The Learning Curve in Prostate MRI Interpretation: Self-Directed Learning Versus Continual Reader Feedback.

OBJECTIVE: The purpose of this study is to evaluate the roles of self-directed learning and continual feedback in the learning curve for tumor detection by novice readers of prostate MRI. MATERIALS AND METHODS: A total of 124 prostate MRI examinations classified as positive (n = 52; single Prostate Imaging Reporting and Data System [PI-RADS] category 3 or higher lesion showing Gleason score ≥ 7 tumor at MRI-targeted biopsy) or negative (n = 72; PI-RADS category 2 or lower and negative biopsy) for detectable tumor were included. These were divided into four equal-sized batches, each with matching numbers of positive and negative examinations. Six second-year radiology residents reviewed examinations to localize tumors. Three of the six readers received feedback after each examination showing the preceding case's solution. The learning curve, plotting accuracy over time, was assessed by the Akaike information criterion (AIC). Logistic regression and mixed-model ANOVA were performed. RESULTS: For readers with and without feedback, the learning curve exhibited an initial rapid improvement that slowed after 40 examinations (change in AIC > 0.2%). Accuracy improved from 58.1% (batch 1) to 71.0-75.3% (batches 2-4) without feedback and from 58.1% to 72.0-77.4% with feedback (p = 0.027-0.046), without a difference in the extent of improvement (p = 0.800). Specificity improved from 53.7% to 68.5-81.5% without feedback and from 55.6% to 74.1-81.5% with feedback (p = 0.006-0.010), without a difference in the extent of improvement (p = 0.891). Sensitivity improved from 59.0-61.5% (batches 1-2) to 71.8-76.9% (batches 3-4) with feedback (p = 0.052), though did not improve without feedback (p = 0.602). Sensitivity for transition zone tumors exhibited larger changes (p = 0.024) with feedback than without feedback. Sensitivity for peripheral zone tumors did not improve in either group (p > 0.3). Reader confidence increased only with feedback (p < 0.001). CONCLUSION: The learning curve in prostate tumor detection largely reflected self-directed learning. Continual feedback had a lesser effect. Clinical prostate MRI interpretation by novice radiologists warrants caution.

Comparison of interreader reproducibility of the prostate imaging reporting and data system and likert scales for evaluation of multiparametric prostate MRI.

OBJECTIVE: The objective of our study was to compare interreader reproducibility of the recently proposed "Prostate Imaging Reporting and Data System," or "PI-RADS," scale incorporating fixed criteria and a standard Likert scale based on overall impression for prostate cancer localization using multiparametric MRI. MATERIALS AND METHODS: Fifty-five patients who underwent a 3-T prostate MRI examination using a pelvic phased-array coil and incorporating T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging were included in the study. Three radiologists (6, 4, and 1 year of experience) independently scored 18 regions (12 in the peripheral zone [PZ] and six in the transition zone [TZ]) using PI-RADS (range, 3-15) and Likert (range, 1-5) scales, which were based on fixed criteria and overall impression, respectively. Interreader reproducibility was evaluated using the concordance correlation coefficient (CCC), which assesses exact agreement between scores (minimal, < 0.2; poor, 0.2-<0.4; moderate, 0.4-<0.6; strong, 0.6-<0.8; almost perfect, ≥ 0.8). RESULTS: Agreement between experienced readers was strong in the PZ and TZ combined and in the PZ for both the PI-RADS and Likert scales (CCC = 0.608-0.677), moderate in the TZ for the Likert scale (CCC = 0.519), and poor in the TZ for PI-RADS (CCC = 0.376). Agreement between experienced and inexperienced readers was moderate to poor in the PZ and TZ combined for PI-RADS (CCC = 0.340-0.477), moderate in the PZ and TZ combined for the Likert scale (CCC = 0.471-0.497), moderate in the PZ for PI-RADS and Likert scales (CCC = 0.472-0.542), minimal to poor in the TZ for PI-RADS (CCC = 0.094-0.283), and poor in the TZ for the Likert scale (CCC = 0.287-0.400). CONCLUSION: Interreader reproducibility tended to be higher for relatively experienced readers than for less experienced readers and to be higher in the PZ than in the TZ. For the relatively experienced readers, reproducibility was similar for PI-RADS and Likert scales in the PZ but was somewhat higher for the Likert scale than for PI-RADS in the TZ.

