Cheng, P.* University of Southern California, Los Angeles, CA

Address correspondence to P. Cheng ([email protected])

Objective: Although DICOM standards have been developed for structured encoding of CT dose data, legacy CT systems continue to encode patient dose information as screen capture bitmaps not directly amenable to automated retrieval and processing. In addition, each CT manufacturer presents dose information using different nonstandardized tabular formats and fonts. We have built a system that can acquire and decode this heterogeneous CT dose information from the PACS and incorporate it into a database for data mining and monitoring.

Materials and Methods: Four bitmap character fonts were constructed from dose screens of Toshiba, General Electric, Siemens, and Philips CT scanners. An exact glyph matching algorithm was implemented in C# to convert screen capture data into raw text. A separate subsystem queries our Fuji Synapse PACS database for CT screen capture data using HTTP-wrapped SQL statements. Retrieved DICOM dose data is processed using the character recognition algorithm and parsed for dose statistics, including dose-length product (DLP). The dose data and patient demographics are incorporated into an SQLite database, with summary statistics and graphs automatically generated by scripts in the R statistical environment. The entire retrieval, decoding, database, and analysis system runs portably from a USB drive on any PACS workstation, without altering client or server configuration.

Results: Using this system, dose data have been processed from 3171 CT studies of the abdomen performed at University Hospital, Norris Cancer Center, and Healthcare Consultation Center II at the University of Southern California between July 2009 and September 2010. The system has been successfully used to detect dose outliers for further analysis, and to compare dose profiles of different CT machines. Several limitations are the subject of further development, most importantly the normalization of DLP values for body size. Other informatics issues discovered include incorrect or insufficiently specific procedure codes for some CT studies, and problems of separability of dose data for different body regions such as the chest and abdomen.

Conclusion: Heterogeneous screen capture dose data in PACS from different CT scanner manufacturers can be automatically retrieved and processed for retrospective data mining and monitoring.

Laks, M.* Montefiore Medical Center, Flushing, NY

Address correspondence to M. Laks ([email protected])

Objective: Increased international attention to diagnostic radiology radiation dosage has spurred interest in monitoring the patient dosage of radiation in diagnostic radiology departments. CT examination dosage is a major contributory component of diagnostic radiation. A few groups have developed and described software to monitor CT dosages in patients at their institutions, using optical character recognition (OCR). Our institution has a mixture of CT machines from multiple vendors and the methods described by those other institutions are not applicable in our environment, as some vendors do not report radiation dosage for CT series in image format.

Materials and Methods: For our institution we developed an open source application, Ezekiel Dose, and we make it available in open source format. We developed the application on the Linux platform, using Perl, and the open source DICOM libraries dcmtk and dcm4che, open source OCR and image manipulation tools and the open source relational database Postgresql, that enables us to calculate and store monitor radiation dosage in CT at Montefiore Medical Center. Montefiore Medical Center has a heterogenous mixture of CT machines from multiple vendors (GE, Siemens, and Philips) Our Philips equipment does not currently display radiation dosage for CT series in image format as displayed by the GE and Siemens equipment.

Results: Our software uses direct calculation of radiation from DICOM header data or OCR if available to determine radiation dosage for CT examinations. We describe the software we have developed and the technique that we use to obtain the data from the medical center enterprise PACS. Our application resides on a stand-alone PC and communicates with the PACs using the open source dcmtk software toolkit. We have utilized Ezekiel Dose to perform quality assurance functions in our department and assure that the patient doses do not exceed guidelines. Our application is available as an open source application.

Conclusion: Ezekiel Dose is an open source application that we have created to monitor the radiation dosages at our medical center. It is of great help in our departmental radiation dose quality assurance program. The power of providing the software in open source format is to enable collaborative sharing and enable other institutions to utilize, customize, and further develop this software to meet their radiation safety needs. This will help us to achieve our common goal to provide a safe radiation environment for our patients.

