RESUMEN
Purpose: To develop radiology domain-specific bidirectional encoder representations from transformers (BERT) models that can identify speech recognition (SR) errors and suggest corrections in radiology reports. Materials and Methods: A pretrained BERT model, Clinical BioBERT, was further pretrained on a corpus of 114 008 radiology reports between April 2016 and August 2019 that were retrospectively collected from two hospitals. Next, the model was fine-tuned on a training dataset of generated insertion, deletion, and substitution errors, creating Radiology BERT. This model was retrospectively evaluated on an independent dataset of radiology reports with generated errors (n = 18 885) and on unaltered report sentences (n = 2000) and prospectively evaluated on true clinical SR errors (n = 92). Correction Radiology BERT was separately trained to suggest corrections for detected deletion and substitution errors. Area under the receiver operating characteristic curve (AUC) and bootstrapped 95% CIs were calculated for each evaluation dataset. Results: Radiology-specific BERT had AUC values of >.99 (95% CI: >0.99, >0.99), 0.94 (95% CI: 0.93, 0.94), 0.98 (95% CI: 0.98, 0.98), and 0.97 (95% CI: 0.97, 0.97) for detecting insertion, deletion, substitution, and all errors, respectively, on the independently generated test set. Testing on unaltered report impressions revealed a sensitivity of 82% (28 of 34; 95% CI: 70%, 93%) and specificity of 88% (1521 of 1728; 95% CI: 87%, 90%). Testing on prospective SR errors showed an accuracy of 75% (69 of 92; 95% CI: 65%, 83%). Finally, the correct word was the top suggestion for 45.6% (475 of 1041; 95% CI: 42.5%, 49.3%) of errors. Conclusion: Radiology-specific BERT models fine-tuned on generated errors were able to identify SR errors in radiology reports and suggest corrections.Keywords: Computer Applications, Technology Assessment Supplemental material is available for this article. © RSNA, 2022See also the commentary by Abajian and Cheung in this issue.
RESUMEN
Our objective was to develop deep learning models with chest radiograph data to predict healthcare costs and classify top-50% spenders. 21,872 frontal chest radiographs were retrospectively collected from 19,524 patients with at least 1-year spending data. Among the patients, 11,003 patients had 3 years of cost data, and 1678 patients had 5 years of cost data. Model performances were measured with area under the receiver operating characteristic curve (ROC-AUC) for classification of top-50% spenders and Spearman ρ for prediction of healthcare cost. The best model predicting 1-year (N = 21,872) expenditure achieved ROC-AUC of 0.806 [95% CI 0.793-0.819] for top-50% spender classification and ρ of 0.561 [0.536-0.586] for regression. Similarly, for predicting 3-year (N = 12,395) expenditure, ROC-AUC of 0.771 [0.750-0.794] and ρ of 0.524 [0.489-0.559]; for predicting 5-year (N = 1779) expenditure ROC-AUC of 0.729 [0.667-0.729] and ρ of 0.424 [0.324-0.529]. Our deep learning model demonstrated the feasibility of predicting health care expenditure as well as classifying top 50% healthcare spenders at 1, 3, and 5 year(s), implying the feasibility of combining deep learning with information-rich imaging data to uncover hidden associations that may allude to physicians. Such a model can be a starting point of making an accurate budget in reimbursement models in healthcare industries.
