Accelerated Diffusion-Weighted Magnetic Resonance Imaging of the Liver at 1.5 T With Deep Learning-Based Image Reconstruction: Impact on Image Quality and Lesion Detection.

OBJECTIVE: To perform image quality comparison between deep learning-based multiband diffusion-weighted sequence (DL-mb-DWI), accelerated multiband diffusion-weighted sequence (accelerated mb-DWI), and conventional multiband diffusion-weighted sequence (conventional mb-DWI) in patients undergoing clinical liver magnetic resonance imaging (MRI). METHODS: Fifty consecutive patients who underwent clinical MRI of the liver at a 1.5-T scanner, between September 1, 2021, and January 31, 2022, were included in this study. Three radiologists independently reviewed images using a 5-point Likert scale for artifacts and image quality factors, in addition to assessing the presence of liver lesions and lesion conspicuity. RESULTS: DL-mb-DWI acquisition time was 65.0 ± 2.4 seconds, significantly (P < 0.001) shorter than conventional mb-DWI (147.5 ± 19.2 seconds) and accelerated mb-DWI (94.3 ± 1.8 seconds). DL-mb-DWI received significantly higher scores than conventional mb-DWI for conspicuity of the left lobe (P < 0.001), sharpness of intrahepatic vessel margin (P < 0.001), sharpness of the pancreatic contour (P < 0.001), in-plane motion artifact (P = 0.002), and overall image quality (P = 0.005) by reader 2. DL-mb-DWI received significantly higher scores for conspicuity of the left lobe (P = 0.006), sharpness of the pancreatic contour (P = 0.020), and in-plane motion artifact (P = 0.042) by reader 3. DL-mb-DWI received significantly higher scores for strength of fat suppression (P = 0.004) and sharpness of the pancreatic contour (P = 0.038) by reader 1. The remaining quality parameters did not reach statistical significance for reader 1. CONCLUSIONS: Novel diffusion-weighted MRI sequence with deep learning-based image reconstruction demonstrated significantly decreased acquisition times compared with conventional and accelerated mb-DWI sequences, while maintaining or improving image quality for routine abdominal MRI. DL-mb-DWI offers a potential alternative to conventional mb-DWI in routine clinical liver MRI.

How Satisfied Are Patients With Their Radiologists? Assessment Using a National Patient Ratings Website.

OBJECTIVE: The purpose of this study is to assess features of patient satisfaction scores for U.S. radiologists using a popular physician rating website. MATERIALS AND METHODS: Patient reviews were retrieved from the website RateMDs for all listed radiologists in all 297 U.S. cities with population 100,000 or greater. Reviews included rating scores of 1-5 (5 = highest) in four categories (staff, punctuality, knowledge, and helpfulness). Additional physician information was obtained from Medicare files. Common words in patient free-text comments were assessed. Statistical analyses were performed. RESULTS: We identified 1891 patient reviews for 1259 radiologists. In all four categories, the most common score was 5 for excellent (62.7-74.3%), and the second most common score was 1 for terrible (13.5-20.4%); scores of 2-4 were far less frequent (1.9-11.6%). Scores for all four categories highly correlated with one another (r = 0.781-0.951). Radiologists in the Northeast scored significantly lower (p < 0.001) than those elsewhere for both staff and punctuality. Radiologists attending a designated top 50 medical school showed nonsignificant trends toward lower scores for helpfulness (p = 0.073) and knowledge (p = 0.062). The most common words in free-text comments for positive reviews were "caring," "knowledgeable," and "professional." For negative reviews, "rude," "pain," and "unprofessional" were most common. CONCLUSION: Overall, most radiologists rated online by their patients score well, but reviews tended to be either strongly positive or negative. Scores across various categories are highly correlated, suggesting that there is a halo effect. Radiologists should recognize the effect of both facility- and radiologist-related factors in influencing patients' overall perceptions.

Academic Radiologist Subspecialty Identification Using a Novel Claims-Based Classification System.

