The Effectiveness of Image Augmentation in Deep Learning Networks for Detecting COVID-19: A Geometric Transformation Perspective

Frontiers in Medicine · 2021 · 112 citations

• Artificial Intelligence
• Computer Science
• Artificial Intelligence

. This is an interesting finding that may improve current deep learning algorithms using geometrical augmentations for detecting COVID-19. We also provide clinical perspectives on geometric augmentation to consider regarding the development of a robust COVID-19 X-ray-based detector.

A Characteristic Chest Radiographic Pattern in the Setting of the COVID-19 Pandemic

Radiology Cardiothoracic Imaging · 2020 · 69 citations

• Medicine
• Radiology
• Nuclear medicine

PURPOSE: To determine the utility of chest radiography in aiding clinical diagnosis of coronavirus disease 2019 (COVID-19) utilizing reverse-transcription polymerase chain reaction (RT-PCR) as the standard of comparison. MATERIALS AND METHODS: A retrospective study was performed of persons under investigation for COVID-19 presenting to this institution during the exponential growth phase of the COVID-19 outbreak in New Orleans (March 13-25, 2020). Three hundred seventy-six in-hospital chest radiographic examinations for 366 individual patients were reviewed along with concurrent RT-PCR tests. Two experienced radiologists categorized each chest radiograph as characteristic, nonspecific, or negative in appearance for COVID-19, utilizing well-documented COVID-19 imaging patterns. Chest radiograph categorization was compared against RT-PCR results to determine the utility of chest radiography in diagnosing COVID-19. RESULTS: Of the 366 patients, the study consisted of 178 male (49%) and 188 female (51%) patients with a mean age of 52.7 years (range, 17 to 98 years). Of the 376 chest radiographic examinations, 37 (10%) exhibited the characteristic COVID-19 appearance; 215 (57%) exhibited the nonspecific appearance; and 124 (33%) were considered negative for a pulmonary abnormality. Of the 376 RT-PCR tests evaluated, 200 (53%) were positive and 176 (47%) were negative. RT-PCR tests took an average of 2.5 days ± 0.7 to provide results. Sensitivity and specificity for correctly identifying COVID-19 with a characteristic chest radiographic pattern was 15.5% (31/200) and 96.6% (170/176), with a positive predictive value and negative predictive value of 83.8% (31/37) and 50.1% (170/339), respectively. CONCLUSION: The presence of patchy and/or confluent, bandlike ground-glass opacity or consolidation in a peripheral and mid to lower lung zone distribution on a chest radiograph obtained in the setting of pandemic COVID-19 was highly suggestive of severe acute respiratory syndrome coronavirus 2 infection and should be used in conjunction with clinical judgment to make a diagnosis.© RSNA, 2020.