Horowitz, J.*; Miller, F.; Casalino, D.; Omar, I.; Yaghmai, V. Northwestern University, Chicago, IL

Address correspondence to J. Horowitz ([email protected])

Objective: The purpose of this study is to determine if educating clinicians about radiation alters CT scan ordering.

Materials and Methods: After a lecture on CT radiation, clinical house staff was surveyed regarding the amount and type of CT scans ordered and use of alternative imaging modalities before and after a lecture and whether they used lecture information to educate patients.

Results: Twenty-one clinical house staff who attended the lecture completed the survey either 2 or 4 months after the lecture. The amount of CT scans ordered after the lecture stayed constant for 90% of respondents and decreased for 10%. The types of CT scans ordered changed after the lecture for 14% of respondents. Thirty-three percent of all respondents reported an increase in use of alternative imaging after the lecture, including 24% increase in MRI and 19% increase in ultrasound. Twenty-nine percent of respondents reported that patients had asked them about radiation since the lecture, and 83% of those who were asked radiation questions used the information learned in the lecture to educate patients.

Conclusion: Most clinicians did not change their CT scan ordering pattern after being educated about radiation. Radiation education of clinicians did not decrease CT scan ordering in 90% of this small group and allowed clinicians to discuss CT benefits and risks with their patients and to choose appropriate CT scan protocols.

Kanfi, A.*; O'Loughlin, M.; Hartford Hospital, Middletown, CT

Objective: To evaluate the CT scan radiation dose in an outpatient setting before and after the implementation of dose-reducing techniques, correlated with radiologist image-quality satisfaction.

Materials and Methods: Prospectively collected radiation dose data, represented as the Dose Length Product (DLP) from six outpatient radiology offices was retrospectively reviewed prior to, and following, the implementation of several additional dose-reduction strategies. The outpatient offices used similar CT protocols, but with various generations of CT scanners. A total of 3,913 studies were reviewed over a period of 3 months. Prior to implementing the dose reduction strategies, the radiologists completed a survey addressing his/her current perceptions on dose-reducing strategies and satisfaction of the quality of the images. Following the recommended alterations to the protocols, the survey was re-administered.

Results: Statistically significant reductions in dose were achieved in the CT scans of the chest (-6.2%, p = 0.045), chest and abdomen (–18.7%, p = 0.031), pelvis (–21.7%, p = 0.023), head (–2.2%, p = 0.032), and lumbar spine (–18.5%, p = 0.012). Additional, but not statistically significant, dose reductions were noted in CT scans of the chest, abdomen, abdomen and pelvis, sinuses, extremities, and CT angiography (CTA) of the chest. The radiologists' satisfaction rate for overall image quality remained unchanged at 90% for both the pre- and post-dose reduction images. There was no statistically significant difference between the image quality satisfaction pre- and post-dose reduction based on the type of examination, with satisfaction scores ranging between 7.2 (CTA chest for pulmonary embolus) and 8.5 (CT cervical spine) (range: 1–10, with 10 representing the highest satisfaction). However, 47% of radiologists felt that the dose reduction may have limited the accuracy in providing a diagnosis in several of their cases.

Conclusion: Implementing additional dose reduction techniques resulted in decreased radiation dose in several examinations. It did not affect overall radiologist image satisfaction; however, it did occasionally reduce their perceived diagnostic accuracy. In the setting of increased public awareness of radiation dose, we believe reduction strategies can be utilized effectively without a significant decrease in radiologist satisfaction.

Khan, A.¹*; Khosa, F.¹; Nasir, K.²; Clouse, M.¹ 1. Beth Israel Deaconess Hospital, Boston, MA; 2. Yale New Haven Hospital, New Haven, CT

Address correspondence to M. Clouse ([email protected])

Objective: MDCT angiography (MDCTA) is a robust, noninvasive method to evaluate coronary artery disease. Image quality and radiation dose are critical concerns in cardiac CT angiography (CTA). The primary aim was to compare the radiation dose and image quality of 320 MDCTA vs 64 MDCTA using prospective gating.