Asunto(s)
Aprendizaje Profundo , Atención a la Salud , Humanos , Proyectos Piloto , Curva ROC , Radiografía , Estudios RetrospectivosRESUMEN
BACKGROUND: The comprehensiveness and maintenance of the American College of Radiology (ACR) Appropriateness Criteria (AC) makes it a unique resource for evidence-based clinical imaging decision support, but it is underutilized by clinicians. To facilitate the use of imaging recommendations, we develop a natural language processing (NLP) search algorithm that automatically matches clinical indications that physicians write into imaging orders to appropriate AC imaging recommendations. METHODS: We apply a hybrid model of semantic similarity from a sent2vec model trained on 223 million scientific sentences, combined with term frequency inverse document frequency features. AC documents are ranked based on their embeddings' cosine distance to query. For model testing, we compiled a dataset of simulated simple and complex indications for each AC document (n = 410) and another with clinical indications from randomly sampled radiology reports (n = 100). We compare our algorithm to a custom google search engine. RESULTS: On the simulated indications, our algorithm ranked ground truth documents as top 3 for 98% of simple queries and 85% of complex queries. Similarly, on the randomly sampled radiology report dataset, the algorithm ranked 86% of indications with a single match as top 3. Vague and distracting phrases present in the free-text indications were main sources of errors. Our algorithm provides more relevant results than a custom Google search engine, especially for complex queries. CONCLUSIONS: We have developed and evaluated an NLP algorithm that matches clinical indications to appropriate AC guidelines. This approach can be integrated into imaging ordering systems for automated access to guidelines.
Asunto(s)
Diagnóstico por Imagen/métodos , Procesamiento de Lenguaje Natural , Radiología/métodos , Adolescente , Adulto , Anciano , Femenino , Humanos , Masculino , Persona de Mediana Edad , Motor de Búsqueda , Semántica , Adulto JovenRESUMEN
RATIONALE AND OBJECTIVES: Radiology turnaround time is an important quality measure that can impact hospital workflow and patient outcomes. We aimed to develop a machine learning model to predict delayed turnaround time during non-business hours and identify factors that contribute to this delay. MATERIALS AND METHODS: This retrospective study consisted of 15,117 CT cases from May 2018 to May 2019 during non-business hours at two hospital campuses after applying exclusion criteria. Of these 15,177 cases, 7,532 were inpatient cases and 7,585 were emergency cases. Order time, scan time, first communication by radiologist, free-text indications, and other clinical metadata were extracted. A combined XGBoost classifier and Random Forest natural language processing model was trained with 85% of the data and tested with 15% of the data. The model predicted two measures of delay: when the exam was ordered to first communication (total time) and when the scan was completed to first communication (interpretation time). The model was analyzed with the area under the curve (AUC) of receiver operating characteristic (ROC) and feature importance. Source code: https://bit.ly/2UrLiVJ RESULTS: The algorithm reached an AUC of 0.85, with a 95% confidence interval [0.83, 0.87], when predicting delays greater than 245 minutes for "total time" and 0.71, with a 95% confidence interval [0.68, 0.73], when predicting delays greater than 57 minutes for "interpretation time". At our institution, CT scan description (e.g. "CTA chest pulmonary embolism protocol"), time of day, and year in training were more predictive features compared to body part, inpatient status, and hospital campus for both interpretation and total time delay. CONCLUSION: This algorithm can be applied clinically when a physician is ordering the scan to reasonably predict delayed turnaround time. Such a model can be leveraged to identify factors associated with delays and emphasize areas for improvement to patient outcomes.
Asunto(s)
Radiología , Humanos , Aprendizaje Automático , Curva ROC , Radiografía , Estudios RetrospectivosRESUMEN
BACKGROUND: A systematic approach to MRI protocol assignment is essential for the efficient delivery of safe patient care. Advances in natural language processing (NLP) allow for the development of accurate automated protocol assignment. We aim to develop, evaluate, and deploy an NLP model that automates protocol assignment, given the clinician indication text. METHODS: We collected 7139 spine MRI protocols (routine or contrast) and 990 head MRI protocols (routine brain, contrast brain, or other) from a single institution. Protocols were split into training (n = 4997 for spine MRI; n = 839 for head MRI), validation (n = 1071 for spine MRI, fivefold cross-validation used for head MRI), and test (n = 1071 for spine MRI; n = 151 for head MRI) sets. fastText and XGBoost were used to develop 2 NLP models to classify spine and head MRI protocols, respectively. A Flask-based web app was developed to be deployed via Heroku. RESULTS: The spine MRI model had an accuracy of 83.38% and a receiver operator characteristic area under the curve (ROC-AUC) of 0.8873. The head MRI model had an accuracy of 85.43% with a routine brain protocol ROC-AUC of 0.9463 and contrast brain protocol ROC-AUC of 0.9284. Cancer, infectious, and inflammatory related keywords were associated with contrast administration. Structural anatomic abnormalities and stroke/altered mental status were indicative of routine spine and brain MRI, respectively. Error analysis revealed increasing the sample size may improve performance for head MRI protocols. A web version of the model is provided for demonstration and deployment. CONCLUSION: We developed and web-deployed two NLP models that accurately predict spine and head MRI protocol assignment, which could improve radiology workflow efficiency.
