Bijoy J. Sundersingh

Document Type: Master's Thesis
Name: Bijoy J. Sundersingh
Email address: bijoy@bme.utmem.edu
Title: Qualitative Evaluation of Data Compression in Real-time Ultrasound Imaging
Degree: Master of Science
Program: Biomedical Engineering
Research Advisor: Herbert D. Zeman, Ph.D.
Advisor's email: hzeman@utmem.edu
Committee Members: Donald Emerson, M.D.
Stephen Kritchevsky, Ph.D.
Lawrence Jordan, Ph.D.
 
 
Keywords: Ultrasound, Data Compression, Image Quality Assessment, Kappa Statistics, GLM, ANOVA, PSNR, Histogram Analysis
Availability: Release the entire work for World Access.

Abstract

    The purpose of this project was to evaluate qualitatively real-time ultrasound imaging using objective and subjective techniques to determine the minimum bandwidth required for clinical diagnosis of various anatomical and pathological states. In the experimental setup live ultrasound video samples representing the most common clinical examinations were compressed at 128, 256, 384, 768, 1152 and 1536 kbps using a compressor-decompressor (CODEC) adhering to International Telecommunication Union (ITU-T) recommendation H.261. A protocol for qualitative evaluation was developed and subjective and objective testing were performed based on this protocol. Subjective methods comprised of inter-rater reliability tests using kappa statistics and three way Analysis of Variance (ANOVA) using General Linear Models (GLM). Objective testing were performed using histogram analysis and estimation of peak signal to noise ratios.

    The kappa scores for all bandwidths greater than 256 kbps indicated good inter-rater reliablity and minimum variation in confidence levels. Using the results from GLM and ANOVA we could not establish a trend in degradation of observer confidence with increasing compression ratios. The histogram analysis showed a linear increase in standard deviation values, indicating a linear scatter in pixel intensity, with increasing compression ratios. Although higher compression levels were evaluated, only video clips with bandwidths greater than 256 kbps displayed satisfactory temporal and spatial resolution, good enough to make clinical diagnosis of various anatomical and pathological states. The evaluations also indicate that compressed real-time ultrasound imagery using H.261 can be transmitted over a T1 or ADSL networks.

Attached File(s)

| Back to top of this page | Back to ETD Index Page | Back to CGHS Home Page |

Revised 23 May 2002