Evaluation of a deep learning method for the automated detection of supraspinatus tears on MRI.

OBJECTIVE: To evaluate if deep learning is a feasible approach for automated detection of supraspinatus tears on MRI. MATERIALS AND METHODS: A total of 200 shoulder MRI studies performed between 2015 and 2019 were retrospectively obtained from our institutional database using a balanced random sampling of studies containing a full-thickness tear, partial-thickness tear, or intact supraspinatus tendon. A 3-stage pipeline was developed comprised of a slice selection network based on a pre-trained residual neural network (ResNet); a segmentation network based on an encoder-decoder network (U-Net); and a custom multi-input convolutional neural network (CNN) classifier. Binary reference labels were created following review of radiologist reports and images by a radiology fellow and consensus validation by two musculoskeletal radiologists. Twenty percent of the data was reserved as a holdout test set with the remaining 80% used for training and optimization under a fivefold cross-validation strategy. Classification and segmentation accuracy were evaluated using area under the receiver operating characteristic curve (AUROC) and Dice similarity coefficient, respectively. Baseline characteristics in correctly versus incorrectly classified cases were compared using independent sample t-test and chi-squared. RESULTS: Test sensitivity and specificity of the classifier at the optimal Youden's index were 85.0% (95% CI: 62.1-96.8%) and 85.0% (95% CI: 62.1-96.8%), respectively. AUROC was 0.943 (95% CI: 0.820-0.991). Dice segmentation accuracy was 0.814 (95% CI: 0.805-0.826). There was no significant difference in AUROC between 1.5 T and 3.0 T studies. Sub-analysis showed superior sensitivity on full-thickness (100%) versus partial-thickness (72.5%) subgroups. DATA CONCLUSION: Deep learning is a feasible approach to detect supraspinatus tears on MRI.

Medical 3D Printing Dimensional Accuracy for Multi-pathological Anatomical Models 3D Printed Using Material Extrusion.

Medical 3D printing of anatomical models is being increasingly applied in healthcare facilities. The accuracy of such 3D-printed anatomical models is an important aspect of their overall quality control. The purpose of this research was to test whether the accuracy of a variety of anatomical models 3D printed using Material Extrusion (MEX) lies within a reasonable tolerance level, defined as less than 1-mm dimensional error. Six medical models spanning across anatomical regions (musculoskeletal, neurological, abdominal, cardiovascular) and sizes (model volumes ranging from ~ 4 to 203 cc) were chosen for the primary study. Three measurement landing blocks were strategically designed within each of the six medical models to allow high-resolution caliper measurements. An 8-cc reference cube was printed as the 7th model in the primary study. In the secondary study, the effect of model rotation and scale was assessed using two of the models from the first study. All models were 3D printed using an Ultimaker 3 printer in triplicates. All absolute measurement errors were found to be less than 1 mm with a maximum error of 0.89 mm. The maximum relative error was 2.78%. The average absolute error was 0.26 mm, and the average relative error was 0.71% in the primary study, and the results were similar in the secondary study with an average absolute error of 0.30 mm and an average relative error of 0.60%. The relative errors demonstrated certain patterns in the data, which were explained based on the mechanics of MEX 3D printing. Results indicate that the MEX process, when carefully assessed on a case-by-case basis, could be suitable for the 3D printing of multi-pathological anatomical models for surgical planning if an accuracy level of 1 mm is deemed sufficient for the application.

Improved Appropriateness of Advanced Diagnostic Imaging After Implementation of Clinical Decision Support Mechanism.