Asunto(s)
Procesamiento de Lenguaje Natural , Radiología , Humanos , Imagen por Resonancia Magnética , Radiografía , Flujo de TrabajoRESUMEN
BACKGROUND: Reidentification of prior nodules for temporal comparison is an important but time-consuming step in lung cancer screening. We develop and evaluate an automated nodule detector that utilizes the axial-slice number of nodules found in radiology reports to generate high precision nodule predictions. METHODS: 888 CTs from Lung Nodule Analysis were used to train a 2-dimensional (2D) object detection neural network. A pipeline of 2D object detection, 3D unsupervised clustering, false positive reduction, and axial-slice numbers were used to generate nodule candidates. 47 CTs from the National Lung Cancer Screening Trial (NLST) were used for model evaluation. RESULTS: Our nodule detector achieved a precision of 0.962 at a recall of 0.573 on the NLST test set for any nodule. When adjusting for unintended nodule predictions, we achieved a precision of 0.931 at a recall 0.561, which corresponds to 0.06 false positives per CT. Error analysis revealed better detection of nodules with soft tissue attenuation compared to ground glass and undeterminable attenuation. Nodule margins, size, location, and patient demographics did not differ between correct and incorrect predictions. CONCLUSIONS: Utilization of axial-slice numbers from radiology reports allowed for development of a lung nodule detector with a low false positive rate compared to prior feature-engineering and machine learning approaches. This high precision nodule detector can reduce time spent on reidentification of prior nodules during lung cancer screening and can rapidly develop new institutional datasets to explore novel applications of computer vision in lung cancer imaging.
Asunto(s)
Neoplasias Pulmonares/diagnóstico por imagen , Aprendizaje Automático , Nódulo Pulmonar Solitario/diagnóstico por imagen , Tomografía Computarizada por Rayos X/métodos , Reacciones Falso Positivas , Femenino , Humanos , Masculino , Persona de Mediana Edad , Redes Neurales de la Computación , Carga TumoralRESUMEN
BACKGROUND: There has been increasing interest in machine learning based natural language processing (NLP) methods in radiology; however, models have often used word embeddings trained on general web corpora due to lack of a radiology-specific corpus. PURPOSE: We examined the potential of Radiopaedia to serve as a general radiology corpus to produce radiology specific word embeddings that could be used to enhance performance on a NLP task on radiological text. MATERIALS AND METHODS: Embeddings of dimension 50, 100, 200, and 300 were trained on articles collected from Radiopaedia using a GloVe algorithm and evaluated on analogy completion. A shallow neural network using input from either our trained embeddings or pre-trained Wikipedia 2014 + Gigaword 5 (WG) embeddings was used to label the Radiopaedia articles. Labeling performance was evaluated based on exact match accuracy and Hamming loss. The McNemar's test with continuity and the Benjamini-Hochberg correction and a 5×2 cross validation paired two-tailed t-test were used to assess statistical significance. RESULTS: For accuracy in the analogy task, 50-dimensional (50-D) Radiopaedia embeddings outperformed WG embeddings on tumor origin analogies (p < 0.05) and organ adjectives (p < 0.01) whereas WG embeddings tended to outperform on inflammation location and bone vs. muscle analogies (p < 0.01). The two embeddings had comparable performance on other subcategories. In the labeling task, the Radiopaedia-based model outperformed the WG based model at 50, 100, 200, and 300-D for exact match accuracy (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively) and Hamming loss (p < 0.001, p < 0.001, p < 0.01, and p < 0.05, respectively). CONCLUSION: We have developed a set of word embeddings from Radiopaedia and shown that they can preserve relevant medical semantics and augment performance on a radiology NLP task. Our results suggest that the cultivation of a radiology-specific corpus can benefit radiology NLP models in the future.
