Practice Pattern, Diagnostic Yield, and Long-Term Prognostic Impact of Coronary Computed Tomographic Angiography

The natural language processing of radiology requests and reports of chest imaging: Comparing five transformer models’ multilabel classification and a proof-of-concept study

Background

Radiology requests and reports contain valuable information about diagnostic findings and indications, and transformer-based language models are promising for more accurate text classification.

Methods

In a retrospective study, 2256 radiologist-annotated radiology requests (8 classes) and reports (10 classes) were divided into training and testing datasets (90% and 10%, respectively) and used to train 32 models. Performance metrics were compared by model type (LSTM, Bertje, RobBERT, BERT-clinical, BERT-multilingual, BERT-base), text length, data prevalence, and training strategy. The best models were used to predict the remaining 40,873 cases’ categories of the datasets of requests and reports.

Results

The RobBERT model performed the best after 4000 training iterations, resulting in AUC values ranging from 0.808 [95% CI (0.757–0.859)] to 0.976 [95% CI (0.956–0.996)] for the requests and 0.746 [95% CI (0.689–0.802)] to 1.0 [95% CI (1.0–1.0)] for the reports. The AUC for the classification of normal reports was 0.95 [95% CI (0.922–0.979)]. The predicted data demonstrated variability of both diagnostic yield for various request classes and request patterns related to COVID-19 hospital admission data.

Conclusion

Transformer-based natural language processing is feasible for the multilabel classification of chest imaging request and report items. Diagnostic yield varies with the information in the requests.

Use of Coronary CT Angiography to Predict Obstructive Lesions in Patients with Chest Pain without Enzyme and ST-Segment Elevation

It is challenging to rule out acute coronary syndrome among chest pain patients without both ST-segment elevation in electrocardiography and troponin elevation at emergency departments (ED). The purpose of this study was to develop a prediction model for rapidly determining the occurrence of significant stenosis in coronary computed tomography angiography (CCTA). Retrospective observational cohort study was conducted with 904 patients who had presented with chest pain without troponin elevation and ST-segment changes and underwent CCTA between January 2017 and December 2018. The primary endpoint was the presence of significant stenosis on CCTA, defined as narrowing above 70% diameter. The logistic regression model was used for development a new predictive model. One hundred and thirty-four patients (14.8%) were shown severe stenosis. The independent associated factors for significant stenosis were age ≥65 years, male, diabetes, history of acute coronary syndrome, and typical chest pain. Based these results, we developed a new prediction model. The area under the curve was 0.782 (95% confidence interval 0.742-0.822). Moreover, score of ≥5 was chosen as cut-off values with 86.6% sensitivity and 56.4% specificity. In conclusion, among chest pain patients without ST changes and troponin elevation, the new score will be helpful to identify potential candidate for CCTA such as patients with significant stenosis.

Practice Pattern, Diagnostic Yield, and Long-Term Prognostic Impact of Coronary Computed Tomographic Angiography

Background Although guidelines recommend the use of coronary computed tomographic angiography (CTA) in patients with stable pain syndromes, the clinical benefits of the use of coronary CTA in a broad spectrum of patients is unknown. We evaluated the contemporary practice pattern and diagnostic yield of coronary CTA and their impact on the subsequent diagnostic-therapeutic cascade and clinical outcomes. Methods and Results We identified 39 906 patients without known coronary artery disease (CAD) who underwent coronary CTA between January 2007 and December 2013. The patients' demographic characteristics, risk factors, symptoms, results of coronary CTA, the appropriateness of downstream diagnostic and therapeutic interventions, and long-term outcomes (death or myocardial infarction) were evaluated. The number of coronary CTAs had increased over time, especially in asymptomatic patients. Coronary CTA revealed that 6108 patients (15.3%) had obstructive CAD (23.7% of symptomatic and 9.3% of asymptomatic patients). Subsequent cardiac catheterization was performed in 19.2% of symptomatic patients (appropriate, 80.6%) and in 3.9% of asymptomatic patients (appropriate, 7.9%). The 5-year rate of death or myocardial infarction was significantly higher in patients with obstructive CAD on CTA than those without (7.2% versus 3.0%; P<0.001; adjusted hazard ratio [95% CI], 1.34 [1.17-1.54]). However, obstructive CAD on CTA had limited added value over conventional risk factors for predicting death or myocardial infarction. Conclusions Although the use of coronary CTA had substantially increased, CTA had a low diagnostic yield for obstructive CAD, especially in asymptomatic patients. The use of CTA in asymptomatic patients seemed to have led to inappropriate subsequent diagnostic or therapeutic interventions without clinical benefit.