A Machine Learning Approach to Image Quality
Principal Investigator
Name
Andrew Brown
Degrees
MD, MBA
Institution
MIT
Position Title
Radiology Resident
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-139
Initial CDAS Request Approval
May 14, 2015
Title
A Machine Learning Approach to Image Quality
Summary
In radiology, image quality analysis and quality assurance are inconsistent and subjective. Often the result of poor image quality is repeat imaging which degrades the patient experience and increases the cost of care. There is a lack of standardized tools for measuring image quality. We propose to investigate the prevalence of deficiencies in image quality and apply statistical machine learning approaches to develop quantitative metrics of image quality. Images from 3,000 NLST participants will be used. A wide range of statistical models will be applied to the data and the results will be compared to subjective quality tests to verify these objective algorithms.
To date, much of the work at the intersection of machine learning and medical images has focused on diagnosis. This work aims to examine a relatively nascent area in radiology - image quality analysis. This work has the potential to improve image quality and support the “Triple Aim”, specifically in improving the patient experience and reducing healthcare costs.
To date, much of the work at the intersection of machine learning and medical images has focused on diagnosis. This work aims to examine a relatively nascent area in radiology - image quality analysis. This work has the potential to improve image quality and support the “Triple Aim”, specifically in improving the patient experience and reducing healthcare costs.
Aims
1) Determine the prevalence of deficient images within the data set
2) Explore the viability of applying machine learning algorithms to evaluate medical image quality
3) Develop an objective measure of image quality
4) Determine whether objective measures of image quality correlate with the subjective opinions of experts