Materials and Methods: The study was institutional review board approved. Using prospective gating, 174 patients were scanned with 320-row MDCT (0.5-mm detectors) and 95 patients with 64-row MDCT (0.6-mm detectors) at a heart rate of 65 beats/min. Three hundred twenty–row MDCT scan parameters were 120 kVp, 400 mA, and gantry rotation 0.350 milliseconds. Sixty-four–row MDCT scan parameters were 120 kVp, 600 mA, and gantry rotation 0.350 milliseconds. Effective dose (ED) estimates were calculated from dose length product and conversion k (0.014 mSv/mGy/cm). Two experienced blinded observers independently assessed image quality of all coronary segments using a 3-point scale (1 = excellent to 3 = nondiagnostic) with the 16-segment model. Discrepancies were settled by consensus.

Results: There were no statistical differences in age, gender, body mass index, and heart rate in the two groups. The mean ED was significantly lower in patients undergoing prospective 320-row MDCTA with a mean of 4.4 (interquartile range [IQR] 3.4–6.2) compared with patients undergoing prospective 64-row MDCTA with 6.2 (5.5–6.9) mSv (p = 0.0001). When data were limited to those with a heart rate below 60 bpm (57% of patients) the median (interquartile range [IQR]) radiation dose among those undergoing 320- vs 64-row MDCTA were 4.1 (IQR, 3.2–5.8) vs 6.2 (IQR, 5.9–6.8) mSv (p = 0.0001). However, among those with a heart rate above 60 beats/min (43% of patients), the respective differences were no longer significant (320-row: 5.2 [IQR, 3.5–8.5] vs 6.0 [IQR, 5.5–6.9] mSv, p = 0.1). Image quality assessed at segment levels was significantly better among those undergoing prospective 320-row MDCTA vs 64-row MDCTA (excellent-diagnostic: 94% vs 86%, nondiagnostic 0.1% vs 3%, all p < 0.0001).

Conclusion: For cardiac CTA, both 320-row MDCT and 64-row MDCT give good image quality but overall radiation dose is less in 320-row MDCT, more in those with heart rate less than 60 beats/min with improved image quality. Every effort should be made to control heart rate in order to minimize radiation dose.

Enterline, J.*; Sarwani, N.; Tappouni, R.; Mahraj, R. Penn State College of Medicine, Hershey, PA

Address correspondence to J. Enterline ([email protected])

Objective: A proposed method to prevent contrast-induced nephropathy (CIN) is the IV administration of fluid prior to and following iodinated contrast injection. Our current hydration protocol involves one hour of IV hydration before scanning and 6 hours after scanning. The objective of this study was to identify areas where the process of referring physician consultation, scheduling, and patient experience could be streamlined and improved upon.

Materials and Methods: The existing CT hydration process was reviewed. An audit of all cases scheduled to receive outpatient hydration from January 1, 2010, to June 30, 2010, was performed. Data collected included the type of CT examination being ordered, admission time, time of scan, time of return to outpatient unit, and time of discharge. After data review, a new proposed workflow for scheduling and performing CT hydration was developed.

Results: Forty-nine patients received IV hydration. There were 24 vascular CT studies, 18 abdominal/pelvic CT scans, and seven head and neck CT scans. The average time from check-in to the outpatient unit to the start of examination was 2 hours 55 minutes. The average length of examination was 28 minutes, and the time from completion of the examination to return to the unit was 24 minutes. Average total time away from the outpatient unit was 54 minutes. The average duration of admission was 8 hours 5 minutes. As a result, proposed improvements include internalizing the process of hydration into the department of radiology, leading to streamlined scheduling, decreased time lost between admission time and time of CT scan, eliminating time lost to patient transportation delays, and decreased patient turnaround.

Conclusion: Internalizing the process of IV hydration prior to and following iodinated contrast material into the radiology department can decrease patient stay by 1 hour. It will also relieve demand on outpatient beds and reliance on hospital transport. In addition, there would be more efficient triage of cases requiring hydration and a streamlined scheduling process. These can lead to increased satisfaction for the referring physicians and clinics, as well as the patient.

