Computer Vision Technologies for Rapid Detection of the Acute Respiratory Distress Syndrome
Abstract
This report summarizes the progress made over the second year on the grant: W81XWH2010496/Computer Vision Technologies for Rapid Detection of the Acute Respiratory Distress Syndrome. The Acute Respiratory Distress Syndrome (ARDS) is a critical illness syndrome with a 35% mortality rate. We proposed to develop computer vision technologies powered by deep convolutional neural networks to automatically identify chest x-ray findings consistent with ARDS with expert-level accuracy. During the second year of the grant, we performed several computational analyses to improve our published ARDS model including 1) increasing the model pre-training time, 2) investigating the effect of training a model on larger chest x-ray images, 3) incorporating a lung-segmentation algorithm into our chest x-ray processing pipeline. Of the three strategies, increasing training time resulted in the largest improvement in model validation performance over our base model.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2022
- Accession Number
- AD1195571
Entities
People
- Michael W. Sjoding
Organizations
- University of Michigan