OBJECTIVE: The objective of the present study is to assess the feasibility of a novel claims-based classification system for payer identification of academic radiologist subspecialties. MATERIALS AND METHODS: Using a categorization scheme based on the Neiman Imaging Types of Service (NITOS) system, we mapped the Medicare Part B services billed by all radiologists from 2012 to 2014, assigning them to the following subspecialty categories: abdominal imaging, breast imaging, cardiothoracic imaging, musculoskeletal imaging, nuclear medicine, interventional radiology, and neuroradiology. The percentage of subspecialty work relative value units (RVUs) to total billed work RVUs was calculated for each radiologist nationwide. For radiologists at the top 20 academic departments funded by the National Institutes of Health, those percentages were compared with subspecialties designated on faculty websites. NITOS-based subspecialty assignments were also compared with the only radiologist subspecialty classifications currently recognized by Medicare (i.e., nuclear medicine and interventional radiology). RESULTS: Of 1012 academic radiologists studied, the median percentage of Medicare-billed NITOS-based subspecialty work RVUs matching the subspecialty designated on radiologists' own websites ranged from 71.3% (for nuclear medicine) to 98.9% (for neuroradiology). A NITOS-based work RVU threshold of 50% correctly classified 89.8% of radiologists (5.9% were not mapped to any subspecialty; subspecialty error rate, 4.2%). In contrast, existing Medicare provider codes identified only 46.7% of nuclear medicine physicians and 39.4% of interventional radiologists. CONCLUSION: Using a framework based on a recently established imaging health services research tool that maps service codes based on imaging modality and body region, Medicare claims data can be used to consistently identify academic radiologists by subspecialty in a manner not possible with the use of existing Medicare physician specialty identifiers. This method may facilitate more appropriate performance metrics for subspecialty academic physicians under emerging value-based payment models.

Interobserver Reproducibility of the PI-RADS Version 2 Lexicon: A Multicenter Study of Six Experienced Prostate Radiologists.

Purpose To determine the interobserver reproducibility of the Prostate Imaging Reporting and Data System (PI-RADS) version 2 lexicon. Materials and Methods This retrospective HIPAA-compliant study was institutional review board-approved. Six radiologists from six separate institutions, all experienced in prostate magnetic resonance (MR) imaging, assessed prostate MR imaging examinations performed at a single center by using the PI-RADS lexicon. Readers were provided screen captures that denoted the location of one specific lesion per case. Analysis entailed two sessions (40 and 80 examinations per session) and an intersession training period for individualized feedback and group discussion. Percent agreement (fraction of pairwise reader combinations with concordant readings) was compared between sessions. κ coefficients were computed. Results No substantial difference in interobserver agreement was observed between sessions, and the sessions were subsequently pooled. Agreement for PI-RADS score of 4 or greater was 0.593 in peripheral zone (PZ) and 0.509 in transition zone (TZ). In PZ, reproducibility was moderate to substantial for features related to diffusion-weighted imaging (κ = 0.535-0.619); fair to moderate for features related to dynamic contrast material-enhanced (DCE) imaging (κ = 0.266-0.439); and fair for definite extraprostatic extension on T2-weighted images (κ = 0.289). In TZ, reproducibility for features related to lesion texture and margins on T2-weighted images ranged from 0.136 (moderately hypointense) to 0.529 (encapsulation). Among 63 lesions that underwent targeted biopsy, classification as PI-RADS score of 4 or greater by a majority of readers yielded tumor with a Gleason score of 3+4 or greater in 45.9% (17 of 37), without missing any tumor with a Gleason score of 3+4 or greater. Conclusion Experienced radiologists achieved moderate reproducibility for PI-RADS version 2, and neither required nor benefitted from a training session. Agreement tended to be better in PZ than TZ, although was weak for DCE in PZ. The findings may help guide future PI-RADS lexicon updates. (©) RSNA, 2016 Online supplemental material is available for this article.

Radiation Therapy for Lung Cancer: Imaging Appearances and Pitfalls.

Radiation therapy is part of a multimodality treatment approach to lung cancer. The radiologist must be aware of both the expected and the unexpected imaging findings of the post-radiation therapy patient, including the time course for development of post- radiation therapy pneumonitis and fibrosis. In this review, a brief discussion of radiation therapy techniques and indications is presented, followed by an image-heavy differential diagnostic approach. The review focuses on computed tomography imaging examples to help distinguish normal postradiation pneumonitis and fibrosis from alternative complications, such as infection, local recurrence, or radiation-induced malignancy.

Impact of 3D printed models on quantitative surgical outcomes for patients undergoing robotic-assisted radical prostatectomy: a cohort study.

