Multicenter Evaluation of Multiparametric MRI Clear Cell Likelihood Scores in Solid Indeterminate Small Renal Masses.

Background Solid small renal masses (SRMs) (≤4 cm) represent benign and malignant tumors. Among SRMs, clear cell renal cell carcinoma (ccRCC) is frequently aggressive. When compared with invasive percutaneous biopsies, the objective of the proposed clear cell likelihood score (ccLS) is to classify ccRCC noninvasively by using multiparametric MRI, but it lacks external validation. Purpose To evaluate the performance of and interobserver agreement for ccLS to diagnose ccRCC among solid SRMs. Materials and Methods This retrospective multicenter cross-sectional study included patients with consecutive solid (≥25% approximate volume enhancement) SRMs undergoing multiparametric MRI between December 2012 and December 2019 at five academic medical centers with histologic confirmation of diagnosis. Masses with macroscopic fat were excluded. After a 1.5-hour training session, two abdominal radiologists per center independently rendered a ccLS for 50 masses. The diagnostic performance for ccRCC was calculated using random-effects logistic regression modeling. The distribution of ccRCC by ccLS was tabulated. Interobserver agreement for ccLS was evaluated with the Fleiss κ statistic. Results A total of 241 patients (mean age, 60 years ± 13 [SD]; 174 men) with 250 solid SRMs were evaluated. The mean size was 25 mm ± 8 (range, 10-39 mm). Of the 250 SRMs, 119 (48%) were ccRCC. The sensitivity, specificity, and positive predictive value for the diagnosis of ccRCC when ccLS was 4 or higher were 75% (95% CI: 68, 81), 78% (72, 84), and 76% (69, 81), respectively. The negative predictive value of a ccLS of 2 or lower was 88% (95% CI: 81, 93). The percentages of ccRCC according to the ccLS were 6% (range, 0%-18%), 38% (range, 0%-100%), 32% (range, 60%-83%), 72% (range, 40%-88%), and 81% (range, 73%-100%) for ccLSs of 1-5, respectively. The mean interobserver agreement was moderate (κ = 0.58; 95% CI: 0.42, 0.75). Conclusion The clear cell likelihood score applied to multiparametric MRI had moderate interobserver agreement and differentiated clear cell renal cell carcinoma from other solid renal masses, with a negative predictive value of 88%. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Mileto and Potretzke in this issue.

Predictors of malignancy in incidental adnexal lesions identified on CT in patients with prior non-ovarian cancer.

PURPOSE: To identify imaging features in incidental adnexal lesions which are associated with malignancy on portal venous phase contrast-enhanced CT in patients with known non-ovarian cancer. MATERIALS AND METHODS: This IRB-approved, HIPAA-compliant retrospective study was performed at a tertiary cancer center. Portal venous phase contrast-enhanced CT from January 2010 to December 2015 was reviewed to identify women with non-ovarian malignancy and incidental adnexal lesion, with mean 18 months (range 1-80 months) to definitive diagnosis or last imaging follow-up. Imaging features of adnexal lesions were recorded (size, laterality, shape, attenuation, and composition) and correlated with outcome (benign or malignant) using univariate and multivariate logistic regression analysis. A point-based system was used to predict likelihood of malignancy. RESULTS: Of 276 women (mean age 45 years), 216 (78.3%) had benign lesions, 58 (21.0%) ovarian metastasis, and 2 (0.7%) had primary ovarian malignancy. On logistic regression model, lesion size > 5 cm (p-value, OR, 95% CI 0.01, 9.11, 1.70-48.87), bilaterality (< 0.0001, 28.34, 7.46-107.67), irregular shape (0.01, 12.31, 1.61-94.05), higher-than-simple-fluid attenuation (< 0.0001, 28.27, 5.65-141.59), and heterogeneous composition (0.0017, 10.75, 2.45-47.23) were associated with malignant outcome (AUC 0.97). A point-based system incorporating these five features (possible 0-5 points) had AUC of 0.97. Rate of malignancy was 0% (0/147) if none of the features of malignancy were present, 12.7% (8/63) if one feature was present, 51.7% (15/29) if two features were present, and 100% (37/37) if three or more features present. CONCLUSION: Risk of malignancy of incidental adnexal lesions in women with prior non-ovarian cancer can be estimated based on lesion features seen on portal venous phase contrast-enhanced CT.

Representing narrative evidence as clinical evidence logic statements.

Objective: Clinical evidence logic statements (CELS) are shareable knowledge artifacts in a semistructured "If-Then" format that can be used for clinical decision support systems. This project aimed to assess factors facilitating CELS representation. Materials and Methods: We described CELS representation of clinical evidence. We assessed factors that facilitate representation, including authoring instruction, evidence structure, and educational level of CELS authors. Five researchers were tasked with representing CELS from published evidence. Represented CELS were compared with the formal representation. After an authoring instruction intervention, the same researchers were asked to represent the same CELS and accuracy was compared with that preintervention using McNemar's test. Moreover, CELS representation accuracy was compared between evidence that is structured versus semistructured, and between CELS authored by specialty-trained versus nonspecialty-trained researchers, using χ2 analysis. Results: 261 CELS were represented from 10 different pieces of published evidence by the researchers pre- and postintervention. CELS representation accuracy significantly increased post-intervention, from 20/261 (8%) to 63/261 (24%, P value < .00001). More CELS were assigned for representation with 379 total CELS subsequently included in the analysis (278 structured and 101 semistructured) postintervention. Representing CELS from structured evidence was associated with significantly higher CELS representation accuracy (P = .002), as well as CELS representation by specialty-trained authors (P = .0004). Discussion: CELS represented from structured evidence had a higher representation accuracy compared with semistructured evidence. Similarly, specialty-trained authors had higher accuracy when representing structured evidence. Conclusion: Authoring instructions significantly improved CELS representation with a 3-fold increase in accuracy. However, CELS representation remains a challenging task.

