Objective assessment of sonographic quality for breast cancer imaging

"The purpose of any medical imaging device is to acquire timely diagnostic information in a manner that poses the lowest cost and risk to patients and society. Ideally, medical imaging systems are designed by maximizing the benefit-to-cost ratio from a set of laboratory measurements that define ""image quality"" in the sense that they predict diagnostic performance in clinical applications. For ionizing radiation modalities, where risks can be high, the science of image quality has been well developed since the early days in the 1970s, and these principles are now integrated throughout industry and the practice of medicine. In medical sonography, however, these methods are not as advanced for at least two reasons. One reason is that ultrasound is a very low-risk and low unit-cost modality. Because there is no risk caused by ionizing radiation in ultrasound imaging , the motivation for building a regulatory system to integrate academic developments into industry has been low. Yet, the use of ultrasonic imaging in medical practice worldwide is second only to X-ray imaging in terms of unit sales and exams per year. Task-based optimization of any highly used technology like sonography can have a major positive effect on society through healthcare cost reduction. The second reason, found in my dissertation, is that the mechanisms of the sound-tissue interactions generating object contrast in sonography are fundamentally different than those of photon-based imaging. These differences pose profound challenges on how laboratory measurements of image quality should be applied in system design and evaluation, and have prevented the image science of medical sonography from being advanced as quickly as other modalities with respect to ideal observer analysis. The unique contributions of my dissertation research are to develop ideal observer analysis for B-mode sonography as a design and evaluation tool. Specifically, our team developed methods for expressing common diagnostic features of tumors as statistical equations so that we could compute the test statistic of the ideal discriminator from log-likelihood ratios that are unique to each clinical exam. We then obtained mathematical approximations to the exact test statistic expressions that could be implemented in signal processing algorithms and applied to the echo signals of images. This approach was shown to improve the information content of the data as well as human observer performance. My contributions to the team effort were to develop and test (a) iterative adaptive filters that are applied to beamformed echo signals to add task-specific information into the image formation process; (b) several methods for measuring the spatiotemporal impulse response of commercial systems; (c) a definition for visual task information in sonography; (d) a closed-form expression that directly links task information to image quality features, which forms a basis for image quality assessments and design specifications; and (e) concepts that unite alternative approaches to array beamforming under a single analytical framework. This dissertation and associated peer-reviewed publications have helped to define the image science of medical sonography. Our applications have thus far focused on benign-malignant discrimination of breast lesions, but we believe the methods described within have much broader potential."

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2012

Author dc:creator

• Nguyen, Nghia

Contributors dc:contributor

• Insana, Michael F.
• Liang, Zhi-Pei
• Carney, Paul S.
• Hasegawa-Johnson, Mark A.
• Abbey, Craig K.

• Image quality
• Objective Assessment
• Ideal observer
• Breast Sonography

Statement dc:rights

• Copyright 2012 Nghia Nguyen

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/31110

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/31110