Development and Implementation of a Multisite Registry Using Structured Templates for Actionable Findings in the Kidney.

PURPOSE: The aim of this study was to scale structured report templates categorizing actionable renal findings across health systems and create a centralized registry of patient and report data. METHODS: In January 2017, three academic radiology departments agreed to prospectively include identical structured templates categorizing the malignant likelihood of renal findings in ≥90% of all adult ultrasound, MRI, and CT reports, a new approach for two sites. Between November 20, 2017, and September 30, 2019, deidentified HL7 report data were transmitted to a centralized ACR registry. An automated algorithm extracted categories. Radiologists were requested to addend reports with missing or incomplete templates after the first month. Separately, each site submitted patient sociodemographic and clinical data 12 months before and at least 3 months after enrollment. RESULTS: A total of 164,982 eligible radiology reports were transmitted to the registry; 4,159 (2.5%) were excluded because of missing categories or radiologist names. The final cohort included 160,823 examinations on 102,619 unique patients. Mean template use before and after addendum requests was 99.3% and 99.9% at SITE1, 86.5% and 94.6% at SITE2, and 91.4% and 96.0% at SITE3. Matching patient sociodemographic and clinical data were obtained on 96.9% of reports from SITE1, 94.2% from SITE2, and 96.0% from SITE3. Regulatory, cultural, and technology barriers to the creation of a multisite registry were identified. CONCLUSIONS: Barriers to the adoption of unified structured report templates for actionable kidney findings can be addressed. Deidentified report and patient data can be securely transmitted to an external registry. These data can facilitate the collection of diverse evidence-based population imaging outcomes.

Automatic Fully-Contextualized Recommendation Extraction from Radiology Reports.

Recommendations are a key component of radiology reports. Automatic extraction of recommendations would facilitate tasks such as recommendation tracking, quality improvement, and large-scale descriptive studies. Existing report-parsing systems are frequently limited to recommendations for follow-up imaging studies, operate at the sentence or document level rather than the individual recommendation level, and do not extract important contextualizing information. We present a neural network architecture capable of extracting fully contextualized recommendations from any type of radiology report. We identified six major "questions" necessary to capture the majority of context associated with a recommendation: recommendation, time period, reason, conditionality, strength, and negation. We developed a unified task representation by allowing questions to refer to answers to other questions. Our representation allows for a single system to perform named entity recognition (NER) and classification tasks. We annotated 2272 radiology reports from all specialties, imaging modalities, and multiple hospitals across our institution. We evaluated the performance of a long short-term memory (LSTM) architecture on the six-question task. The single-task LSTM model achieves a token-level performance of 89.2% at recommendation extraction, and token-level performances between 85 and 95% F1 on extracting modifying features. Our model extracts all types of recommendations, including follow-up imaging, tissue biopsies, and clinical correlation, and can operate in real time. It is feasible to extract complete contextualized recommendations of all types from arbitrary radiology reports. The approach is likely generalizable to other clinical entities referenced in radiology reports, such as radiologic findings or diagnoses.

Patient Exposure from Radiologic and Nuclear Medicine Procedures in the United States: Procedure Volume and Effective Dose for the Period 2006-2016.

Background Comprehensive assessments of the frequency and associated doses from radiologic and nuclear medicine procedures are rarely conducted. The use of these procedures and the population-based radiation dose increased remarkably from 1980 to 2006. Purpose To determine the change in per capita radiation exposure in the United States from 2006 to 2016. Materials and Methods The U.S. National Council on Radiation Protection and Measurements conducted a retrospective assessment for 2016 and compared the results to previously published data for the year 2006. Effective dose values for procedures were obtained from the literature, and frequency data were obtained from commercial, governmental, and professional society data. Results In the United States in 2006, an estimated 377 million diagnostic and interventional radiologic examinations were performed. This value remained essentially the same for 2016 even though the U.S. population had increased by about 24 million people. The number of CT scans performed increased from 67 million to 84 million, but the number of other procedures (eg, diagnostic fluoroscopy) and nuclear medicine procedures decreased from 17 million to 13.5 million. The number of dental radiographic and dental CT examinations performed was estimated to be about 320 million in 2016. Using the tissue-weighting factors from Publication 60 of the International Commission on Radiological Protection, the U.S. annual individual (per capita) effective dose from diagnostic and interventional medical procedures was estimated to have been 2.9 mSv in 2006 and 2.3 mSv in 2016, with the collective doses being 885 000 and 755 000 person-sievert, respectively. Conclusion The trend from 1980 to 2006 of increasing dose from medical radiation has reversed. Estimated 2016 total collective effective dose and radiation dose per capita dose are lower than in 2006. © RSNA, 2020 See also the editorial by Einstein in this issue.