The Protecting Access to Medicare Act (PAMA) mandates clinical decision support mechanism (CDSM) consultation for all advanced imaging. There are a growing number of studies examining the association of CDSM use with imaging appropriateness, but a paucity of multicenter data. This observational study evaluates the association between changes in advanced imaging appropriateness scores with increasing provider exposure to CDSM. Each provider's first 200 consecutive anonymized requisitions for advanced imaging (CT, MRI, ultrasound, nuclear medicine) using a single CDSM (CareSelect, Change Healthcare) between January 1, 2017 and December 31, 2019 were collected from 288 US institutions. Changes in imaging requisition proportions among four appropriateness categories ("usually appropriate" [green], "may be appropriate" [yellow], "usually not appropriate" [red], and unmapped [gray]) were evaluated in relation to the chronological order of the requisition for each provider and total provider exposure to CDSM using logistic regression fits and Wald tests. The number of providers and requisitions included was 244,158 and 7,345,437, respectively. For 10,123 providers with ≥ 200 requisitions (2,024,600 total requisitions), the fraction of green, yellow, and red requisitions among the last 10 requisitions changed by +3.0% (95% confidence interval +2.6% to +3.4%), -0.8% (95% CI -0.5% to -1.1%), and -3.0% (95% CI 3.3% to -2.7%) in comparison with the first 10, respectively. Providers with > 190 requisitions had 8.5% (95% CI 6.3% to 10.7%) more green requisitions, 2.3% (0.7% to 3.9%) fewer yellow requisitions, and 0.5% (95% CI -1.0% to 2.0%) fewer red (not statistically significant) requisitions relative to providers with ≤ 10 requisitions. Increasing provider exposure to CDSM is associated with improved appropriateness scores for advanced imaging requisitions.

Gender Disparity Among Leaders of Canadian Academic Radiology Departments.

OBJECTIVE. Underrepresentation of women in the top hierarchy of academic medicine exists despite women comprising more than half of the medical school graduates and residency positions. The purpose of this study is to analyze and quantify the relationship of gender, research productivity, and career advancement in Canadian academic radiology departments. MATERIALS AND METHODS. Seventeen academic radiology departments with affiliated residency programs in Canada were searched for publicly available data on faculty to generate a database for gender and academic profiles of the radiologists. Bibliometric data were collected using Scopus archives. The associations of gender, academic ranks, and leadership positions were assessed, and a p value of ≤ 0.05 was defined as significant. Significant variables were analyzed using a multivariate linear regression model. RESULTS. Of 1266 faculty members, gender information and academic rank were available for 932 faculty members: 597 (64.05%) were men and 335 (35.95%) were women (χ2 = 21.82; p < 0.0001). Of a total of 563 assistant professors, 331 (58.79%) were men and 232 (41.21%) were women; of 258 associate professors, 177 (68.60%) were men and 81 (31.40%) were women; and of 111 professors, 89 (80.18%) were men and 22 (19.82%) were women. The gender gap widens at higher academic ranks, displaying a threefold drop in the ratio of women holding the rank of full professor (6.57%) compared with 14.91% male professors; 29.55% of women radiologists have first-in-command leadership positions compared with 70.45% of men. A comparable or higher h-index is noted for women Canadian radiologists after adjusting for number of citations, number of publications, and years of active research. CONCLUSION. Canadian academic radiology departments have fewer women radiologists in senior faculty and leadership positions. Our study results show that Canadian female radiologists at the professor level have more publications than their male counterparts.

RadioGraphics Update: Medical 3D Printing for the Radiologist.

Editor's Note.-Articles in the RadioGraphics Update section provide current knowledge to supplement or update information found in full-length articles previously published in RadioGraphics. Authors of the previously published article provide a brief synopsis that emphasizes important new information such as technological advances, revised imaging protocols, new clinical guidelines involving imaging, or updated classification schemes. Articles in this section are published solely online and are linked to the original article.

Patient Preferences for Coronary CT Angiography with Stress Perfusion, SPECT, or Invasive Coronary Angiography.

