A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot.

BACKGROUND: Cardiac shape modeling is a useful computational tool that has provided quantitative insights into the mechanisms underlying dysfunction in heart disease. The manual input and time required to make cardiac shape models, however, limits their clinical utility. Here we present an end-to-end pipeline that uses deep learning for automated view classification, slice selection, phase selection, anatomical landmark localization, and myocardial image segmentation for the automated generation of three-dimensional, biventricular shape models. With this approach, we aim to make cardiac shape modeling a more robust and broadly applicable tool that has processing times consistent with clinical workflows. METHODS: Cardiovascular magnetic resonance (CMR) images from a cohort of 123 patients with repaired tetralogy of Fallot (rTOF) from two internal sites were used to train and validate each step in the automated pipeline. The complete automated pipeline was tested using CMR images from a cohort of 12 rTOF patients from an internal site and 18 rTOF patients from an external site. Manually and automatically generated shape models from the test set were compared using Euclidean projection distances, global ventricular measurements, and atlas-based shape mode scores. RESULTS: The mean absolute error (MAE) between manually and automatically generated shape models in the test set was similar to the voxel resolution of the original CMR images for end-diastolic models (MAE = 1.9 ± 0.5 mm) and end-systolic models (MAE = 2.1 ± 0.7 mm). Global ventricular measurements computed from automated models were in good agreement with those computed from manual models. The average mean absolute difference in shape mode Z-score between manually and automatically generated models was 0.5 standard deviations for the first 20 modes of a reference statistical shape atlas. CONCLUSIONS: Using deep learning, accurate three-dimensional, biventricular shape models can be reliably created. This fully automated end-to-end approach dramatically reduces the manual input required to create shape models, thereby enabling the rapid analysis of large-scale datasets and the potential to deploy statistical atlas-based analyses in point-of-care clinical settings. Training data and networks are available from cardiacatlas.org.

Deep Learning Subtraction Angiography: Improved Generalizability with Transfer Learning.

PURPOSE: To investigate the utility and generalizability of deep learning subtraction angiography (DLSA) for generating synthetic digital subtraction angiography (DSA) images without misalignment artifacts. MATERIALS AND METHODS: DSA images and native digital angiograms of the cerebral, hepatic, and splenic vasculature, both with and without motion artifacts, were retrospectively collected. Images were divided into a motion-free training set (n = 66 patients, 9,161 images) and a motion artifact-containing test set (n = 22 patients, 3,322 images). Using the motion-free set, the deep neural network pix2pix was trained to produce synthetic DSA images without misalignment artifacts directly from native digital angiograms. After training, the algorithm was tested on digital angiograms of hepatic and splenic vasculature with substantial motion. Four board-certified radiologists evaluated performance via visual assessment using a 5-grade Likert scale. Subgroup analyses were performed to analyze the impact of transfer learning and generalizability to novel vasculature. RESULTS: Compared with the traditional DSA method, the proposed approach was found to generate synthetic DSA images with significantly fewer background artifacts (a mean rating of 1.9 [95% CI, 1.1-2.6] vs 3.5 [3.5-4.4]; P = .01) without a significant difference in foreground vascular detail (mean rating of 3.1 [2.6-3.5] vs 3.3 [2.8-3.8], P = .19) in both the hepatic and splenic vasculature. Transfer learning significantly improved the quality of generated images (P < .001). CONCLUSIONS: DLSA successfully generates synthetic angiograms without misalignment artifacts, is improved through transfer learning, and generalizes reliably to novel vasculature that was not included in the training data.

Machine Learning-Based Analysis and Prediction of Unplanned 30-Day Readmissions After Pituitary Adenoma Resection: A Multi-Institutional Retrospective Study With External Validation.

