1059

IEEE TRANSACTlOr-;S 0:-1 BIOMEDICAL ENGINEERING. VOL. 3i. NO. II. NOVEMBER 199[)

Ultrasonic Reflection Mode Computed Tomography Through a Skullbone JUHA YLITALO,

MEMBER,

IEEE,

JOHN KOIVUKANGAS,

Abstract-An ultrasonic reflection-mode CT method was applied to transskull imaging

of brain. The method involves only

a single trans­

ducer and a single scan to acquire data from the ohject. In reconstruc­ tion an ordinary

Fourier slice theorem is applied. The average velocit�

changes of ultrasound due to the skull bone can be compensated. In

experiments the object immersed in water was scanned by a wide-angle transducer through the viewing angle of

360".

When ima�ing through

bone a simplified approach was employed in which a piece of skull hone (thickness

3-4

mm) was attached

firmly to the transducer. For com­

parison, the same object was then imaged without the skullbone. two-point resolution better than

3

A

mm was achieved for transskull im­

aging using I MHz ultrasound. The experiments with brain specimens show that transskull images compare well with the images of the same specimens obtained without the bone interference. The findings

are

clinically significant in terms of pediatric train diagnosis and postop­ erative follow up. Based on the method, a clinical prototype imager is currently being developed especially for diagnosis of children's hrain diseases.

INTRODUCTION

REVIOUS work on the application of ultrasound hol­ ographic B scan (UHB) to clinical imaging [I], [2] and the development of coherent ultrasound reflected­ mode CT [3] has prompted the application of these prin­ ciples to transskull imaging. It must be stressed here, however, that unlike the methods above the current im­ aging technique employs rectified, incoherent ultrasound. Ultrasonic imaging through a skullbone has been em­ ployed since the 1940's [4]-[6]. The one-dimensional A-scan method was applied to thc determination of the brain midline. It was found that the relatively strong echo from midline structures could be resolved in the presence of noise and strong echoes from the skullbone. The de­ velopment of the two-dimensional ultrasound imaging method for transskull viewing. however, has proved to be difficult, especially in adults [7]-[9]. The main difficulties arise from reflection, refraction. attenuation, absorption, dispersion, delays. and reverberations of ultrasound in a skullbone. These effects, however. are considerably re­ duced in infant skullbone [9]. Moreover, the invention of X-ray computerized tomography improved dramatically the imaging of the brain. Even though the research on

P

Manuf>cript

received September 14.

1989; revised February 12. 1990.

This work was supported by the TechnOlogy Development Centre

land.

of Fin­

J. Ylitalo and J. Oksman are with the Department of Electrical Engi­

neering. Cnlverslty of Oulu, 90570 Oulu. Finland. J. Koivukangas is with

the

Department of Neurosurgery, OulU Univer­

sity Central Hospital. 90220 Oulu. Finland. IEEE Log Number 9038589.

AND

JUHANI OKSMAN

diagnostic ultrasonic imaging methods has since been di­ rected mainly to soft tissue applications, several trials to modify present-day B-scanners for two-dimensional transskull imaging have been made [7], [8]. In order to acquire images through bone, the operating frequency of the transducer has to be reduced from the typical soft tis­ sue range (2.5-10.0 MHz) to a range of 0.5-1 MHz. However, frequencies up to 5 MHz have been success­ fully applied to infant brain imaging [9]. The acoustical properties of the human skull were care­ fully studied by Fry and Barger [10]. They showed that the dominant feature in determining the acoustic behavior of human adult skull is the middle layer (diploe). Diploe causes strong attenuation and scattering of ultrasound. However, the diploe of small children introduces no sig­ nificant attenuation or scattering in the acoustic path. They also concluded that with the selection of appropriate fre­ quencies ( 0.5-1.0 MHz) and beam configuration it will be possible to perform clinically significant transskull di­ agnostic imaging and interrogation in the adult human brain. Indeed, Fry et at. later showed that with usual B-scan apparatus, compound B-scans through the adult skull can be obtained with significant diagnostic resolu­ tion [8]. Transmission mode ultrasound CT has also been ap­ plied to brain imaging. Robinson and Greenleaf showed three-dimensional stacks of ultrasound computerized (ve­ locity) tomograms of excised human brain [11]. Com­ pound B-scan images of a similar experiment were later shown by Koivukangas et al. [12]. To our knowledge the only ultrasound transskull tomograms of the human brain are the attenuation CT images of infant and adult human cadaver heads presented by Dines et al. [9]. They used an ultrasonic frequency of 0.75 MHz for the adult head and as high as 5 MHz for the infant head. The ultrasonic tomograms were taken from the "standard" CT plane, i. e. , the plane determined approximately by the top of the ears and the middle of the forehead. As discussed by Dines et al., the amount of skullbone and its curvature in adults introduce severe distortions when the image plane is moved away from the "standard" CT plane. In the infant experiment, however, the CT projection data were well behaved. The ultrasound tomograms agree quite well with the X-ray CT images obtained from the same planes and, in addition, the reconstructed attenuation coefficients seem to be consistent with the values reported in the literature. Even if the scan took 24 h, they concluded that ultrasound tomography offers promise for the diagnosis of intracran-