Closing the Loop on Unscheduled Diagnostic Imaging Orders: A Systems-Based Approach.

PURPOSE: The aims of this study were to (1) describe the System for Coordinating Orders for Radiology Exams (SCORE), the objective of which is to manage unscheduled orders for outpatient diagnostic imaging in an electronic health record (EHR) with embedded computerized physician order entry; (2) assess the impact of SCORE and other related factors (eg, demographics) on the rate of unscheduled orders; and (3) assess the clinical necessity of orders canceled, expired, scheduled, and performed. METHODS: This institutional review board-approved retrospective study was conducted at a large academic institution between October 1, 2017, and July 1, 2019. The design and implementation of SCORE are described, including people (eg, competencies), processes (eg, standardized procedures), and tools (eg, EHR interfaces, dashboard). The rate of unscheduled imaging orders was compared before SCORE (October 1, 2017, to September 30, 2018) and after SCORE (October 1, 2018, to Jun 30, 2019) using χ2 analysis. For 447 randomly selected orders, mode of resolution was obtained from the EHR, and factors related to order resolution were assessed using multivariate analysis. Finally, clinical necessity was manually assessed by two physicians. RESULTS: Before SCORE, 52,204 of 607,020 examination orders (8.6%) were unscheduled, compared with 20,900 of 475,000 examination orders (4.4%) after SCORE (P < .00001, χ2 test), a 49% reduction in unscheduled orders. Among 447 randomly selected orders, orders were addressed via cancellation (57%), expiration (21%), scheduling (1%), and performance (11%). Order resolution was not significantly associated with other factors. About 32% of cancellations and 27.7% of expired orders remained clinically necessary, which was attributed to scheduling and patient-related factors. CONCLUSIONS: SCORE significantly reduced unscheduled diagnostic imaging orders. This patient safety initiative may help reduce errors resulting from diagnostic delays due to unscheduled examination orders.

Three-Dimensional Neural Network to Automatically Assess Liver Tumor Burden Change on Consecutive Liver MRIs.

BACKGROUND: Tumor response to therapy is often assessed by measuring change in liver lesion size between consecutive MRIs. However, these evaluations are both tedious and time-consuming for clinical radiologists. PURPOSE: In this study, we sought to develop a convolutional neural network to detect liver metastases on MRI and applied this algorithm to assess change in tumor size on consecutive examinations. METHODS: We annotated a data set of 64 patients with neuroendocrine tumors who underwent at least two consecutive liver MRIs with gadoxetic acid. We then developed a 3-D neural network using a U-Net architecture with ResNet-18 building blocks that first detected the liver and then lesions within the liver. Liver lesion labels for each examination were then matched in 3-D space using an iterative closest point algorithm followed by Kuhn-Munkres algorithm. RESULTS: We developed a deep learning algorithm that detected liver metastases, co-registered the detected lesions, and then assessed the interval change in tumor burden between two multiparametric liver MRI examinations. Our deep learning algorithm was concordant in 91% with the radiologists' manual assessment about the interval change of disease burden. It had a sensitivity of 0.85 (95% confidence interval (95% CI): 0.77; 0.93) and specificity of 0.92 (95% CI: 0.87; 0.96) to classify liver segments as diseased or healthy. The mean DICE coefficient for individual lesions ranged between 0.73 and 0.81. CONCLUSIONS: Our algorithm displayed high agreement with human readers for detecting change in liver lesions on MRI, offering evidence that artificial intelligence-based detectors may perform these tasks as part of routine clinical care in the future.

Prostate Cancer Detection using Deep Convolutional Neural Networks.

Prostate cancer is one of the most common forms of cancer and the third leading cause of cancer death in North America. As an integrated part of computer-aided detection (CAD) tools, diffusion-weighted magnetic resonance imaging (DWI) has been intensively studied for accurate detection of prostate cancer. With deep convolutional neural networks (CNNs) significant success in computer vision tasks such as object detection and segmentation, different CNN architectures are increasingly investigated in medical imaging research community as promising solutions for designing more accurate CAD tools for cancer detection. In this work, we developed and implemented an automated CNN-based pipeline for detection of clinically significant prostate cancer (PCa) for a given axial DWI image and for each patient. DWI images of 427 patients were used as the dataset, which contained 175 patients with PCa and 252 patients without PCa. To measure the performance of the proposed pipeline, a test set of 108 (out of 427) patients were set aside and not used in the training phase. The proposed pipeline achieved area under the receiver operating characteristic curve (AUC) of 0.87 (95[Formula: see text] Confidence Interval (CI): 0.84-0.90) and 0.84 (95[Formula: see text] CI: 0.76-0.91) at slice level and patient level, respectively.