BACKGROUND: Three-dimensional (3D) printed anatomic models can facilitate presurgical planning by providing surgeons with detailed knowledge of the exact location of pertinent anatomical structures. Although 3D printed anatomic models have been shown to be useful for pre-operative planning, few studies have demonstrated how these models can influence quantitative surgical metrics. OBJECTIVE: To prospectively assess whether patient-specific 3D printed prostate cancer models can improve quantitative surgical metrics in patients undergoing robotic-assisted radical prostatectomy (RARP). METHODS: Patients with MRI-visible prostate cancer (PI-RADS V2 ≥ 3) scheduled to undergo RARP were prospectively enrolled in our IRB approved study (n = 82). Quantitative surgical metrics included the rate of positive surgical margins (PSMs), operative times, and blood loss. A qualitative Likert scale survey to assess understanding of anatomy and confidence regarding surgical approach was also implemented. RESULTS: The rate of PSMs was lower for the 3D printed model group (8.11%) compared to that with imaging only (28.6%), p = 0.128. The 3D printed model group had a 9-min reduction in operating time (213 ± 42 min vs. 222 ± 47 min) and a 5 mL reduction in average blood loss (227 ± 148 mL vs. 232 ± 114 mL). Surgeon anatomical understanding and confidence improved after reviewing the 3D printed models (3.60 ± 0.74 to 4.20 ± 0.56, p = 0.62 and 3.86 ± 0.53 to 4.20 ± 0.56, p = 0.22). CONCLUSIONS: 3D printed prostate cancer models can positively impact quantitative patient outcomes such as PSMs, operative times, and blood loss in patients undergoing RARP.

Accelerated T2-weighted MRI of the liver at 3 T using a single-shot technique with deep learning-based image reconstruction: impact on the image quality and lesion detection.

PURPOSE: Fat-suppressed T2-weighted imaging (T2-FS) requires a long scan time and can be wrought with motion artifacts, urging the development of a shorter and more motion robust sequence. We compare the image quality of a single-shot T2-weighted MRI prototype with deep-learning-based image reconstruction (DL HASTE-FS) with a standard T2-FS sequence for 3 T liver MRI. METHODS: 41 consecutive patients with 3 T abdominal MRI examinations including standard T2-FS and DL HASTE-FS, between 5/6/2020 and 11/23/2020, comprised the study cohort. Three radiologists independently reviewed images using a 5-point Likert scale for artifact and image quality measures, while also assessing for liver lesions. RESULTS: DL HASTE-FS acquisition time was 54.93 ± 16.69, significantly (p < .001) shorter than standard T2-FS (114.00 ± 32.98 s). DL HASTE-FS received significantly higher scores for sharpness of liver margin (4.3 vs 3.3; p < .001), hepatic vessel margin (4.2 vs 3.3; p < .001), pancreatic duct margin (4.0 vs 1.9; p < .001); in-plane (4.0 vs 3.2; p < .001) and through-plane (3.9 vs 3.4; p < .001) motion artifacts; other ghosting artifacts (4.3 vs 2.9; p < .001); and overall image quality (4.0 vs 2.9; p < .001), in addition to receiving a higher score for homogeneity of fat suppression (3.7 vs 3.4; p = .04) and liver-fat contrast (p = .03). For liver lesions, DL HASTE-FS received significantly higher scores for sharpness of lesion margin (4.4 vs 3.7; p = .03). CONCLUSION: Novel single-shot T2-weighted MRI with deep-learning-based image reconstruction demonstrated superior image quality compared with the standard T2-FS sequence for 3 T liver MRI, while being acquired in less than half the time.

Feasibility of Accelerated Prostate Diffusion-Weighted Imaging on 0.55 T MRI Enabled With Random Matrix Theory Denoising.