0018-9294/9011100-1059$01.00 © 1990 IEEE

1060

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING. VOL

ial abnormalities in infants [9]. This would be clinically significant especially at the age following closure of the fontanelles, the acoustic windows used in routine ultra­ sound of newborns. The incident angle between the ultrasonic beam and the skullbone seem to play an important role. Dines et al. showed that in adults ultrasound transmission CT can be applied only near the standard CT plane where the inci­ dent angle of ultrasound is nearly perpendicular. In in­ fants the problem is not so severe [9]. Even at the stan­ dard CT plane severe distortions arise in transmission CT when profiles near the outline of the head are collected. Considering the transskull imaging methods presented above [8], [9], the present work introduces a method were only a single transducer and a single transducer position gives an echo profile. The main advantage is that the in­ sonifying beam is approximately perpendicular to the bone surface throughout the scan. Therefore the difficulties due to the ultrasound-bone interaction are considerably re­ duced which, at least in infants, should contribute to a better quality of images. In addition, the time needed for data acquisition can be made as fast as 30 s or even less (the scanning time of the present scanner is about 5 min). METHODS

In these experiments an ultrasonic computed tomogra­ phy (CT) method was applied. In place of ordinary ultra­ sonic CT based on attenuation or speed measurements re­ flection mode CT was used (Fig. 1). In the literature this type of ultrasonic CT has been applied using a plane-wave beam [\3] and fan-shaped beam [14] mainly in NDE ap­ plications, but also in a coherent mode in medical appli­ cations (15]. A similar method was also presented by Dines and Goss who also showed an in vitro experiment with a beef liver specimen [16]. The tomographic approach gives several benefits com­ pared to ordinary B-scan methods: increased spatial res­ olution, reduced speckle, reduction of imaging errors, and wide dynamic range (16]. Furthermore, the ultrasonic re­ flection CT introduces some additional advances over transmission CT. First, only a single transducer and a sin­ gle scan around the object is needed, which reduces con­ siderably the time required for data acquisition. Secondly, reflection mode offers better spatial resolution. On the other hand, one drawback is that only qualitative images can be obtained. Since the present work involves well-known ultrasound reflection CT methods producing mainly experimental re­ sults, the theoretical background is presentcd briefly. The reader is asked to review (14]-(16]. The laboratory imaging apparatus is based on a PDP 11173 minicomputer. The minicomputer is connected to a self-made digitizer unit which involves also a timing card and a TTL-level pulser. A separate pulse amplifier trans­ mits the pulses to a transducer. The transducer can be ro­ tated by computer control in a water tank made of epoxy glass. Before digitization the reflected echoes can be am­ plified in two separate preamplifiers. However. in these

,7. NO.

II. NOVEMBER 1990

experiments a no time-gain-compensation (TGC) ampli­ fier was available. The strategy of experiments was as follows: the speci­ men was imaged first without the intervening skullbone, and then a piece of human adult skullbone was placed in front of the transducer and the imaging process was re­ peated. This approach assures equal imaging conditions of more reliable comparison. A piece of adult parietal skullbone with an average thickness of 3.5 mm was se­ lected in order to simulate the bone effect of children. Specimens of calf brain and various human tumors (e.g., glioma) were imaged after the method was investigated with test objects including nylon and steel wires.

Data Acquisition The principle and the experimental setup of the under­ lying ultrasound reflection CT are depicted in Fig. 1_ The procedure for data acquisition is as follows: the reflected echoes are rectified and recorded to form the first line (A-mode line) in the CT projection data (sinogram data). This A-mode line is regarded as the first CT projection for a rotational angle 80 0 rad. The transducer is then ro­ tated to an angle 81 I X 27r / N rad where N denotes the number of projections around the object. The corre­ sponding A-mode line is stored in the sinogram data and the transducer is rotated to an angle 82 2 X 27r / N. The process is continued until all the projections around the object have been collected. In the sinogram data the strong echo from the skull bone draws a straight echo front at time 1 I, if the bone is kept at a constant distance from the transducer. In Fig. 1 to denotes the transmission of the ultrasonic pulse and 12 the delay of the echo from the cen­ ter of rotation. The echoes from the object form sinusoids around the delay 12 in the sinogram data. Fig. 2 shows an experimental sinogram of an object which consisted of three steel wires ( diameter 2 mm) and the corresponding image. The strong bone echoes have been reduced considerably by a diode limiter. Since no time-gain-compensation (TGC) amplifier was available each echo profile was amplified numerically after detec­ tion by 1 dB / cm. Various other amplification factors ranging from 0 dB / cm to 3 dB / cm were also applied to sinogram data of a brain tissue specimen but with no sig­ nificant difference in thc output image. Therefore the am­ plification of I dB / cm was chosen, being also a typical figure for soft tissue at 1 MHz. =

=

=

Transducer In data acquisition the most critical factor is the effec­ tive beam profile of the transducer. Ideally, the transducer sends a very narrow plane (or fan) beam into the object. The beam profiles of our self-made transducer were mea­ sured at distances typical to the experimental setup (Fig. 1). In the imaging plane the -6 dB beamwidth was ap­ proximately 13° and the -12 dB beamwidth was 17°. These beamwidths correspond to transaxial imaging ap­ ertures of about 25 mm ( -6 dB) and 31 mm ( - 1 2 dB) if the center of rotation is at a distance of 80 mm from the

l061

YLITALO el al.: COMPUTED TOMOGRAPHY THROUGH SKULLBONE

The Olr.ction of Rotation

Center of Rotation

Object

t

,

Transducer

Piece of skullbone

jl



11

12

Ultrasonic reflection mode computed tomography through a skull bone.

An ultrasonic reflection-mode CT method was applied to transskull imaging of brain. The method involves only a single transducer and a single scan to ...
695KB Sizes 0 Downloads 0 Views