Wang, C.¹*; Schopp, J.¹; Petscavage, J.^1,2; Paladin, A.¹; Richardson, M.¹; Bush, W.¹ 1. University of Washington, Seattle, WA; 2. Penn State University, Hershey, PA

Address correspondence to C. Wang ([email protected])

Objective: Assess if high-fidelity simulation-based training is more effective than traditional didactic lecture to train radiology residents in the management of contrast reactions.

Materials and Methods: Prospective study of 44 radiology residents randomized into a simulation versus lecture group based on questionnaire responses. All residents attended a contrast reaction didactic lecture, and baseline knowledge was assessed with a written pretest. Four months later, 21 residents in the lecture group had a repeat didactic lecture, and 23 residents in the simulation group underwent high-fidelity simulationbased training with five contrast reaction scenarios, followed by a written posttest. Two months later, both groups took a delay written posttest and underwent performance testing with a high-fidelity severe contrast reaction scenario graded on predefined critical actions.

Results: There was no statistical difference between the simulation and lecture group scores on the pretest, immediate posttest, or delay posttest. The simulation group performed statistically better than the lecture group on the severe contrast reaction simulation scenario (p = 0.001). The simulation group reported statistically improved comfort in identifying and managing contrast reactions and administering medications following the simulation training (p = 0.04) and were statistically more comfortable than the control group (p = 0.03), which reported no change in their comfort level following the repeat didactic lecture.

Conclusion: High-fidelity simulation-based training of contrast reaction management demonstrates equal results on written test scores, but improved performance during a high-fidelity severe contrast reaction simulation scenario when compared with didactic lecture.

Merritt, C.¹*; Baram-Clothier, E.² 1. Thomas Jefferson University, Philadelphia, PA; 2. American Medical Foundation, Philadelphia, PA

Address correspondence to C. Merritt ([email protected])

Objective: Determination of the interpretation error rates of practicing radiologists by external peer review of randomly selected examinations.

Materials and Methods: Review of more than 12,270 interpretations of 42 radiologists in five group practices in different geographic regions was performed by The American Medical Foundation for Peer Review and Education between 1995 and 2008. For each radiologist, 200–300 randomly selected cases were reviewed by a panel of experienced academic radiologists. Interpretation discrepancies were classified as follows: 0 = minor incidental finding; 1 = miss, diagnosis not expected by a general radiologist; 2 = miss, subtle finding with no impact on care; 3 = miss of apparent finding; 4 = gross error; 5 = overdiagnosis.

Results: The overall significant (class 3 and 4) error rate for all radiologists was 3.32%, with overall error rates for individual radiologists from less than 1% to greater than 8%. Lowest error rates were found for ultrasound (1.68%) and general radiographic examinations (2.56%). Discrepancies were highest for body imaging (4.72%) and neuroradiology (5.85%) examinations. Overcalls were noted in 0.65% of reports. Among the departments surveyed, overall error rates ranged from 2.52% to 4.16%.

Conclusion: Our results agree with previous studies of discrepancy rates measured by external review with overall significant interpretation errors in 3–4% of studies. Of interest is the finding that these values are up to 10 times higher than those reported by internal peer review using the American College of Radiology Radpeer process, suggesting external review as a more objective process for assessment of physician performance.

Agarwal, A.*; Kapoor, T.; Ozonoff, A.; Subramaniam, R. Boston University School of Medicine, Boston, MA

Address correspondence to A. Agarwal ([email protected])

Objective: To determine the disparities in the utilization of PET/CT across different ethnic and socioeconomic groups at an academic medical center.

