C Abbey¹*, M Anastasio²*, M Kupinski³*, (1) University of California - Santa Barbara, Santa Barbara, CA, (2) University of Illinois at Urbana-Champaign, Chicago, IL, (3) University of Arizona, Tucson, AZ

• 
  C Abbey
• 
  M Anastasio
• 
  M Kupinski

TH-AB-TRACK 4-0 (Thursday, 7/29/2021) 10:30 AM - 12:30 PM [Eastern Time (GMT-4)]

Medical imaging systems and image processing operations are widely employed to support clinical practice.
In medical imaging, images are often acquired for specific purposes and the use of objective measures of image quality (IQ) has been widely advocated for assessing imaging systems and image processing algorithms. Task-based measures of IQ quantify the ability of an observer to perform specific tasks.

Advanced deep-learning (DL) methods are being rapidly developed for a large number of medical imaging applications. The optimization of medical imaging systems and algorithms is still mainly guided by traditional physical measures of IQ, and the objective evaluation of IQ of medical imaging and image processing methods remains largely lacking. Traditional IQ measures such as spatial resolution, contrast, dice coefficient, and noise levels are widely employed; however, it is well-known, but perhaps under-appreciated, that their clinical value is limited because they are not always sensitive to the intended use of the image.

This symposium will address recent developments in task-based IQ assessment in diagnostic imaging and radiation therapy, and its applications for the assessment and optimization of advanced DL-based image processing algorithms. In addition, the challenges of using task-based IQ measures to guide the optimization and assessment of imaging system and image processing will be discussed as well.

Learning Objectives:
1. Understand basic principles of task-based image quality assessment and its applications in optimizing medical imaging systems.
2. Understand the difference and relationship between objective/task-based and traditional physical metrics-based image quality assessment.
3. Understand the difference and relationship between model observers and human observers, understand the challenges in using task-based image quality assessment theory to guide clinical applications and assess DL-based imaging operators such as image reconstruction, image denoising, super-resolution, and others.

Keywords

Imaging Theory, Optimization, Observer Performance

Taxonomy

IM/TH- Mathematical/Statistical Foundational Skills: Observer Models

Contact Email