A clinical evaluation of Contrast-Detail analysis of ultrasound images

A clinical evaluation of Contrast-Detail analysis of ultrasound images

ABSTRACTS, ULTRASONIC IMAGING AND TISSUE CHARACTERIZATION SYMPOSIUM the superimposing on the original points the original image component tangential ...

159KB Sizes 2 Downloads 98 Views

ABSTRACTS, ULTRASONIC IMAGING AND TISSUE CHARACTERIZATION SYMPOSIUM

the superimposing on the original points the original image component tangential level value; 3) by the average radial function of time. sector by sector; or

local

velocity vector at even1 y-spaced image; 2) by replacing each point of by its velocity magnitude, its radial or velocity component represented by a grey plotting the average velocity magnitude or and tangential velocity components as a This can be done for the whole myocardium and 4) by building functional i mages level of each point the time of where the grey indicates occurence or magnitude of its maximum velocity. This work was supported in part by the National Sciences and Engineering Research Council of Canada and the Canadian Heart Foundat i on through a Research Scholarship to Guy E. Mailloux. 2:

SESSION

STATISTICAL

FUNDAMENTAL Wagner+, R. F. ‘Office of

M. F.

Science

Radiological

Health, Duke

of Radiology, 27706. There backscattered is contained

is

in displayed

typically statistical not readi The lower the size carry

moments

shape

information

video

or

scatterers.

depends

the

1 y accessed frequencies of diffuse

envelope contributions signatures samp 1 es density

a signals

avai of on

the

incoherently determination, samp 1 es available

II

PHYSICS OF MEDICCIL ULTRASONIC IMCIGES, Smithzv2, Insanal, D. G. Brown* and S.W. and Technol ogy , Center for Devices & FD&, Rockville, MD 20857 and =Department University Medical Center, Durham, NC wealth of physical i nf ormat i on in in medical ultrasound. This information radi of requency spectrum--which is not the viewer--as to well as in the higher of the envelope or video signal--which are human viewer of typical B-scans. by the of the rf spectrum carry information on scatterers and the higher frequencies on these small-scale scatterers. The carries information on the relative

of coherent The density is determined 1 able and

speckle

IMAGING

or

system compounded. there are

and

diffuse

information

of by

the

in

number

this is determined correlation cells. resol ut i on and the For a typical

of

incoherent) these tissue independent data in turn by the The latter density number of i Rages liver structure

all

of

about a thousand correlation cells or 5 cm= region. In conventional radiography, there are about forty thousand samples avai 1 able over 1 mm’. Thcrs, the ultrasound inverse scattering problem is relatively ill-posed and tissue characterization techniques must be based on statistical methods working at low

over

a

resolution.

A

CLINIC&L

EVCILUATION OF CONTRAST-DETAIL CIN&LYSIS OF H. Lopezl**, M.H. Loew” , P.F. Butler" and =Center for Devices and Radiological Health, FDA, Rockville, MD 20895 “Gear ge Washington and University, Washington, DC 20520. A major advantage of Contrast-Detail (C/D) analysis of ultrasound imaging systems is the quantitative information obtained, in the form of a Contrast-Detail plot, of the imaging system's ability to display a range of varying-size targets as a function of contrast. The C/D plot, however, is obtained by using human observer readings. Our study shows that data obtained from human observers is difficult to match ULTRASOUND t-l. H. Hill=,

IMAGES,

49

ABSTRACTS,

ULTRASONIC

IMAGING AND TISSUE

CHARACTERIZATION

SYMPOSIUM

with that of other observers in test that the and wow neither a larger test group nor repeated readings reduce the overall reading error as predicted by theory. Other investigators postulated that overall error should Cl,21 have decrease with the number of observers, number of observations, and with the independent images, based on number of simplifying assumptions about the statistics of the observations, and the reliability of the observers. Our results indicate that the determination of absolute threshold diameter values for targets of differing contrast in C/D analysis by human observers yields a high degree of error, which is not predicted by this theoretical relationship. We find that systematic the process of error is introduced by learning during the experiment and by variables other not included in the error prediction. These results suggest that C/D analysis may be impractical in the clinical environment, especially for observers without extensive training in a specific task. We suggest that a computer-based observer may be more reliable image quality for objective measurement of for evaluating ultrasound images, and propose a model of such an observer. Cl1 Swets, J.A. and Pickett, R-M. (Academic Press, NY, 1982). C21 Loo, L-D., Doi, K,, et al, Proc. SPIE XI,, (April, 1983). REAL-TIME ENHANCEMENT OF MEDICAL ULTRASOUND IMAGES, Tim Electrical Engineering, University Blankenship, Department of of Missouri-Rolla, Rolla, MO 65401. two techniques are discussed that produce In this paper, real-time or real-time improvements in the image near sharpness and contrast of a microprocessor-based medical ultrasound instrument. The first technique is histogram hyperbolization. This simply improves contrast by producing with a grey images level distribution that appears uniform (equal ired) to the human eye. The second technique, statistical differencing , primarily enhances (sharpens) edges. This locally-varying operation divides the image into subregions and applies a gain and bias to the center pixel of each subregion. The gain and bias are functions of the subregion's standard deviation and mean. These processes are evaluated by varying their respective parameters and comparing original image. A number of the resulting images with the comparisons are Image quality resulting from these made. histogram hyberbolization is examined when performed combination with statistical differencing. separately, and in possible hardware Also considered are some of the implementations. the Missouri Research This work was supported by Assistance Act and Linscan Systems, Inc., Rolla, MO. BY NARROWBAND FILTERING, F.G. IMAGES CREATED ULTRASONIC Departments of Radiology and Sommer, R.A. Stern and H.S. Chen, Stanford, CA Engineering, Stanford University, Electrical 94305. commercial scanner, unfiltered Using a specially-adapted were from region5 of interest ultrasonic waveform data from ultrasonic digitized to a standardized dynamic range and cirrhotic human livers. phantoms and from normal filtering the then created by first Ultrasonic images were filter, of 3.4 MHZ center digitized data with a narrowband and El00 kl-lz bandwidth, prior to thresholding on the frequency

50