Background Patient preference is pivotal for widespread adoption of tests in clinical practice. Patient preferences for invasive versus other noninvasive tests for coronary artery disease are not known. Purpose To compare patient acceptance and preferences for noninvasive and invasive cardiac imaging in North and South America, Asia, and Europe. Materials and Methods This was a prospective 16-center trial in 381 study participants undergoing coronary CT angiography with stress perfusion, SPECT, and invasive coronary angiography (ICA). Patient preferences were collected by using a previously validated questionnaire translated into eight languages. Responses were converted to ordinal scales and were modeled with generalized linear mixed models. Results In patients in whom at least one test was associated with pain, CT and SPECT showed reduced median pain levels, reported on 0-100 visual analog scales, from 20 for ICA (interquartile range [IQR], 4-50) to 6 for CT (IQR, 0-27.5) and 5 for SPECT (IQR, 0-25) (P < .001). Patients from Asia reported significantly more pain than patients from other continents for ICA (median, 25; IQR, 10-50; P = .01), CT (median, 10; IQR, 0-30; P = .02), and SPECT (median, 7; IQR, 0-28; P = .03). Satisfaction with preparation differed by continent and test (P = .01), with patients from Asia reporting generally lower ratings. Patients from North America had greater percentages of "very high" or "high" satisfaction than patients from other continents for ICA (96% vs 82%, respectively; P < .001) and SPECT (95% vs 79%, respectively; P = .04) but not for CT (89% vs 86%, respectively; P = .70). Among all patients, CT was preferred by 54% of patients, compared with 18% for SPECT and 28% for ICA (P < .001). Conclusion For cardiac imaging, patients generally favored CT angiography with stress perfusion, while study participants from Asia generally reported lowest satisfaction. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Woodard and Nguyen in this issue.

EXTraction of EMR numerical data: an efficient and generalizable tool to EXTEND clinical research.

BACKGROUND: Electronic medical records (EMR) contain numerical data important for clinical outcomes research, such as vital signs and cardiac ejection fractions (EF), which tend to be embedded in narrative clinical notes. In current practice, this data is often manually extracted for use in research studies. However, due to the large volume of notes in datasets, manually extracting numerical data often becomes infeasible. The objective of this study is to develop and validate a natural language processing (NLP) tool that can efficiently extract numerical clinical data from narrative notes. RESULTS: To validate the accuracy of the tool EXTraction of EMR Numerical Data (EXTEND), we developed a reference standard by manually extracting vital signs from 285 notes, EF values from 300 notes, glycated hemoglobin (HbA1C), and serum creatinine from 890 notes. For each parameter of interest, we calculated the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score of EXTEND using two metrics. (1) completion of data extraction, and (2) accuracy of data extraction compared to the actual values in the note verified by chart review. At the note level, extraction by EXTEND was considered correct only if it accurately detected and extracted all values of interest in a note. Using manually-annotated labels as the gold standard, the note-level accuracy of EXTEND in capturing the numerical vital sign values, EF, HbA1C and creatinine ranged from 0.88 to 0.95 for sensitivity, 0.95 to 1.0 for specificity, 0.95 to 1.0 for PPV, 0.89 to 0.99 for NPV, and 0.92 to 0.96 in F1 scores. Compared to the actual value level, the sensitivity, PPV, and F1 score of EXTEND ranged from 0.91 to 0.95, 0.95 to 1.0 and 0.95 to 0.96. CONCLUSIONS: EXTEND is an efficient, flexible tool that uses knowledge-based rules to extract clinical numerical parameters with high accuracy. By increasing dictionary terms and developing new rules, the usage of EXTEND can easily be expanded to extract additional numerical data important in clinical outcomes research.

Applying Modern Virtual and Augmented Reality Technologies to Medical Images and Models.

Recent technological innovations have created new opportunities for the increased adoption of virtual reality (VR) and augmented reality (AR) applications in medicine. While medical applications of VR have historically seen greater adoption from patient-as-user applications, the new era of VR/AR technology has created the conditions for wider adoption of clinician-as-user applications. Historically, adoption to clinical use has been limited in part by the ability of the technology to achieve a sufficient quality of experience. This article reviews the definitions of virtual and augmented reality and briefly covers the history of their development. Currently available options for consumer-level virtual and augmented reality systems are presented, along with a discussion of technical considerations for their adoption in the clinical environment. Finally, a brief review of the literature of medical VR/AR applications is presented prior to introducing a comprehensive conceptual framework for the viewing and manipulation of medical images in virtual and augmented reality. Using this framework, we outline considerations for placing these methods directly into a radiology-based workflow and show how it can be applied to a variety of clinical scenarios.

