Ten most commonly performed adult CT examinations in the United States were analyzed from the DIR - head brain without contrast, cervical spine without contrast, neck with contrast, chest without contrast, chest with contrast, chest with pulmonary embolism protocol, chest abdomen pelvis with contrast, abdomen pelvis with contrast, abdomen pelvis without contrast, and abdomen pelvis nephrolithiasis protocol without contrast. For the head exams, lateral thickness dimension was used as an indicator of patient head size. For neck, c-spine, chest, abdomen and pelvis exams, effective diameter was used. Descriptive statistics were calculated for 4 facility characteristics (facility category, location, census region, and average volume of examinations per month) for all the exams included.  Data from over 1.3 million examinations were used to determine median (AD) as well as mean, 25th and 75th (DRL) percentiles of CTDIvol, DLP and SSDE. All analyses were done using SAS 9.3.

Dose data were collected using the same dose management software (DoseWatch and DoseWatch Explore-cloud-based, GE Healthcare) from 11 countries (Finland, Spain, Italy, Luxembourg, France, Belgium, UK, Germany, Sweden, Hungary and Switzerland), 61 CT (7 GE, 3 Siemens, 3 Toshiba and 2 Philips models) for a total of 12817 CT exams (19100 series). For each systems protocol Radlex mapping for the following anatomical region occurred: head (axial and spiral), sinus, chest, abdomen-pelvis and lumbar spine. The estimated European and national DRLs based on collected data (median CTDIvol and DLP) for the investigated RPIDs were compared to European (DataMed II) and national DRLs. The one-sample Wilcoxon signed rank test was used to assess statistical significant differences.

The DCE phantom (Shelley Medical) was imaged on a 320 slice Toshiba Aquillion One CT at a single bed position. Wash-in and wash-out flow to the phantom was set to 100 mL/min. 100 CT volumes were acquired over 360 seconds immediately after contrast (Omnipaque 300) injection at varying temporal sampling frequency between frames (45 × 1.5s, 35 × 3.5s, and 20 × 5s). Imaging was repeated at 80, 100 and 120 kVp with constant 300 mA tube current.  Dynamic scans were retrospectively modified by excluding frames to simulate reduced temporal sampling (1/2, 1/4, 1/5, 1/10, 1/20 of frames). Dynamic images were processed using custom developed software to derive input and output time-attenuation-curves to which a modified 1-tissue-compartment kinetic model with wash-in (K1) and wash-out (k2) parameters were fitted along with transport time delay. Image derived flow estimates were compared to flow meter measured flow rates (ground truth) to determine flow accuracy. 

Using the topograms collected through a dose tracking system (DoseWatch, GE Healthcare) an algorithm was developed to detect the following anatomical regions: head, shoulders, chest, abdomen, pelvis, lower limbs. Using a patient’s anterior-posterior localizer we estimated patient contours, gray-scale intensity profile and bone symmetries, and edges. For each identified anatomical region, the percent of region detected was estimated and the percentage of region irradiated, through comparison with the scanned area. Extracted patient-specific landmarks from DoseWatch, along with exposure parameters, were used to estimate patient-specific organ doses for a sample of patients, to assess the workflow.

Monte Carlo simulation methods were used to estimate effective and organ doses from a low-dose lung cancer screening protocol for a 64-slice CT (Sensation 64, Siemens Healthcare) that used TCM. Scanning parameters were from the AAPM’s Alliance for Quality CT on-line protocols. Ten GSF voxelized patient models that had all radiosensitive organs identified were used to facilitate estimating both organ and effective doses. Predicted TCM schemes for each patient model were generated using a validated method wherein tissue attenuation and scanner limitations were used to determine the TCM output as a function of table position and source angle. The water equivalent diameter (WED) was determined by estimating the attenuation at the center of the scan volume for each patient model. Monte Carlo simulations were performed using the unique TCM scheme for each patient model. All organ doses were tallied and effective doses were estimated using ICRP 103 tissue weighting factors. All dose values were normalized by scan-specific dose-length product (DLP) from 32 cm CTDIvol values that used the average tube current across the entire length of the simulated scan. Absolute and normalized doses were reported as a function of WED for each patient model.

Patients undergoing a single post-venous phase WBCT exam on the 3rd and 2nd generation dual-source-CT (DSCT) (Siemens Somatom Force and Flash, Siemens Healthcare, Forchheim, Germany), as well as on the 64-slice single-source-CT (SSCT) (Siemens Sensation 64, Siemens Healthcare, Forchheim, Germany) were included into the retrospective study. Acquisitions on both DSCT-systems were performed with automated tube voltage selection and automated tube current selection, whereas SSCT protocol included solely the automatic tube current modulation. All images were reconstructed with a 3 mm slice thickness and an increment of 1.5 mm, using the iterative method on both DSCT-systems and filter-back-projection on the SSCT. Commercially available automated dose tracking software (Radimetrics, Bayer Healthcare, Whippany, NJ) was used to calculate the size-specific-dose-estimate. Subjective image quality of axillary and mediastinal lymph nodes, and adrenal glands was rated by two experienced radiologists in a blinded fashion: 5= Excellent image quality with excellent delineation, no blurriness; 4= Good image quality with good delineation, slight blurriness, diagnostically usable; 3= Acceptable image quality with acceptable delineation or blurriness, diagnostically still usable; 2= Insufficient image quality with non-definable delineation or blurriness and not recommended for diagnostic usage. 1= Non-usable image quality.

The aggregation of CT dose data on an institutional level has now become common place using 3rd party dose monitoring products or the ACR DIR. This work describes an attempt to use such a system to compare 13 single phase, adult and pediatric, chest, abdomen, and neuro CT protocols between two academic hospitals. We also explore challenges with dose comparison related to issues with naming conventions and differences in data aggregation.  We hope that other sites can learn from this exercise and use our experience to better evaluate their own CT dose.

This is an IRB-approved retrospective study. Fifty non-contrast head CT’s (NCHCT), non-contrast chest CT’s (ChCT), non-contrast abdominopelvic CT’s (ncCTAP), and contrast-enhanced abdominopelvic CT’s (c+CTAP) performed on GE LightSpeed 64-slice and GE BrightSpeed 16-slice scanners from April 2015 to December 2015 were enrolled. Fifty CT pulmonary angiography (CTPA) studies from the 64-slice scanner and the 12 CTPA studies from the 16-slice scanner during the study period were enrolled.Radiation dose monitoring software, Radimetrics (Bayer, Whippany, NJ), was used to extract the exam dose length product (DLP). The effective dose (ED) was calculated using the standard formula (ED = DLP*k). Radimetrics software provided Monte Carlo-simulated calculations of ED for each exam using a library of phantoms with pre-run Monte Carlo simulations for various scan parameters best matched to the patient exam. The standard formulaic calculation of ED for each exam on each scanner was compared with the Monte Carlo calculation. Bland-Altman plots and paired t-test analysis were performed.