Asunto(s)
Procesamiento de Lenguaje Natural , Radiología , Aprendizaje Automático , Semántica , Unified Medical Language SystemRESUMEN
Although machine learning (ML) has made significant improvements in radiology, few algorithms have been integrated into clinical radiology workflow. Complex radiology IT environments and Picture Archiving and Communication System (PACS) pose unique challenges in creating a practical ML schema. However, clinical integration and testing are critical to ensuring the safety and accuracy of ML algorithms. This study aims to propose, develop, and demonstrate a simple, efficient, and understandable hardware and software system for integrating ML models into the standard radiology workflow and PACS that can serve as a framework for testing ML algorithms. A Digital Imaging and Communications in Medicine/Graphics Processing Unit (DICOM/GPU) server and software pipeline was established at a metropolitan county hospital intranet to demonstrate clinical integration of ML algorithms in radiology. A clinical ML integration schema, agnostic to the hospital IT system and specific ML models/frameworks, was implemented and tested with a breast density classification algorithm and prospectively evaluated for time delays using 100 digital 2D mammograms. An open-source clinical ML integration schema was successfully implemented and demonstrated. This schema allows for simple uploading of custom ML models. With the proposed setup, the ML pipeline took an average of 26.52 s per second to process a batch of 100 studies. The most significant processing time delays were noted in model load and study stability times. The code is made available at " http://bit.ly/2Z121hX ". We demonstrated the feasibility to deploy and utilize ML models in radiology without disrupting existing radiology workflow.
Asunto(s)
Sistemas de Información Radiológica , Radiología , Programas Informáticos , Inteligencia Artificial , Humanos , Integración de Sistemas , Flujo de TrabajoRESUMEN
OBJECTIVE: Differentiate high- versus low-volume radiologists who interpret neurological (Neuro) MRI or musculoskeletal (MSK) MRI and measure the proportion of Neuro and MSK MRIs read by low-volume radiologists. METHODS: We queried the 2015 Medicare Physician and Other Supplier Public Use File for radiologists who submitted claims for Neuro or MSK MRIs. Radiologists were classified as high-volume versus low-volume based on their work relative value units (wRVUs) focus or volume of studies interpreted using three different methodologies: Method 1, percentage of wRVUs in Neuro or MSK MRI; Method 2, absolute number of Neuro or MSK MRIs interpreted; and Method 3, both percentage and absolute number. Multiple thresholds with each methodology were tested, and the percentage of Neuro or MSK MRIs interpreted by low-volume radiologists was calculated for each threshold. RESULTS: With Method 1, 33% of Neuro MRI and 50% of MSK MRI studies were interpreted by a radiologist whose wRVUs in Neuro or MSK MRI were less than 20% (Method 1). With Method 2, 22% of Neuro MRIs and 37% of MSK MRIs were interpreted by radiologists who read fewer than the mean number of Neuro or MSK MRIs interpreted by an "average full-time radiologist" whose wRVUs in Neuro or MSK MRI were approximately 20%. With Method 3, 38% of Neuro MRIs and 57% of MSK MRIs were interpreted by "low-volume" radiologists. If instead a 50% wRVU threshold is used for Methods One, Two, and Three, then 70%, 58%, and 77% of Neuro MRIs and 86%, 80%, and 90% of MSK MRIs are read by low-volume radiologists. DISCUSSION: A large number of radiologists read a low volume of Neuro or MSK MRIs; these low-volume Neuro or MSK MRI radiologists read a substantial portion of Neuro or MSK MRIs. It is unknown which of the methods for distinguishing low-volume radiologists, combined with which threshold, may best correlate with high-performing or low-performing radiologists.