INTRODUCTION: Prostate cancer diffusion weighted imaging (DWI) MRI is typically performed at high-field strength (3.0 T) in order to overcome low signal-to-noise ratio (SNR). In this study, we demonstrate the feasibility of prostate DWI at low field enabled by random matrix theory (RMT)-based denoising, relying on the MP-PCA algorithm applied during image reconstruction from multiple coils. METHODS: Twenty-one volunteers and 2 prostate cancer patients were imaged with a 6-channel pelvic surface array coil and an 18-channel spine array on a prototype 0.55 T system created by ramping down a commercial magnetic resonance imaging system (1.5 T MAGNETOM Aera Siemens Healthcare) with 45 mT/m gradients and 200 T/m/s slew rate. Diffusion-weighted images were acquired with 4 non-collinear directions, for which b = 50 s/mm 2 was used with 8 averages and b = 1000 s/mm 2 with 40 averages; 2 extra b = 50 s/mm 2 were used as part of the dynamic field correction. Standard and RMT-based reconstructions were applied on DWI over different ranges of averages. Accuracy/precision was evaluated using the apparent diffusion coefficient (ADC), and image quality was evaluated over 5 separate reconstructions by 3 radiologists with a 5-point Likert scale. For the 2 patients, we compare image quality and lesion visibility of the RMT reconstruction versus the standard one on 0.55 T and on clinical 3.0 T. RESULTS: The RMT-based reconstruction in this study reduces the noise floor by a factor of 5.8, thereby alleviating the bias on prostate ADC. Moreover, the precision of the ADC in prostate tissue after RMT increases over a range of 30%-130%, with the increase in both signal-to-noise ratio and precision being more prominent for a low number of averages. Raters found that the images were consistently of moderate to good overall quality (3-4 on the Likert scale). Moreover, they determined that b = 1000 s/mm 2 images from a 1:55-minute scan with the RMT-based reconstruction were on par with the corresponding images from a 14:20-minute scan with standard reconstruction. Prostate cancer was visible on ADC and calculated b = 1500 images even with the abbreviated 1:55-minute scan reconstructed with RMT. CONCLUSIONS: Prostate imaging using DWI is feasible at low field and can be performed more rapidly with noninferior image quality compared with standard reconstruction.

Structured versus non-structured reporting of pelvic MRI for ileal pouch evaluation: clarity and effectiveness.

PURPOSE: Given that ileal pouch-anal anastomosis (IPAA) surgery is a technically challenging and high-morbidity procedure, there are numerous pertinent imaging findings that need to be clearly and efficiently communicated to the IBD surgeons for essential patient management and surgical planning. Structured reporting has been increasingly used over the past decade throughout various radiology subspecialties to improve reporting clarity and completeness. We compare structured versus non-structured reporting of pelvic MRI for ileal pouch to evaluate for clarity and effectiveness. METHODS: 164 consecutive pelvic MRI's for ileal pouch evaluation, excluding subsequent exams for the same patient, acquired between 1/1/2019 and 7/31/2021 at one institution were included, before and after implementation (11/15/2020) of a structured reporting template, which was created with institutional IBD surgeons. Reports were assessed for the presence of 18 key features required for complete ileal pouch assessment: anastomosis (IPAA, tip of J, pouch body), cuff (length, cuffitis), pouch body (size, pouchitis, stricture), pouch inlet/pre-pouch ileum (stricture, inflammation, sharp angulation), pouch outlet (stricture), peripouch mesentery (position, mesentery twist), pelvic abscess, peri-anal fistula, pelvic lymph nodes, and skeletal abnormalities. Subgroup analysis was performed based on reader experience and divided into three categories: experienced (n = 2), other intra-institutional (n = 20), or affiliate site (n = 6). RESULTS: 57 (35%) structured and 107 (65%) non-structured pelvic MRI reports were reviewed. Structured reports contained 16.6 [SD:4.0] key features whereas non-structured reports contained 6.3 [SD:2.5] key features (p < .001). The largest improvement following template implementation was for reporting sharp angulation of the pouch inlet (91.2% vs. 0.9%, p < .001), tip of J suture line and pouch body anastomosis (both improved to 91.2% from 3.7%). Structured versus non-structured reports contained mean 17.7 versus 9.1 key features for experienced readers, 17.0 versus 5.9 for other intra-institutional readers, and 8.7 versus 5.3 for affiliate site readers. CONCLUSION: Structured reporting of pelvic MRI guides a systematic search pattern and comprehensive evaluation of ileal pouches, and therefore facilitates surgical planning and clinical management. This standardized reporting template can serve as baseline at other institutions for adaptation based on specific radiology and surgery preferences, fostering a collaborative environment between radiology and surgery, and ultimately improving patient care.

Brainstorming Our Way to Improved Quality, Safety, and Resident Wellness in a Resource-Limited Emergency Department.

PURPOSE: To implement a more efficient standardized computed tomography (CT) protocoling system for emergency department (ED) patients in order to improve resident work satisfaction and wellness, and decrease lag time between ordering and protocoling a study. METHODS AND MATERIALS: Residents recorded lag times between time of order and time of protocol for 176 CT scans between November 2018 and January 2019. Pre- and postintervention resident surveys of 7 questions utilizing a 5-point Likert scale were used to assess the perceived efficiency and overall satisfaction with the protocolling system. CT technologists received a 2-step Standardized ED CT Protocoling Guidance Sheet for common indications and would consult the radiologist for any questions. RESULTS: Lag time between order and protocol averaged 17.8 minutes. Postintervention surveys demonstrated that residents were more satisfied with the new system (100% vs 6.1%), had an overall higher job satisfaction in the ED (91% vs 12.1%), thought the system was more efficient for a single study (100% vs 15.2%) and for an entire shift (100% vs 6.1%), volume of studies was maximized (91% vs 6.1%), and the workflow allowed residents to focus on interpreting studies and communicating findings (91% vs 3%). CONCLUSION: The implementation of an auto-protocolling system at our institution's ED took a system which was disruptive, inefficient, and unreliable, and eliminated both lag time and variation in time between ordering and protocoling, improving time to final report. It simultaneously decreased interruptions, allowing residents to focus on study interpretation, which increased resident work satisfaction, wellness, and educational benefit.

