DETECT-ARDS - Analytic for Detecting Acute Respiratory Distress Syndrome
Value Proposition
DETECT-ARDS is a new approach for identifying ARDS findings on chest x-rays. With ARDS often missed or under-diagnosed, DETECT_ARDS has the potential to transform patient outcomes for the better. By training powerful algorithms called deep convolutional neural networks (CNNs), our system can identify findings consistent with ARDS with high accuracy.
Competitive Advantage
Deep learning networks trained using transfer learning for ARDS detection will be a fundamental leap forward in ARDS detection and care
Can provide critical diagnostics, enabling rapid identification and triage of ARDS patients
Ensures prompt treatment, leading to improved patient outcomes
Unique Features
Integrates with EHR and bedside-monitoring data
Simple interface
Predicts ARDS risk
Principal Investigators
Michael Sjoding, MD
Sardar Ansari, PhD
Intellectual Property
Invention Disclosure # 2020-026
Patent Application Submitted
Solution Sheet
Download Solution Sheet (PDF)
According to a 2016 WHO study, over 3.6 Billion x-rays are performed each year, and 40% of them are chest x-rays.
Acute respiratory distress syndrome (ARDS) is a common, yet under-recognised, critical illness syndrome associated with high mortality. An important factor in its under-recognition is the variability in chest x-ray interpretation. DETECT-ARDS is trained to detect ARDS finding on chest x-rays, achieving expert physician-level performance and supporting real-time identification of patients to both support care and ongoing ARDS research.
Please contact the Licensing Manager, Drew Bennett, for more information.
Funding History
$545,326 in non-dilutive funding
2020 $545,326 DOD
Substantial additional departmental, school and center based support
Completed Milestones:
Detection of ARDS from raw Chest X-Rays
Pre-training on publicly available data
Training on UM ARDS data
Validation on UM ARDS data
Validation on external data
Detection of ARDS from Chest X-Rays with lung segmentation
Segmentation of lungs in chest x-rays
Pre-training ARDS on publicly available data using lung segmentations
Training dual-lung model using UM ARDS data
Funding Organizations
This work is supported by the Office of the Assistant Secretary of Defense for Health Affairs through the JPC-6 Combat Casualty Care / Defense Medical Research and Development Program under an assistance agreement from the U.S. Army Medical Research Activity, Award No. W81XWH2010496.