Materials and Methods: This is an institutional review board–approved, HIPAA-compliant retrospective analysis. A list of patients who underwent PET/CT imaging and a list of patients diagnosed with cancer between August 2004 and September 2009 were obtained from our institutional tumor registry. Patients were classified as black, white, or other (Hispanic, Asian, American Indian, Middle Eastern, and unknown) and their payment method was categorized as Medicare, Private, or Free care (Boston Medical Center HealthNet Plan, Massachusetts Free Care, and Medicaid). Rates of utilization were calculated using the number of patients receiving PET/CT scans in each calendar year as numerator and total number of cancer patients in that year as denominator. Log-linear (Poisson) regression models were used to estimate trends over time while controlling for race and payer status. Data were analyzed using SAS version 9.2, with two-tailed hypothesis tests and a significance level of 0.05.

Results: A total of 4567 eligible subjects with cancer were identified. Overall, 51% were male and 49% were female. The mean age for male subjects was 60.68 years (SD = 13.00) and the mean age for female subjects was 59.70 (SD = 15.36). The racial distribution of patients was 55% white, 26% black, and 19% other. Forty-seven percent of the study population paid by Medicare, compared with 22% using Free Care, and 31% paying with private insurance. Of the sample, 2248 patients (49%) received PET/CT scans. Overall, 50.2% were men and 49.8% were women. The mean age for men was 59.48 years (SD = 12.97) and the mean age for women was 59.22 years (SD = 14.01). The distribution of patients by ethnic group was 51% white, 29% black, and 20% other. The distribution of patients by payment method was 56% Medicare, 19% Free Care, and 25% private insurance. Utilization of PET/CT scans increased at an adjusted annual rate of 17% over the course of the study period. Both black and other groups had significantly higher rates of utilization compared with whites (18.7% [p = 0.0009] and 26.0% [p = 0.0001] higher, respectively). Free Care and private insurance payers had similar rates of utilization (p = 0.70), but Medicare payers had 47% higher rates (p < 0.0001).

Conclusion: Disparities exist in the utilization of PET/CT scans depending on a patient's race and insurance type. This suggests that ability to pay for PET/CT scans may affect access to PET/CT scans.

Friedman, D.; Smith, N. Jefferson Medical College/Thomas Jefferson University Hospital, Wynnewood, PA

Objective: The sentinel effect can be defined as a decrease in services given by providers as a result of a utilization management program. In this project, we evaluated the sentinel effect caused by a prior authorization (PA) process in a radiology benefit management (RBM) company.

Materials and Methods: Using evidence-based guidelines, an RBM company (HealthHelp, LLC) provides real-time, peer-to-peer decision support for physicians ordering high-cost outpatient imaging studies on patients enrolled in national and local health plans. After initial consultation between RBM personnel (level I, customer service representative; level II, nurse) and the provider's staff, studies not meeting appropriateness criteria are referred to an academic radiologist (level III) for further review. The radiologist can approve the study based upon the electronic chart evaluation or call the provider's office for further information; the determination of appropriateness is then made. If a suitable individual is not available to take the radiologist's call, and there is subsequently “no callback” from the provider's office with 48 hours, the study is administratively withdrawn. Studies are not denied by the radiologist. We analyzed the rate of “procedure withdrawn by consensus with the provider” and the rate of “no callback” for a three year interval (January 2007 – December 2009). We also assessed how often a study was reordered after being withdrawn simply due to “no callback.”

Results: A total of 28,120 studies were reviewed during the 3-year interval. There were 3906 procedures “withdrawn by consensus” (13.9%). There were “no callbacks” for 4216 (15.0%). Dividing each year into quarters, the percentage of “no callbacks” did not significantly change over the study period (mean, 14.9%; median, 15.1%, range, 11.5 – 18.3%). Of 4216 studies withdrawn due to “no callback,” 1557 (36.9%) were subsequently reordered within the next 30 days. Hence, 2659/28,120 (9.5%) of all studies were not performed simply due to “no callback.”

Conclusion: The PA process acts as a “sentinel” by imposing a minor barrier to ordering high-cost outpatient imaging studies. In our project, at Level III, almost one out of four studies was not performed due to this barrier. Moreover, this study does not address the sentinel effect at the point of entry into the PA process (Level I). Our data suggest that RBMs can slow the rapid growth in high-cost outpatient imaging.