Canadian Association of Radiologists White Paper on Ethical and Legal Issues Related to Artificial Intelligence in Radiology.

Artificial intelligence (AI) software that analyzes medical images is becoming increasingly prevalent. Unlike earlier generations of AI software, which relied on expert knowledge to identify imaging features, machine learning approaches automatically learn to recognize these features. However, the promise of accurate personalized medicine can only be fulfilled with access to large quantities of medical data from patients. This data could be used for purposes such as predicting disease, diagnosis, treatment optimization, and prognostication. Radiology is positioned to lead development and implementation of AI algorithms and to manage the associated ethical and legal challenges. This white paper from the Canadian Association of Radiologists provides a framework for study of the legal and ethical issues related to AI in medical imaging, related to patient data (privacy, confidentiality, ownership, and sharing); algorithms (levels of autonomy, liability, and jurisprudence); practice (best practices and current legal framework); and finally, opportunities in AI from the perspective of a universal health care system.

Fractional Flow Reserve Estimated at Coronary CT Angiography in Intermediate Lesions: Comparison of Diagnostic Accuracy of Different Methods to Determine Coronary Flow Distribution.

Purpose To compare the diagnostic accuracy of different computed tomographic (CT) fractional flow reserve (FFR) algorithms for vessels with intermediate stenosis. Materials and Methods This cross-sectional HIPAA-compliant and human research committee-approved study applied a four-step CT FFR algorithm in 61 patients (mean age, 69 years ± 10; age range, 29-89 years) with a lesion of intermediate-diameter stenosis (25%-69%) at CT angiography who underwent FFR measurement within 90 days. The per-lesion diagnostic performance of CT FFR was tested for three different approaches to estimate blood flow distribution for CT FFR calculation. The first two, the Murray law and the Huo-Kassab rule, used coronary anatomy; the third used contrast material opacification gradients. CT FFR algorithms and CT angiography percentage diameter stenosis (DS) measurements were compared by using the area under the receiver operating characteristic curve (AUC) to detect FFRs of 0.8 or lower. Results Twenty-five lesions (41%) had FFRs of 0.8 or lower. The AUC of CT FFR determination by using contrast material gradients (AUC = 0.953) was significantly higher than that of the Huo-Kassab (AUC = 0.882, P = .043) and Murray law models (AUC = 0.871, P = .033). All three AUCs were higher than that for 50% or greater DS at CT angiography (AUC = 0.596, P < .001). Correlation of CT FFR with FFR was highest for gradients (Spearman ρ = 0.80), followed by the Huo-Kassab rule (ρ = 0.68) and Murray law (ρ = 0.67) models. All CT FFR algorithms had small biases, ranging from -0.015 (Murray) to -0.049 (Huo-Kassab). Limits of agreement were narrowest for gradients (-0.182, 0.147), followed by the Huo-Kassab rule (-0.246, 0.149) and the Murray law (-0.285, 0.256) models. Conclusion Clinicians can perform CT FFR by using a four-step approach on site to accurately detect hemodynamically significant intermediate-stenosis lesions. Estimating blood flow distribution by using coronary contrast opacification variations may improve CT FFR accuracy. © RSNA, 2017 Online supplemental material is available for this article.

Common First-Pass CT Angiography Findings Associated With Rapid Growth Rate in Abdominal Aorta Aneurysms Between 3 and 5 cm in Largest Diameter.

