Michael W Sjoding1, Daniel Taylor2, Jonathan Motyka2, Elizabeth Lee3, Ivan Co4, Dru Claar5, Jakob I McSparron6, Sardar Ansari7, Meeta Prasad Kerlin8, John P Reilly9, Michael G S Shashaty8, Brian J Anderson8, Tiffanie K Jones9, Harrison M Drebin9, Caroline A G Ittner9, Nuala J Meyer9, Theodore J Iwashyna10, Kevin R Ward11, Christopher E Gillies12. 1. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Center for Computational Medicine and Bioinformatics, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA; Institute for Healthcare Policy and Innovation, University of Michigan, Ann Arbor, MI, USA. Electronic address: msjoding@umich.edu. 2. Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA. 3. Department of Radiology, University of Michigan Medical School, Ann Arbor, MI, USA. 4. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Department of Emergency Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA. 5. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA. 6. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA. 7. Department of Emergency Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA. 8. Division of Pulmonary, Allergy, and Critical Care Medicine, Department of Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA. 9. Division of Pulmonary, Allergy, and Critical Care Medicine, Department of Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA; Center for Translational Lung Biology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA. 10. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; VA Center for Clinic Management Research, Ann Arbor, MI, USA; Institute for Social Research, Ann Arbor, MI, USA. 11. Center for Computational Medicine and Bioinformatics, University of Michigan Medical School, Ann Arbor, MI, USA; Department of Emergency Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA. 12. Department of Emergency Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; Michigan Center for Integrative Research in Critical Care; Ann Arbor, MI, USA; Michigan Institute for Data Science, University of Michigan, Ann Arbor, MI, USA.
Abstract
BACKGROUND: Acute respiratory distress syndrome (ARDS) is a common, but under-recognised, critical illness syndrome associated with high mortality. An important factor in its under-recognition is the variability in chest radiograph interpretation for ARDS. We sought to train a deep convolutional neural network (CNN) to detect ARDS findings on chest radiographs. METHODS: CNNs were pretrained on 595 506 radiographs from two centres to identify common chest findings (eg, opacity and effusion), and then trained on 8072 radiographs annotated for ARDS by multiple physicians using various transfer learning approaches. The best performing CNN was tested on chest radiographs in an internal and external cohort, including a subset reviewed by six physicians, including a chest radiologist and physicians trained in intensive care medicine. Chest radiograph data were acquired from four US hospitals. FINDINGS: In an internal test set of 1560 chest radiographs from 455 patients with acute hypoxaemic respiratory failure, a CNN could detect ARDS with an area under the receiver operator characteristics curve (AUROC) of 0·92 (95% CI 0·89-0·94). In the subgroup of 413 images reviewed by at least six physicians, its AUROC was 0·93 (95% CI 0·88-0·96), sensitivity 83·0% (95% CI 74·0-91·1), and specificity 88·3% (95% CI 83·1-92·8). Among images with zero of six ARDS annotations (n=155), the median CNN probability was 11%, with six (4%) assigned a probability above 50%. Among images with six of six ARDS annotations (n=27), the median CNN probability was 91%, with two (7%) assigned a probability below 50%. In an external cohort of 958 chest radiographs from 431 patients with sepsis, the AUROC was 0·88 (95% CI 0·85-0·91). When radiographs annotated as equivocal were excluded, the AUROC was 0·93 (0·92-0·95). INTERPRETATION: A CNN can be trained to achieve expert physician-level performance in ARDS detection on chest radiographs. Further research is needed to evaluate the use of these algorithms to support real-time identification of ARDS patients to ensure fidelity with evidence-based care or to support ongoing ARDS research. FUNDING: National Institutes of Health, Department of Defense, and Department of Veterans Affairs.
BACKGROUND:Acute respiratory distress syndrome (ARDS) is a common, but under-recognised, critical illness syndrome associated with high mortality. An important factor in its under-recognition is the variability in chest radiograph interpretation for ARDS. We sought to train a deep convolutional neural network (CNN) to detect ARDS findings on chest radiographs. METHODS: CNNs were pretrained on 595 506 radiographs from two centres to identify common chest findings (eg, opacity and effusion), and then trained on 8072 radiographs annotated for ARDS by multiple physicians using various transfer learning approaches. The best performing CNN was tested on chest radiographs in an internal and external cohort, including a subset reviewed by six physicians, including a chest radiologist and physicians trained in intensive care medicine. Chest radiograph data were acquired from four US hospitals. FINDINGS: In an internal test set of 1560 chest radiographs from 455 patients with acute hypoxaemic respiratory failure, a CNN could detect ARDS with an area under the receiver operator characteristics curve (AUROC) of 0·92 (95% CI 0·89-0·94). In the subgroup of 413 images reviewed by at least six physicians, its AUROC was 0·93 (95% CI 0·88-0·96), sensitivity 83·0% (95% CI 74·0-91·1), and specificity 88·3% (95% CI 83·1-92·8). Among images with zero of six ARDS annotations (n=155), the median CNN probability was 11%, with six (4%) assigned a probability above 50%. Among images with six of six ARDS annotations (n=27), the median CNN probability was 91%, with two (7%) assigned a probability below 50%. In an external cohort of 958 chest radiographs from 431 patients with sepsis, the AUROC was 0·88 (95% CI 0·85-0·91). When radiographs annotated as equivocal were excluded, the AUROC was 0·93 (0·92-0·95). INTERPRETATION: A CNN can be trained to achieve expert physician-level performance in ARDS detection on chest radiographs. Further research is needed to evaluate the use of these algorithms to support real-time identification of ARDSpatients to ensure fidelity with evidence-based care or to support ongoing ARDS research. FUNDING: National Institutes of Health, Department of Defense, and Department of Veterans Affairs.
Authors: Mark P Khurana; Daniel E Raaschou-Pedersen; Jørgen Kurtzhals; Jakob E Bardram; Sisse R Ostrowski; Johan S Bundgaard Journal: BMC Med Educ Date: 2022-02-26 Impact factor: 2.463