Automated Detection of Radiology Reports that Require Follow-up Imaging Using Natural Language Processing Feature Engineering and Machine Learning Classification.

While radiologists regularly issue follow-up recommendations, our preliminary research has shown that anywhere from 35 to 50% of patients who receive follow-up recommendations for findings of possible cancer on abdominopelvic imaging do not return for follow-up. As such, they remain at risk for adverse outcomes related to missed or delayed cancer diagnosis. In this study, we develop an algorithm to automatically detect free text radiology reports that have a follow-up recommendation using natural language processing (NLP) techniques and machine learning models. The data set used in this study consists of 6000 free text reports from the author's institution. NLP techniques are used to engineer 1500 features, which include the most informative unigrams, bigrams, and trigrams in the training corpus after performing tokenization and Porter stemming. On this data set, we train naive Bayes, decision tree, and maximum entropy models. The decision tree model, with an F1 score of 0.458 and accuracy of 0.862, outperforms both the naive Bayes (F1 score of 0.381) and maximum entropy (F1 score of 0.387) models. The models were analyzed to determine predictive features, with term frequency of n-grams such as "renal neoplasm" and "evalu with enhanc" being most predictive of a follow-up recommendation. Key to maximizing performance was feature engineering that extracts predictive information and appropriate selection of machine learning algorithms based on the feature set.

Assessing cardiovascular risk in the prerenal transplant population: Comparison of myocardial perfusion imaging and coronary angiography with risk factor stratification.

INTRODUCTION: Patients with end-stage renal disease (ESRD) have a higher incidence of coronary artery disease (CAD). Hence, it is crucial to evaluate CAD before renal transplantation. This study compares the utility of pharmacologic single-photon emission computed-tomography (SPECT) imaging directly to coronary angiography for diagnosis of CAD with correlation to cardiovascular risk factors. METHOD: Retrospective review of asymptomatic renal failure patients who underwent both SPECT and coronary angiography to identify obstructive CAD between the years 2008-2016. Ninety-four ESRD subjects were evaluated. RESULTS: Myocardial perfusion SPECT study found, when compared to coronary angiography demonstrated for CAD, the sensitivity of 93.3% with a specificity of 73.4%. Importantly, the negative predictive value for coronary artery disease was 96%. With seven or more cardiac risk factors, 66.7% of patients had obstructive coronary artery disease. Among all the risk factors examined, patients with a previous history of coronary artery disease had a 68% risk of obstructive coronary artery disease. CONCLUSION: Comparing myocardial perfusion imaging SPECT findings with coronary angiography in patients with ESRD, a sensitivity of 93.3% and a specificity of 73% were observed. Of all the risk factors examined, patient with the previous history of CAD was the single most significant risk factor for CAD in 68% of cases.

Toward Complete Structured Information Extraction from Radiology Reports Using Machine Learning.

Unstructured and semi-structured radiology reports represent an underutilized trove of information for machine learning (ML)-based clinical informatics applications, including abnormality tracking systems, research cohort identification, point-of-care summarization, semi-automated report writing, and as a source of weak data labels for training image processing systems. Clinical ML systems must be interpretable to ensure user trust. To create interpretable models applicable to all of these tasks, we can build general-purpose systems which extract all relevant human-level assertions or "facts" documented in reports; identifying these facts is an information extraction (IE) task. Previous IE work in radiology has focused on a limited set of information, and extracts isolated entities (i.e., single words such as "lesion" or "cyst") rather than complete facts, which require the linking of multiple entities and modifiers. Here, we develop a prototype system to extract all useful information in abdominopelvic radiology reports (findings, recommendations, clinical history, procedures, imaging indications and limitations, etc.), in the form of complete, contextualized facts. We construct an information schema to capture the bulk of information in reports, develop real-time ML models to extract this information, and demonstrate the feasibility and performance of the system.