OBJECTIVE: The purpose of this study was to describe CT angiography (CTA) findings of lumen contrast heterogeneity and intraluminal thrombus volume and to evaluate their relationship with rapid aneurysm growth in patients with abdominal aortic aneurysms (AAA) between 3 and 5 cm. MATERIALS AND METHODS: This institutional review board-approved and HIPAA-compliant single-center retrospective study included CTA studies obtained between January 2004 and December 2014 in 140 patients with AAA (101 men, 39 women; mean age ± SD, 70 ± 9 years old; age range, 22-87 years old). Standardized measurements for aneurysm intraluminal thrombus volume and a relatively new metric termed "lumen contrast heterogeneity" were obtained from the CTA images. AAA growth rate data were acquired from all subsequent cross-sectional studies. The association between the imaging findings and rapid aneurysm growth (> 0.4 cm/y) was evaluated using multivariate logistic regression. Patient comorbidities and medications were added to the regression model to assess for further associations with rapid growth rate. RESULTS: Using a baseline logistic regression model, lumen contrast heterogeneity (odds ratio [OR], 1.16; 95% CI, 1.05-1.32), intraluminal thrombus volume (OR, 2.15; 95% CI, 1.26-3.86), and maximum AAA diameter (OR, 1.69; 95% CI, 1.03-2.84) were independently associated with increased likelihood of rapid aneurysm growth. None of the patient comorbidities or medications were significantly associated with the outcome when added to the baseline model. CONCLUSION: Both intraluminal thrombus and lumen contrast heterogeneity are seen on AAA CTA studies and can be quantified; both of these metrics are independently associated with rapid growth rate and should be recognized by radiologists evaluating patients with AAA during surveillance.

Early animal model evaluation of an implantable contrast agent to enhance magnetic resonance imaging of arterial bypass vein grafts.

Background Non-invasive monitoring of autologous vein graft (VG) bypass grafts is largely limited to detecting late luminal narrowing. Although magnetic resonance imaging (MRI) delineates vein graft intima, media, and adventitia, which may detect early failure, the scan time required to achieve sufficient resolution is at present impractical. Purpose To study VG visualization enhancement in vivo and delineate whether a covalently attached MRI contrast agent would enable quicker longitudinal imaging of the VG wall. Material and Methods Sixteen 12-week-old male C57BL/6J mice underwent carotid interposition vein grafting. The inferior vena cava of nine donor mice was treated with a gadolinium-diethylenetriaminepentaacetic acid (Gd-DTPA)-based contrast agent, with control VGs labeled with a vehicle. T1-weighted (T1W) MRI was performed serially at postoperative weeks 1, 4, 12, and 20. A portion of animals was sacrificed for histopathology following each imaging time point. Results MRI signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were significantly higher for treated VGs in the first three time points (1.73 × higher SNR, P = 0.0006, and 5.83 × higher CNR at the first time point, P = 0.0006). However, MRI signal enhancement decreased consistently in the study period, to 1.29 × higher SNR and 2.64 × higher CNR, by the final time point. There were no apparent differences in graft morphometric analyses in Masson's trichrome-stained sections. Conclusion A MRI contrast agent that binds covalently to the VG wall provides significant increase in T1W MRI signal with no observed adverse effects in a mouse model. Further optimization of the contrast agent to enhance its durability is required.

Prognostic Value of Combined CT Angiography and Myocardial Perfusion Imaging versus Invasive Coronary Angiography and Nuclear Stress Perfusion Imaging in the Prediction of Major Adverse Cardiovascular Events: The CORE320 Multicenter Study.