Exploring Which Medical Schools Cost the Most: An Assessment of Medical School Characteristics Associated With School Tuition.

OBJECTIVE: To assess medical school characteristics associated with school tuition. MATERIALS AND METHODS: US medical schools' tuitions, and various medical school characteristics, were extracted from the Association of American Medical Colleges' online MSAR database, using in-state tuition when applicable. US News ranking and National Institutes of Health (NIH) award ranking from the Blue Ridge Institute for Medical Research were obtained, when available. Geographic population density was obtained using Governing magazine's online database. Cost of living estimates were obtained from online American Chamber of Commerce Research Association Cost of Living Index. Spearman correlations were determined, and multivariable linear regression was performed. RESULTS: Among 148 included medical schools, adjusted average ± standard deviation tuition was $47,612 ± $23,765 (range $12,761-$141,464). Tuition demonstrated positive correlations with regional population density (r = +0.577) and years established (r = +0.265). Among ranked schools, tuition showed negative correlations with US News rank (r = -0.469) and NIH rank (r = -0.336). Average tuition varied by geographic region: Northeast: $49,662, Midwest: $43,560, West: $37,701, and South: $34,270. Among states with at least 3 medical schools, average tuition was highest in MA ($53,520), PA ($53,034), $51,547 (DC), and lowest in TX ($21,002), FL ($30,440), LA ($36,066). At multivariable linear regression, the strongest independent predictor of tuition was US News rank (ß = -396.0, P= 0.05). CONCLUSIONS: US medical school tuition is highly variable by over a 10:1 ratio. Tuition is greater in higher ranked, longer established schools, in more densely populated regions. Objective data regarding medical education quality may be warranted to assess whether higher tuition in schools with higher US News and NIH rankings is justified.

Instructional Vignettes in Publication and Journalism Ethics in Radiology Research: Assessment via a Survey of Radiology Trainees.

RATIONALE AND OBJECTIVES: The aim of the present study was to assess the potential usefulness of written instructional vignettes relating to publication and journalism ethics in radiology via a survey of radiology trainees. MATERIALS AND METHODS: A literature review was conducted to guide the development of vignettes, each describing a scenario relating to an ethical issue in research and publication, with subsequent commentary on the underlying ethical issue and potential approaches to its handling. Radiology trainees at a single institution were surveyed regarding the vignettes' perceived usefulness. RESULTS: A total of 21 vignettes were prepared, addressing institutional review board and human subjects protection, authorship issues, usage of previous work, manuscript review, and other miscellaneous topics. Of the solicited trainees, 24.7% (16/65) completed the survey. On average among the vignettes, 94.0% of the participants found the vignette helpful; 19.9 received prior formal instruction on the issue during medical training; 40.0% received prior informal guidance from a research mentor; and 42.0% indicated that the issue had arisen in their own or a peer's prior research experience. The most common previously experienced specific issue was authorship order (93.8%). Free-text responses were largely favorable regarding the value of the vignettes, although also indicated numerous challenges in properly handling the ethical issues: impact of hierarchy, pressure to publish, internal politics, reluctance to conduct sensitive conversations with colleagues, and variability in journal and professional society policies. CONCLUSION: Radiology trainees overall found the vignettes helpful, addressing commonly encountered topics for which formal and informal guidance were otherwise lacking. The vignettes are publicly available through the Association of University Radiologists (AUR) website and may foster greater insights by investigators into ethical aspects of the publication and journalism process, thus contributing to higher quality radiology research.

Associations Between Academic Rank and Advanced Bibliometric Indices Among United States Academic Radiologists.