Asunto(s)
Medicare , Radiología , Anciano , Humanos , Imagen por Resonancia Magnética , Radiografía , Radiólogos , Estados UnidosRESUMEN
Magnetic resonance imaging (MRI) protocoling can be time- and resource-intensive, and protocols can often be suboptimal dependent upon the expertise or preferences of the protocoling radiologist. Providing a best-practice recommendation for an MRI protocol has the potential to improve efficiency and decrease the likelihood of a suboptimal or erroneous study. The goal of this study was to develop and validate a machine learning-based natural language classifier that can automatically assign the use of intravenous contrast for musculoskeletal MRI protocols based upon the free-text clinical indication of the study, thereby improving efficiency of the protocoling radiologist and potentially decreasing errors. We utilized a deep learning-based natural language classification system from IBM Watson, a question-answering supercomputer that gained fame after challenging the best human players on Jeopardy! in 2011. We compared this solution to a series of traditional machine learning-based natural language processing techniques that utilize a term-document frequency matrix. Each classifier was trained with 1240 MRI protocols plus their respective clinical indications and validated with a test set of 280. Ground truth of contrast assignment was obtained from the clinical record. For evaluation of inter-reader agreement, a blinded second reader radiologist analyzed all cases and determined contrast assignment based on only the free-text clinical indication. In the test set, Watson demonstrated overall accuracy of 83.2% when compared to the original protocol. This was similar to the overall accuracy of 80.2% achieved by an ensemble of eight traditional machine learning algorithms based on a term-document matrix. When compared to the second reader's contrast assignment, Watson achieved 88.6% agreement. When evaluating only the subset of cases where the original protocol and second reader were concordant (n = 251), agreement climbed further to 90.0%. The classifier was relatively robust to spelling and grammatical errors, which were frequent. Implementation of this automated MR contrast determination system as a clinical decision support tool may save considerable time and effort of the radiologist while potentially decreasing error rates, and require no change in order entry or workflow.
Asunto(s)
Aumento de la Imagen/métodos , Interpretación de Imagen Asistida por Computador/métodos , Imagen por Resonancia Magnética/métodos , Enfermedades Musculoesqueléticas/diagnóstico por imagen , Procesamiento de Lenguaje Natural , Algoritmos , Medios de Contraste/administración & dosificación , Humanos , Inyecciones Intravenosas , Sistema Musculoesquelético/diagnóstico por imagen , Reproducibilidad de los Resultados , Estudios RetrospectivosRESUMEN
Pneumocystis jirovecii pneumonia (PCP) typically presents as an interstitial and alveolar process with ground glass opacities on chest computed tomography (CT). The absence of ground glass opacities on chest CT is thought to have a high negative predictive value for PCP in individuals with AIDS. Here, we report a case of PCP in a man with AIDS who presented to our hospital with subacute shortness of breath and a nonproductive cough. While his chest CT revealed diffuse nodular rather than ground glass opacities, bronchoscopy with bronchoalveolar lavage and transbronchial biopsies confirmed the diagnosis of PCP and did not identify additional pathogens. PCP was not the expected diagnosis based on chest CT, but it otherwise fit well with the patient's clinical and laboratory presentation. In the era of combination antiretroviral therapy, routine prophylaxis for PCP, and increased use of computed tomography, it may be that PCP will increasingly present with nonclassical chest radiographic patterns. Clinicians should be aware of this presentation when selecting diagnostic and management strategies.