Utility of quantitative MRI metrics for assessment of stage and grade of urothelial carcinoma of the bladder: preliminary results.

OBJECTIVE: The purpose of this study was to assess associations between quantitative MRI metrics and pathologic indicators of aggressiveness of urothelial carcinoma of the bladder. MATERIALS AND METHODS: In this retrospective biinstitutional study, 37 patients (28 men and nine women; mean age, 73 ± 12 years) who underwent pelvic MRI including diffusion-weighted imaging (b values 0, 400, and 800 s/mm(2)) and T2-weighted imaging before transurethral resection or cystectomy for urothelial carcinoma of the bladder were identified. Tumor diameter (measured on T2-weighted imaging), normalized T2 signal intensity (to muscle; hereafter labeled normalized T2) and apparent diffusion coefficient (ADC) were measured for all tumors. Mann-Whitney test and receiver operating characteristic analyses were used to identify associations between these metrics and histopathologic tumor stage and grade. RESULTS: Thirty-seven tumors were assessed (mean size, 35 ± 23 mm; range 8-88 mm). At histopathologic analysis, 16 of 37 (43%) tumors were stage T2 or greater and 21 of 37 (57%) were stage T1 or lower, whereas 34 of 37 (92%) were high grade and three of 37 (8%) were low grade. High-stage (≥ T2) tumors showed greater tumor diameter, lower normalized T2, and lower ADC (p = 0.005-0.032) than low-stage (≤ T1) tumors. Tumor diameter and ADC were significant independent predictors of stage (p ≤ 0.043), with their combination giving an area-under the-curve (AUC) of 0.804. High-grade tumors showed significantly lower ADC (p = 0.023) but no significant difference in tumor diameter or normalized T2 (p = 0.201-0.559). AUC for differentiating low- and high-grade tumors was higher for ADC (0.902) than for tumor diameter (0.603) or normalized T2 (0.725). CONCLUSION: A combination of size and quantitative MRI metrics can potentially be used as markers of stage and grade of bladder cancer.

Comparison of CT-based methodologies for detection of growth of solid renal masses on active surveillance.

OBJECTIVE: The purpose of this study is to retrospectively compare 1D, 2D, and 3D measurements on CT for detection of growth of solid renal masses on active surveillance. MATERIALS AND METHODS: Forty solid renal masses measuring at least 1 cm in patients who underwent two CT studies at least 3 months apart were included. Two radiologists independently assessed the studies for interval growth using gestalt visual assessment and 1D, 2D, and 3D measurements. Prospective reports were also evaluated for indications of growth. The summation-of-areas technique was used to calculate volumes of lesions, which served as reference standard in determination of growth. Logistic regression analysis for correlated data was used to compare accuracy of methodologies for detection of lesion growth. Interreader agreement was assessed using kappa coefficients and intraclass correlation coefficients. RESULTS: The accuracy of gestalt visual, 1D, 2D, and 3D assessments for detection of interval growth was 72.5%, 70.0%, 82.5%, and 85% for reader 1 and 77.5%, 70.0%, 90.0%, and 95.0% for reader 2. These differences were significant or nearly significant (p = 0.003-0.054) for the greater accuracy of 2D or 3D measurements than for 1D measurements for reader 1 as well as the greater accuracy of 2D measurements than 1D measurements and 3D measurements than gestalt visual assessment or 1D measurements for reader 2. The accuracy of prospective reports for detection of growth was 65.0%. Reader agreement was fair for gestalt visual assessment (κ = 0.31) and nearly perfect for 1D, 2D, and 3D measurements (intraclass correlation coefficient = 0.97-0.99). CONCLUSION: Our results show that 2D or 3D measurements may be preferable to 1D measurements on CT in assessment for growth of solid renal masses on active surveillance.

Factors Influencing Patients' Perspectives of Radiology Imaging Centers: Evaluation Using an Online Social Media Ratings Website.

