICOM headers. Here, we compare three different automated methods to extract the patient size in terms of their accuracy and precision.
Methods: Patient size was determined using three automated algorithms. Method A used a thresholded scout image to determine the patient size. A correction factor of 0.85 was applied to the thickness for the chest region to account for the reduced attenuation of the lungs. Method B converted the central slice of the data set into a water equivalent thickness. Method C converted all slices to a water equivalent thickness and computed an average patient thickness. Testing on a clinical data set of 77 chest and 177 abdomen-pelvis CT exams, the patient size and SSDE were calculated using all three algorithms. Additionally, the local SSDE was calculated using the local CTDI and the patient thickness for each slice. The SSDEs calculated using the three methods were compared with the average of the local SSDE to evaluate the accuracy of each method.
Results: The average absolute error in patient size was 4.2%, 3.1% and 0.0% for methods A, B, and C, respectively. The average absolute error in SSDE was 4.3%, 3.2%, and 0.5% for method A, B, and C respectively.
Conclusions: All three methods accurately calculated SSDE to within 5%. Method A, which uses a thresholded scout image, offers the simplest and most computationally efficient approach to measure patient size for dose monitoring purposes.