Purpose To compare the prognostic importance (time to major adverse cardiovascular event [MACE]) of combined computed tomography (CT) angiography and CT myocardial stress perfusion imaging with that of combined invasive coronary angiography (ICA) and stress single photon emission CT myocardial perfusion imaging. Materials and Methods This study was approved by all institutional review boards, and written informed consent was obtained. Between November 2009 and July 2011, 381 participants clinically referred for ICA and aged 45-85 years were enrolled in the Combined Noninvasive Coronary Angiography and Myocardial Perfusion Imaging Using 320-Detector Row Computed Tomography (CORE320) prospective multicenter diagnostic study. All images were analyzed in blinded independent core laboratories, and a panel of physicians adjudicated all adverse events. MACE was defined as revascularization (>30 days after index ICA), myocardial infarction, or cardiac death; hospitalization for chest pain or congestive heart failure; or arrhythmia. Late MACE was defined similarly, except for patients who underwent revascularization within the first 182 days after ICA, who were excluded. Comparisons of 2-year survival (time to MACE) used standard Kaplan-Meier curves and restricted mean survival times bootstrapped with 2000 replicates. Results An MACE (49 revascularizations, five myocardial infarctions, one cardiac death, nine hospitalizations for chest pain or congestive heart failure, and one arrhythmia) occurred in 51 of 379 patients (13.5%). The 2-year MACE-free rates for combined CT angiography and CT perfusion findings were 94% negative for coronary artery disease (CAD) versus 82% positive for CAD and were similar to combined ICA and single photon emission CT findings (93% negative for CAD vs 77% positive for CAD, P < .001 for both). Event-free rates for CT angiography and CT perfusion versus ICA and single photon emission CT for either positive or negative results were not significantly different for MACE or late MACE (P > .05 for all). The area under the receiver operating characteristic curve (AUC) for combined CT angiography and CT perfusion (AUC = 68; 95% confidence interval [CI]: 62, 75) was similar (P = .36) to that for combined ICA and single photon emission CT (AUC = 71; 95% CI: 65, 79) in the identification of MACE at 2-year follow-up. Conclusion Combined CT angiography and CT perfusion enables similar prediction of 2-year MACE, late MACE, and event-free survival similar to that enabled by ICA and single photon emission CT. © RSNA, 2017 Online supplemental material is available for this article.

Measuring and Establishing the Accuracy and Reproducibility of 3D Printed Medical Models.

Despite the rapid growth of three-dimensional (3D) printing applications in medicine, the accuracy and reproducibility of 3D printed medical models have not been thoroughly investigated. Although current technologies enable 3D models to be created with accuracy within the limits of clinical imaging spatial resolutions, this is not always achieved in practice. Inaccuracies are due to errors that occur during the imaging, segmentation, postprocessing, and 3D printing steps. Radiologists' understanding of the factors that influence 3D printed model accuracy and the metrics used to measure this accuracy is key in directing appropriate practices and establishing reference standards and validation procedures. The authors review the various factors in each step of the 3D model printing process that contribute to model inaccuracy, including the intrinsic limitations of each printing technology. In addition, common sources of model inaccuracy are illustrated. Metrics involving comparisons of model dimensions and morphology that have been developed to quantify differences between 3D models also are described and illustrated. These metrics can be used to define the accuracy of a model, as compared with the reference standard, and to measure the variability of models created by different observers or using different workflows. The accuracies reported for specific indications of 3D printing are summarized, and potential guidelines for quality assurance and workflow assessment are discussed. Online supplemental material is available for this article. ©RSNA, 2017.

Limited Hospital Variation in the Use and Yield of CT for Pulmonary Embolism in Patients Undergoing Total Hip or Total Knee Replacement Surgery.

