Deep learning to detect left ventricular structural abnormalities in chest X-rays.

BACKGROUND AND AIMS: Early identification of cardiac structural abnormalities indicative of heart failure is crucial to improving patient outcomes. Chest X-rays (CXRs) are routinely conducted on a broad population of patients, presenting an opportunity to build scalable screening tools for structural abnormalities indicative of Stage B or worse heart failure with deep learning methods. In this study, a model was developed to identify severe left ventricular hypertrophy (SLVH) and dilated left ventricle (DLV) using CXRs. METHODS: A total of 71 589 unique CXRs from 24 689 different patients completed within 1 year of echocardiograms were identified. Labels for SLVH, DLV, and a composite label indicating the presence of either were extracted from echocardiograms. A deep learning model was developed and evaluated using area under the receiver operating characteristic curve (AUROC). Performance was additionally validated on 8003 CXRs from an external site and compared against visual assessment by 15 board-certified radiologists. RESULTS: The model yielded an AUROC of 0.79 (0.76-0.81) for SLVH, 0.80 (0.77-0.84) for DLV, and 0.80 (0.78-0.83) for the composite label, with similar performance on an external data set. The model outperformed all 15 individual radiologists for predicting the composite label and achieved a sensitivity of 71% vs. 66% against the consensus vote across all radiologists at a fixed specificity of 73%. CONCLUSIONS: Deep learning analysis of CXRs can accurately detect the presence of certain structural abnormalities and may be useful in early identification of patients with LV hypertrophy and dilation. As a resource to promote further innovation, 71 589 CXRs with adjoining echocardiographic labels have been made publicly available.

COVID-19 on Chest CT: Translating Known Microscopic Findings to Imaging Observations.

PURPOSE: To describe the imaging findings of COVID-19 and correlate them with their known pathology observations. METHODS: This is an IRB-approved retrospective study performed at Columbia University Irving Medical Center (IRB # AAAS9652) that included symptomatic adult patients (21 years or older) who presented to our emergency room and tested positive for COVID-19 and were either admitted or discharged with at least one chest CT from 11 March 2020 through 1 July 2020. CT scans were ordered by the physicians caring for the patients; our COVID-19 care protocols did not specify the timing for chest CT scans. A scoring system was used to document the extent of pulmonary involvement. The total CT grade was the sum of the individual lobar grades and ranged from 0 (no involvement) to 16 (maximum involvement). The distribution of lung abnormalities was described as peripheral (involving the outer one-third of the lung), central (inner two-thirds of the lung), or both. Additional CT findings, including the presence of pleural fluid, atelectasis, fibrosis, cysts, and pneumothorax, were recorded. Contrast-enhanced CT scans were evaluated for the presence of a pulmonary embolism, while non-contrast chest CT scans were evaluated for hyperdense vessels. RESULTS: 209 patients with 232 CT scans met the inclusion criteria. The average age was 61 years (range 23-97 years), and 56% of the patients were male. The average score reflecting the extent of the disease on the CT was 10.2 (out of a potential grade of 16). Further, 73% of the patients received contrast, which allowed the identification of a pulmonary embolism in 21%. Of those without contrast, 33% had hyperdense vessels, which might suggest a chronic pulmonary embolism. Further, 47% had peripheral opacities and 9% had a Hampton's hump, and 78% of the patients had central consolidation, while 28% had round consolidations. Atelectasis was, overall, infrequent at 5%. Fibrosis was observed in 11% of those studied, with 6% having cysts and 3% pneumothorax. CONCLUSIONS: The CT manifestations of COVID-19 can be divided into findings related to endothelial and epithelial injury, as were seen on prior post-mortem reports. Endothelial injury may benefit from treatments to stabilize the endothelium. Epithelial injury is more prone to developing pulmonary fibrotic changes.