Radiation protection for the echocardiographers: "To each their own".

This article illustrates the effectiveness of targeted radioprotective strategies for the interventional echocardiographer. The reader should recognize the importance of engagement of all team members in the multifaceted process of radiation exposure mitigation. Future efforts/studies should focus on the impact of team oriented training, lab design, and development of novel supplies and equipment to mitigate radiation exposure of all personnel in the cardiac catheterization lab, particularly during more complex interventional procedures.

Automated Organ-Level Classification of Free-Text Pathology Reports to Support a Radiology Follow-up Tracking Engine.

PURPOSE: To evaluate the performance of machine learning algorithms on organ-level classification of semistructured pathology reports, to incorporate surgical pathology monitoring into an automated imaging recommendation follow-up engine. MATERIALS AND METHODS: This retrospective study included 2013 pathology reports from patients who underwent abdominal imaging at a large tertiary care center between 2012 and 2018. The reports were labeled by two annotators as relevant to four abdominal organs: liver, kidneys, pancreas and/or adrenal glands, or none. Automated classification methods were compared: simple string matching, random forests, extreme gradient boosting, support vector machines, and two neural network architectures-convolutional neural networks and long short-term memory networks. Three methods from the literature were used to provide interpretability and qualitative validation of the learned network features. RESULTS: The neural networks performed well on the four-organ classification task (F1 score: 96.3% for convolutional neural network and 96.7% for long short-term memory vs 89.9% for support vector machines, 93.9% for extreme gradient boosting, 82.8% for random forests, and 75.2% for simple string matching). Multiple methods were used to visualize the decision-making process of the network, verifying that the networks used similar heuristics to a human annotator. The neural networks were able to classify, with a high degree of accuracy, pathology reports written in unseen formats, suggesting the networks had learned a generalizable encoding of the salient features. CONCLUSION: Neural network-based approaches achieve high performance on organ-level pathology report classification, suggesting that it is feasible to use them within automated tracking systems.© RSNA, 2019Supplemental material is available for this article.See also the commentary by Liu in this issue.

Reducing radiation dose: Equipment, procedure, and operator Perfecting the Trifecta.

This article illustrates the positive impact of fluoroscopic imaging equipment on radiation dose reduction in CTO PCI. The reader should recognize the importance of purchasing and maintaining the best equipment, understanding procedure/patient complexity, and assuring operator training in radiation dose reduction. Future efforts/studies should focus upon all three areas of dose reduction for best results.

The art of delivering evidence-based dual antiplatelet therapy.

This review, which details 2 DAPT risk scoring systems and includes a treatment guide, can help ensure that you deliver the right treatment to the right patients.

2018 ACC/HRS/NASCI/SCAI/SCCT Expert Consensus Document on Optimal Use of Ionizing Radiation in Cardiovascular Imaging-Best Practices for Safety and Effectiveness, Part 2: Radiological Equipment Operation, Dose-Sparing Methodologies, Patient and Medical Personnel Protection.