Purpose To evaluate the variation among U.S. hospitals in overall use and yield of in-hospital computed tomographic (CT) pulmonary angiography (PA) in patients undergoing total hip replacement (THR) or total knee replacement (TKR) surgery. Materials and Methods Patients in the Premier Research Database who underwent elective TKR or THR between 2007 and 2011 were enrolled in this HIPAA-compliant, institutional review board-approved retrospective observational study. The informed consent requirement was waived. Hospitals were categorized into low, medium, and high tertiles of CT PA use to compare baseline patient- and hospital-level characteristics and pulmonary embolism (PE) positivity rates. To further investigate between-hospital variation in CT PA use, a hierarchical logistic regression model that included hospital-specific random effects and fixed patient- and hospital-level effects was used. The intraclass correlation coefficient (ICC) was used to measure the amount of variability in CT PA use attributable to between-hospital variation. Results The cohort included 205 198 patients discharged from 178 hospitals (median of 734.5 patients discharged per hospital; interquartile range, 316-1461 patients) with 3647 CT PA studies (1.8%). The crude frequency of CT PA scans among the hospitals ranged from 0% to 6.2% (median, 1.6%); more than 90% of the hospitals performed CT PA in less than 3% of their patients. The mean hospital-level PE positivity rate was 12.3% (median, 9.1%); there was no significant difference in PE positivity rate across low through high CT PA use tertiles (11.3%, 11.9%, 12.9%, P = .37). After adjustment for hospital- and patient-level factors, the remaining amount of interhospital variation was relatively low (ICC, 9.0%). Conclusion Limited interhospital variation in use and yield of in-hospital CT PA was observed among patients undergoing TKR or THR in the United States. © RSNA, 2016 Online supplemental material is available for this article.

Natural Language Processing Technologies in Radiology Research and Clinical Applications.

The migration of imaging reports to electronic medical record systems holds great potential in terms of advancing radiology research and practice by leveraging the large volume of data continuously being updated, integrated, and shared. However, there are significant challenges as well, largely due to the heterogeneity of how these data are formatted. Indeed, although there is movement toward structured reporting in radiology (ie, hierarchically itemized reporting with use of standardized terminology), the majority of radiology reports remain unstructured and use free-form language. To effectively "mine" these large datasets for hypothesis testing, a robust strategy for extracting the necessary information is needed. Manual extraction of information is a time-consuming and often unmanageable task. "Intelligent" search engines that instead rely on natural language processing (NLP), a computer-based approach to analyzing free-form text or speech, can be used to automate this data mining task. The overall goal of NLP is to translate natural human language into a structured format (ie, a fixed collection of elements), each with a standardized set of choices for its value, that is easily manipulated by computer programs to (among other things) order into subcategories or query for the presence or absence of a finding. The authors review the fundamentals of NLP and describe various techniques that constitute NLP in radiology, along with some key applications.

Implementation and Performance of Automated Software for Computing Right-to-Left Ventricular Diameter Ratio From Computed Tomography Pulmonary Angiography Images.

OBJECTIVE: The aim of this study was to prospectively test the performance and potential for clinical integration of software that automatically calculates the right-to-left ventricular (RV/LV) diameter ratio from computed tomography pulmonary angiography images. METHODS: Using 115 computed tomography pulmonary angiography images that were positive for acute pulmonary embolism, we prospectively evaluated RV/LV ratio measurements that were obtained as follows: (1) completely manual measurement (reference standard), (2) completely automated measurement using the software, and (3 and 4) using a customized software interface that allowed 2 independent radiologists to manually adjust the automatically positioned calipers. RESULTS: Automated measurements underestimated (P < 0.001) the reference standard (1.09 [0.25] vs1.03 [0.35]). With manual correction of the automatically positioned calipers, the mean ratio became closer to the reference standard (1.06 [0.29] by read 1 and 1.07 [0.30] by read 2), and the correlation improved (r = 0.675 to 0.872 and 0.887). The mean time required for manual adjustment (37 [20] seconds) was significantly less than the time required to perform measurements entirely manually (100 [23] seconds). CONCLUSIONS: Automated CT RV/LV diameter ratio software shows promise for integration into the clinical workflow for patients with acute pulmonary embolism.

Three patients with full facial transplantation.

Unlike conventional reconstruction, facial transplantation seeks to correct severe deformities in a single operation. We report on three patients who received full-face transplants at our institution in 2011 in operations that aimed for functional restoration by coaptation of all main available motor and sensory nerves. We enumerate the technical challenges and postoperative complications and their management, including single episodes of acute rejection in two patients. At 6 months of follow-up, all facial allografts were surviving, facial appearance and function were improved, and glucocorticoids were successfully withdrawn in all patients.