BACKGROUND: Unplanned readmission after transsphenoidal resection of pituitary adenoma can occur in up to 10% of patients but is unpredictable. OBJECTIVE: To develop a reliable system for predicting unplanned readmission and create a validated method for stratifying patients by risk. METHODS: Data sets were retrospectively collected from the National Surgical Quality Improvement Program and 2 tertiary academic medical centers. Eight machine learning classifiers were fit to the National Surgical Quality Improvement Program data, optimized using Bayesian parameter optimization and evaluated on the external data. Permutation analysis identified the relative importance of predictive variables, and a risk stratification system was built using the trained machine learning models. RESULTS: Readmissions were accurately predicted by several classification models with an area under the receiving operator characteristic curve of 0.76 (95% CI 0.68-0.83) on the external data set. Permutation analysis identified the most important variables for predicting readmission as preoperative sodium level, returning to the operating room, and total operation time. High-risk and medium-risk patients, as identified by the proposed risk stratification system, were more likely to be readmitted than low-risk patients, with relative risks of 12.2 (95% CI 5.9-26.5) and 4.2 (95% CI 2.3-8.7), respectively. Overall risk stratification showed high discriminative capability with a C-statistic of 0.73. CONCLUSION: In this multi-institutional study with outside validation, unplanned readmissions after pituitary adenoma resection were accurately predicted using machine learning techniques. The features identified in this study and the risk stratification system developed could guide clinical and surgical decision making, reduce healthcare costs, and improve the quality of patient care by better identifying high-risk patients for closer perioperative management.

Comparison of International Classification of Diseases and Related Health Problems, Tenth Revision Codes With Electronic Medical Records Among Patients With Symptoms of Coronavirus Disease 2019.

Importance: International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes are used to characterize coronavirus disease 2019 (COVID-19)-related symptoms. Their accuracy is unknown, which could affect downstream analyses. Objective: To compare the performance of fever-, cough-, and dyspnea-specific ICD-10 codes with medical record review among patients tested for COVID-19. Design, Setting, and Participants: This cohort study included patients who underwent quantitative reverse transcriptase-polymerase chain reaction testing for severe acute respiratory syndrome coronavirus 2 at University of Utah Health from March 10 to April 6, 2020. Data analysis was performed in April 2020. Main Outcomes and Measures: The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of ICD-10 codes for fever (R50*), cough (R05*), and dyspnea (R06.0*) were compared with manual medical record review. Performance was calculated overall and stratified by COVID-19 test result, sex, age group (<50, 50-64, and >64 years), and inpatient status. Bootstrapping was used to generate 95% CIs, and Pearson χ2 tests were used to compare different subgroups. Results: Among 2201 patients tested for COVD-19, the mean (SD) age was 42 (17) years; 1201 (55%) were female, 1569 (71%) were White, and 282 (13%) were Hispanic or Latino. The prevalence of fever was 66% (1444 patients), that of cough was 88% (1930 patients), and that of dyspnea was 64% (1399 patients). For fever, the sensitivity of ICD-10 codes was 0.26 (95% CI, 0.24-0.29), specificity was 0.98 (95% CI, 0.96-0.99), PPV was 0.96 (95% CI, 0.93-0.97), and NPV was 0.41 (95% CI, 0.39-0.43). For cough, the sensitivity of ICD-10 codes was 0.44 (95% CI, 0.42-0.46), specificity was 0.88 (95% CI, 0.84-0.92), PPV was 0.96 (95% CI, 0.95-0.97), and NPV was 0.18 (95% CI, 0.16-0.20). For dyspnea, the sensitivity of ICD-10 codes was 0.24 (95% CI, 0.22-0.26), specificity was 0.97 (95% CI, 0.96-0.98), PPV was 0.93 (95% CI, 0.90-0.96), and NPV was 0.42 (95% CI, 0.40-0.44). ICD-10 code performance was better for inpatients than for outpatients for fever (χ2 = 41.30; P < .001) and dyspnea (χ2 = 14.25; P = .003) but not for cough (χ2 = 5.13; P = .16). Conclusions and Relevance: These findings suggest that ICD-10 codes lack sensitivity and have poor NPV for symptoms associated with COVID-19. This inaccuracy has implications for any downstream data model, scientific discovery, or surveillance that relies on these codes.