The stimulus to create this document was the recognition that ionizing radiation-guided cardiovascular procedures are being performed with increasing frequency, leading to greater patient radiation exposure and, potentially, to greater exposure to clinical personnel. While the clinical benefit of these procedures is substantial, there is concern about the implications of medical radiation exposure. ACC leadership concluded that it is important to provide practitioners with an educational resource that assembles and interprets the current radiation knowledge base relevant to cardiovascular procedures. By applying this knowledge base, cardiovascular practitioners will be able to select procedures optimally, and minimize radiation exposure to patients and to clinical personnel. "Optimal Use of Ionizing Radiation in Cardiovascular Imaging - Best Practices for Safety and Effectiveness" is a comprehensive overview of ionizing radiation use in cardiovascular procedures and is published online. To provide the most value to our members, we divided the print version of this document into 2 focused parts. "Part I: Radiation Physics and Radiation Biology" addresses radiation physics, dosimetry and detrimental biologic effects. "Part II: Radiologic Equipment Operation, Dose-Sparing Methodologies, Patient and Medical Personnel Protection" covers the basics of operation and radiation delivery for the 3 cardiovascular imaging modalities (x-ray fluoroscopy, x-ray computed tomography, and nuclear scintigraphy). For each modality, it includes the determinants of radiation exposure and techniques to minimize exposure to both patients and to medical personnel.

2018 ACC/HRS/NASCI/SCAI/SCCT Expert Consensus Document on Optimal Use of Ionizing Radiation in Cardiovascular Imaging-Best Practices for Safety and Effectiveness, Part 1: Radiation Physics and Radiation Biology: A Report of the American College of Cardiology Task Force on Expert Consensus Decision Pathways Developed in Collaboration With Mended Hearts.

The stimulus to create this document was the recognition that ionizing radiation-guided cardiovascular procedures are being performed with increasing frequency, leading to greater patient radiation exposure and, potentially, to greater exposure for clinical personnel. Although the clinical benefit of these procedures is substantial, there is concern about the implications of medical radiation exposure. The American College of Cardiology leadership concluded that it is important to provide practitioners with an educational resource that assembles and interprets the current radiation knowledge base relevant to cardiovascular procedures. By applying this knowledge base, cardiovascular practitioners will be able to select procedures optimally, and minimize radiation exposure to patients and to clinical personnel. Optimal Use of Ionizing Radiation in Cardiovascular Imaging: Best Practices for Safety and Effectiveness is a comprehensive overview of ionizing radiation use in cardiovascular procedures and is published online. To provide the most value to our members, we divided the print version of this document into 2 focused parts. Part I: Radiation Physics and Radiation Biology addresses the issue of medical radiation exposure, the basics of radiation physics and dosimetry, and the basics of radiation biology and radiation-induced adverse effects. Part II: Radiological Equipment Operation, Dose-Sparing Methodologies, Patient and Medical Personnel Protection covers the basics of operation and radiation delivery for the 3 cardiovascular imaging modalities (x-ray fluoroscopy, x-ray computed tomography, and nuclear scintigraphy) and will be published in the next issue of the Journal.

Radiation Safety in the Catheterization Laboratory: Current Perspectives and Practices.

BACKGROUND: There is great variability in radiation safety practices in cardiac catheterization laboratories around the world. METHODS: We performed an international online survey on radiation safety including interventional cardiologists, electrophysiologists, interventional radiologists, and vascular surgeons. RESULTS: A total of 570 responses were received from various geographic locations, including the United States (77.9%), Asia (7.9%), Europe (6.8%), Canada (2.8%), and Mexico and Central America (2.1%). Most respondents (73%) were interventional cardiologists and 23% were electrophysiologists, with 14.4 ± 10.2 years in practice. Most respondents (75%) were not aware of their radiation dose during the past year and 21.2% had never attended a radiation safety course; 58.9% are "somewhat worried" and 31.5% are "very worried" about chronic radiation exposure. Back pain due to lead use was reported by 43.0% and radiation-related health complications including cataracts and malignancies were reported by 6.3%. Only 37.5% of respondents had an established radiation dose threshold for initiating patient follow-up. When comparing United States operators with the other respondents, the former were more likely to attend radiation safety courses (P<.001), wear dosimeters (P<.001), know their annual personal radiation exposure (P<.001), and have an established patient radiation dose threshold (P<.001). They were also more likely to use the fluoro store function, under-table shields, leaded glasses, ceiling lead glass, and disposable radiation shields, and were more concerned about the adverse effects of radiation. CONCLUSIONS: Radiation safety is of concern to catheterization laboratory personnel, yet there is significant variability in radiation protection practices, highlighting several opportunities for standardization and improvement.