RATIONALE AND OBJECTIVES: This study aimed to evaluate associations between traditional and advanced bibliometric indices with academic rank for radiologists in the United States. METHODS: Faculty web pages were searched to classify 538 members of the Association of University Radiologists as assistant (n = 212), associate (n = 128), or full (n = 198) professors. Radiologists' publication and citation records were extracted from Scopus to compute the following indices: publication count, citation count, h-index, i-10 index, hc-index, m-quotient, e-index, and g-index. Analysis of variance, multivariable logistic regression, and receiver operating characteristic curve analysis were performed. RESULTS: All indices were significantly different among the three groups (P ≤.001), progressively increasing with increasing rank (eg, mean publication count of 17, 41, and 128 among assistant, associate, and full professors, respectively; mean citation count of 205, 687, and 3622, respectively; mean h-index of 5, 11, and 27, respectively). At multivariable analysis, the h-index (reflecting publications and citations) was a strong significant independent positive predictor of associate (ß=+0.32, P <.001) or full professor (ß=+0.26, P <.001) status, whereas the m-quotient (adjusted h-index that is greater for more rapid publication) was a strong significant independent negative predictor of associate (ß=-1.87, P = .009) or full professor (ß =-4.97, P <.001) status. The models exhibited moderate goodness-of-fit (r2 = 0.534-0.655; P <.001). The model for predicting at least associate professor achieved area under the curve 0.876 (sensitivity 74.6%, specificity 88.8%). The model for predicting full professor achieved area under the curve 0.925 (sensitivity 85.5%, specificity 86.1%). CONCLUSION: When controlling for the h-index, more rapid publication, as indicated by the m-quotient, was negatively associated with radiologists' academic rank, indicating the additional influence of career duration in promotions decisions.

Evaluation for suspected acute appendicitis in the emergency department setting: a comparison of outcomes among three imaging pathways.

PURPOSE: To compare outcomes of imaging pathways in suspected acute appendicitis. METHODS: Computerized tomography (CT) alone, ultrasound alone, and ultrasound followed by CT were compared in 570 emergency department (ED) patients with suspected acute appendicitis. RESULTS: After initial ultrasound, 9.3% of men and 41.0% of women underwent CT. Body mass index (BMI) (P≤.036): 25.3±5.7kg/m(2) (CT), 19.1±3.3kg/m(2) (ultrasound), and 22.4±3.2kg/m(2) (ultrasound then CT). Age (P<.001): 35.9±14.5 years (CT), 12.8±6.2y (ultrasound), and 21.2±9.0 years (ultrasound then CT). ED length-of-stay: 7.0±2.8h (CT), 5.9±2.8h (ultrasound), and 8.4±3.5h (ultrasound then CT). Admission, same-day discharge, appendectomy, pathology positive for appendicitis or complicated appendicitis, 30-day repeat ED visit/hospitalization: no difference between pathways (P=.062-1.00). CONCLUSION: Ultrasound, selected in patients with lowest age/BMI, had shortest length-of-stay but otherwise similar outcomes.

Use of a Machine-learning Method for Predicting Highly Cited Articles Within General Radiology Journals.

RATIONALE AND OBJECTIVES: This study aimed to assess the performance of a text classification machine-learning model in predicting highly cited articles within the recent radiological literature and to identify the model's most influential article features. MATERIALS AND METHODS: We downloaded from PubMed the title, abstract, and medical subject heading terms for 10,065 articles published in 25 general radiology journals in 2012 and 2013. Three machine-learning models were applied to predict the top 10% of included articles in terms of the number of citations to the article in 2014 (reflecting the 2-year time window in conventional impact factor calculations). The model having the highest area under the curve was selected to derive a list of article features (words) predicting high citation volume, which was iteratively reduced to identify the smallest possible core feature list maintaining predictive power. Overall themes were qualitatively assigned to the core features. RESULTS: The regularized logistic regression (Bayesian binary regression) model had highest performance, achieving an area under the curve of 0.814 in predicting articles in the top 10% of citation volume. We reduced the initial 14,083 features to 210 features that maintain predictivity. These features corresponded with topics relating to various imaging techniques (eg, diffusion-weighted magnetic resonance imaging, hyperpolarized magnetic resonance imaging, dual-energy computed tomography, computed tomography reconstruction algorithms, tomosynthesis, elastography, and computer-aided diagnosis), particular pathologies (prostate cancer; thyroid nodules; hepatic adenoma, hepatocellular carcinoma, non-alcoholic fatty liver disease), and other topics (radiation dose, electroporation, education, general oncology, gadolinium, statistics). CONCLUSIONS: Machine learning can be successfully applied to create specific feature-based models for predicting articles likely to achieve high influence within the radiological literature.