PURPOSE: The goal of this study was to use patient reviews posted on Yelp.com, an online ratings website, to identify factors most commonly associated with positive versus negative patient perceptions of radiology imaging centers across the United States. METHODS: A total of 126 outpatient radiology centers from the 46 largest US cities were identified using Yelp.com; 1,009 patient reviews comprising 2,582 individual comments were evaluated. Comments were coded as pertaining to either the radiologist or other service items, and as expressing either a positive or negative opinion. Distribution of comments was compared with center ratings using Fisher's exact test. RESULTS: Overall, 14% of comments were radiologist related; 86% pertained to other aspects of service quality. Radiologist-related negative comments more frequent in low-performing centers (mean rating ≤2 on 1-5 scale) than high-performing centers (rating ≥4) pertained to imaging equipment (25% versus 7%), report content (25% versus 2%), and radiologist professionalism (25% versus 2%) (P < .010). Other service-related negative comments more frequent in low-performing centers pertained to receptionist professionalism (70% versus 21%), billing (65% versus 10%), wait times (60% versus 26%), technologist professionalism (55% versus 12%), scheduling (50% versus 17%), and physical office conditions (50% versus 5%) (P < .020). Positive comments more frequent in high-performing centers included technologist professionalism (98% versus 55%), receptionist professionalism (79% versus 50%), wait times (72% versus 40%), and physical office conditions (64% versus 25%) (P < .020). CONCLUSIONS: Patients' perception of radiology imaging centers is largely shaped by aspects of service quality. Schedulers, receptionists, technologists, and billers heavily influence patient satisfaction in radiology. Thus, radiologists must promote a service-oriented culture throughout their practice.

A Multidisciplinary Approach to Improving Appropriate Follow-Up Imaging of Ovarian Cysts: A Quality Improvement Initiative.

PURPOSE: Incidental ovarian cysts are frequently detected on imaging. Despite published follow-up consensus statements, there remains variability in radiologist follow-up recommendations and clinician practice patterns. The aim of this study was to evaluate if collaborative ovarian cyst management recommendations and a radiologist decision support tool can improve adherence to follow-up recommendations. METHODS: Gynecologic oncologists and abdominal radiologists convened to develop collaborative institutional recommendations for the management of incidental, asymptomatic simple ovarian cysts detected on ultrasound, CT, and MRI. The recommendations were developed by modifying the published consensus recommendations developed by the Society of Radiologists in Ultrasound on the basis of local practice patterns and the experience of the group members. A less formal process involved the circulation of the published consensus recommendations, followed by suggestions for revisions and subsequent consensus, in similar fashion to the ACR Incidental Findings Committee II. The recommendations were developed by building on the published work of experienced groups to provide the authors' medical community with a set of recommendations that could be endorsed by both the Department of Gynecology and the Department of Radiology to provide supportive guidance to the clinicians who manage incidental ovarian cysts. The recommendations were integrated into a radiologist decision support tool accessible from the dictation software. Nine months after tool launch, institutional review board approval was obtained, and radiology reports mentioning ovarian cysts in the prior 34 months were retrospectively reviewed. For cysts detected on ultrasound, adherence rates to Society of Radiologists in Ultrasound recommendations were calculated for examinations before tool launch and compared with adherence rates to the collaborative institutional recommendations after tool launch. Additionally, electronic medical records were reviewed to determine the follow-up chosen by the clinician. RESULTS: For cysts detected on ultrasound, radiologist adherence to recommendations improved from 50% (98 of 197) to 80% (111 of 139) (P < .05). Overmanagement decreased from 34% (67 of 197) to 10% (14 of 139) (P < .05). A recommendation was considered "overmanaged" if the radiologist recommended follow-up when it was not indicated or if the recommended follow-up time was at a shorter interval than indicated. Clinician adherence to radiologist recommendations showed statistically nonsignificant improvement from 49% (36 of 73) to 57% (27 of 47) (P = .5034). CONCLUSIONS: Management recommendations developed through collaboration with clinicians may help standardize follow-up of ovarian cysts and reduce overutilization.