Faculty Bibliography

AIM: To evaluate the performance of multidetector computed tomography (MDCT) in the measurement of endometrial thickness and assessment for endometrial disease. MATERIALS AND METHODS: Seventy-nine MDCT examinations, including sagittal reformats from isotropic data, were retrospectively evaluated for the presence of endometrial abnormality, endometrial thickness, and recommendation for transvaginal ultrasound (TVUS) after CT. The endometrial thickness was measured on sagittal images using two different methods, between the inner-to-inner hypoattenuating stripe, and when visible, between the outer-to-outer hyperattenuating stripe. TVUS performed within 48 h of CT in premenopausal and 1 month in postmenopausal patients served as reference standard. Interobserver agreement for endometrial thickness and abnormalities was assessed using concordance correlation (CC) and kappa statistics. RESULTS: Interobserver agreement for endometrial thickness on sagittal CT images was excellent (CC 0.98), and highly accurate using the inner-to-inner measurement. For determination of abnormal thickening, the positive predictive value and negative predictive value were 67-100% and 99.5-100%. For detection of any endometrial abnormality, the positive predictive value and negative predictive value were 79-90% and 84-95%, respectively. False-negative missed abnormalities included small volume hydrometra, a polyp, and endometrial distortion by a fibroid. CONCLUSION: At MDCT, sagittal reformatted images provide reliable endometrial measurement using the inner-to-inner hypoattenuating stripe and are accurately categorized as normal or abnormal thickness using the same numerical criteria as at sonography.

AIM: To evaluate the degree of variability in clinicians' interpretation of expressions used by radiologists to communicate their level of diagnostic confidence within radiological reports. MATERIALS AND METHODS: Clinicians were solicited to complete a prospective survey asking them to select the approximate perceived level of certainty, expressed as a percentage, associated with 20 expressions used by radiologists to communicate their level of diagnostic confidence within radiological reports. The median and inter-decile range (IDR) were computed for each expression, with a smaller IDR indicating greater reproducibility. Clinicians were also asked questions regarding their attitudes about radiologists' communication of diagnostic confidence. RESULTS: Forty-nine surveys were completed. Median confidence associated with the expressions ranged from 10-90%. Reproducibility of the expressions was variable, as IDR ranged from 15-53%, although a median IDR of 40% indicated overall poor reproducibility. Expressions with relatively higher reproducibility included "most likely", "likely", and "unlikely" (IDR 15-20%), whereas expressions with relatively lower reproducibility included "compatible with", "suspicious for", "possibly," and "can be seen in the setting of" (IDR >/=45%). Only 20% of clinicians agreed or strongly agreed that radiologists consistently use such expressions within their reports. Fifty-five percent of clinicians preferred that diagnostic confidence be communicated as a percentage rather than as a textual expression. CONCLUSION: There was poor reproducibility in clinicians' interpretations of many expressions used by radiologists to communicate their level of diagnostic confidence. Use of percentages to convey diagnostic confidence within reports may mitigate this source of ambiguity in radiologists' communication with clinicians.

Purpose: To compare the recently proposed Prostate Imaging Reporting and Data System (PI-RADS) scale that incorporates fixed criteria and a standard Likert scale based on overall impression in prostate cancer localization using multiparametric magnetic resonance (MR) imaging. Materials and Methods: This retrospective study was HIPAA compliant and institutional review board approved. Seventy patients who underwent 3-T pelvic MR imaging, including T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast material-enhanced imaging, with a pelvic phased-array coil before radical prostatectomy were included. Three radiologists, each with 6 years of experience, independently scored 18 regions (12 peripheral zone [PZ], six transition zone [TZ]) using PI-RADS (range, scores 3-15) and Likert (range, scores 1-5) scales. Logistic regression for correlated data was used to compare scales for detection of tumors larger than 3 mm in maximal diameter at prostatectomy. Results: Maximal accuracy was achieved with score thresholds of 8 and higher and of 3 and higher for PI-RADS and Likert scales, respectively. At these thresholds, in the PZ, similar accuracy was achieved with the PI-RADS scale and the Likert scale for radiologist 1 (89.0% vs 88.2%, P = .223) and radiologist 3 (88.5% vs 88.2%, P = .739) and greater accuracy was achieved with the PI-RADS scale than the Likert scale for radiologist 2 (89.6% vs 87.1%, P = .008). In the TZ, accuracy was lower with the PI-RADS scale than with the Likert scale for radiologist 1 (70.0% vs 87.1%, P < .001), radiologist 2 (87.6% vs 92.6%, P = .002), and radiologist 3 (82.9% vs 91.2%, P < .001). For tumors with Gleason score of at least 7, sensitivity was higher with the PI-RADS scale than with the Likert scale for radiologist 1 (88.6% vs 82.6%, P = .032), and sensitivity was similar for radiologist 2 (78.0% vs 76.5, P = .467) and radiologist 3 (77.3% vs 81.1%, P = .125). Conclusion: Radiologists performed well with both PI-RADS and Likert scales for tumor localization, although, in the TZ, performance was better with the Likert scale than the PI-RADS scale. (c) RSNA, 2013 Supplemental material: http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.13122233/-/DC1.