Neuroradiology 16,524-536 (1978)

Nearoradinlngv

© by Springer-Verlag1978

Third Panel Discussion The Ideal Head CT Scanner Chairman: Giovanni Di Chiro, Bethesda, Maryland Participants: R. E. Anderson, Salt Lake City; C. R. Archer, St. Louis; E. S. Cabanis, Pards; N. E. Chase, New York; R. J. Churchill, Maywood, Ill.; A. J. Cooke, Akron; V. H. Haughton, Milwaukee; S. K. Hilal, New York; J. Isherwood, Manchester; K. R. Maravilla, Dallas; P. F. J. New, Boston; SI L. G. Rothman, New Haven; M. M. Sehechter, New York; M. A. Weinstein, Cleveland Opening Remarks: Scope and Rules of the Panel Di Chiro: The idea of this panel originated because of some personal experiences which I am sure have also occurred to other neuroradiologists. These experiences are related to a fundamental change which has occurred in our specialty, i.e., the noncomparability of the devices used for computed tomography (CT). Ten years ago, when neuroradiologic studies, e.g., arteriograms, were sent for consultation to a neuroradiologist, there was no need for him to know what type of X-ray equipment had been used for the examination. Neither was it necessary, in most instances, to indicate in a scientific report the type of X-ray machine employed. Nowadays, a certain number of head CT scans are sent to me for consultation from different centers. Regretfully, in a relatively large number of eases, I am not able to offer my assistance due to the fact that I am not totally familiar with the performance characteristics of the machine used to obtain the submitted scans. The problem is particularly difficult when the changes in question are subtle; indeed changes may be recognizable or not, depending upon the resolution capability of the tomograph. If a head CT scan obtained with one machine appears normal, this does not necessarily mean it would also appear normal if another machine had been used. A few months ago, I was asked to evaluate and participate in a joint research project involving some twenty USA medical centers which treat patients with leukemia using the standard CNS prophylaxis (radiation plus intrathecal methotrexate). The purpose of the study was to evaluate, by CT scans, early leukoencephalopathic cerebral changes possibly associated with this prophylaxis. After careful consideration of the diversity of the tomographs available in the different institutions, it became apparent that embarking on such a study would have little value due to the large number of unavoidable variables. We had to renounce carrying out the project. For all the above it appeared to me that getting together a panel of individuals who have been working

with different CT devices would be useful. Each speaker has been asked to show images, discuss his experience, and emphasize one or two outstanding features of the particular machine he has been working with. Thirteen devices will be discussed. Dr. Isherwood will summarize the salient points and choose the features which he considers of particular value and worthwhile incorporating in an 'Ideal Head CT Scanner.' The hope is that as a remit of this panel, neuroradiologists will have a better knowledge, from statements of their peers, about features and performances of machines different from the ones they work with every day. Another desirable goal is that some of the commercial companies will perhaps modify and/or add new obviously advantageous capabilities to their products. Without further ado, I give the floor to Dr. Hilal who will discuss the AS and E scanner and its resolution capability. Dr. Hilal will also deal with the stability of the CT numbers, a problem which represents a concern of many of us who are trying to improve CT diagnosis through quantitative analysis of CT data. AS and E CT Scanner (American Science and Engineering, Inc.): High Resolution Scanning and Stability of CT Numbers Hilal: The Pfizer/AS and E scanner is the first stationary detector array scanner. It was developed by the joint effort of American Science and Engineering and our team at the Columbia-Presbyterian Medical Center in New York. The two main characteristics of this scanner are its high resolution and the stability of the CT numbers regardless, of object size, slice thickness, kVp and mA of the X-ray source.

Resolution of the Pfizer/AS and E Scanner: The factors affecting the resolution in a CT scanner are: (a) the effective beam width; (b) the sampling rate; (c) the algorithm filter; (d) the number of views; and (e) the pixel size. 0028-3940/78/0016/0524/$02.60

Third Panel Discussion. The Ideal Head CT Scanner The effective beam width depends on the size of the X-ray source, the size of the detector, and the distance the beam travels during one integration period (Fig. 1). Consequently, if one is to build a scanner with a high resolution one would select a machine that will accept the smallest focal spot possible. Two-motion machines (translate-rotate) unfortunately cannot easily incorporate in their design an X-ray tube with a small focal spot. The small focal spot requires a rotating anode X-ray tube, which can only be incorporated in single-motion scanners. Single-motion scanners in the current designs are available i n two basic configurations: (a) a single-motion scanner where the source and the detectors are mechanically linked and rotate together; (b) the stationary detector array design where only the source rotates and is mechanically uncoupled from the detectors. The machines with mechanically coupled and rotating source and detectors are limited in their linear samplings of the rays constituting a single view. In these scanners the number of linear samples in a given view is the number of detectors. One can, therefore, obtain only one sample per beam width. A beam width in this design is defined by the size of a single detector and the size of the X-ray focal spot. When the machine rotates to acquire the next set of samples, the new set is another view and is considered a different angular sample. A fundamental rule of physics dictates that in order to obtain the maximum information from a given set o f data the system should have at least two linear samples per beam width. This is not achievable in single-motion scanners with coupled source and detectors. The linear samplings in this system cannot be improved by offsetting the center of rotation of the scanner by a quarter of a beam width. With the stationary detector array systems, the linear sampling rate can be increased at will to provide any number of samples for each beam width. The system, therefore, permits maximum extraction of information from the set of data obtained for a given beam width. All scanners obtain the X-ray transmission measurements while the scanner is moving. Each measurement, therefore, represents the density of the area covered by the X-ray beam during the period of integration. For a given beam width and a given amount of beam displacement, there is less effect of blurring due to scanner motion with a stationary detector array machine than with any other scanner design. In summary, the stationary detector array scanner provides the highest possible resolution for a given beam width and a given number of detectors. It also permits small X-ray focal spots which allow tree high resolution in a three-dimensional geometry. X-ray slices of 2 mm are achievable with little penumbra. Some of the two-motion scanners could potentially resolve phantoms of fine line spacing equivalent to the resolution obtained in the stationary detector array scanner in current usage; the problem with the twomotion scanner, however, is the fact that the source is typically longer than 15 mm in the z-axis orientation, thus precluding slice thickness finer than 8 or 5 mm. Ob-

525

I

L t)-

Fig. 1. Effective beam width. In a theoretical completely stationary system (left diagram) the beam width is defined by the size of the X-ray source (on top), the collimator, and the detector (D). The shaded area represents the region of the object traversed by the X-ray beam and represented by a single X-ray measurement. In practice, however, all CT scanners sample the X-ray beam while the scanner is in motion. Between any two consecutive measurements of the transmitted X-rays, the X-ray source and the detectors move slightly (right diagram). The X-ray measurement obtained in this system represents a larger area of the object than that obtained in the theoretical completely stationary system. The effective beam width is the combination of the beam size and the distance the beam moves between two samples

viously, for a spherical object or most anatomic structures a tree three-dimensional f'me resolution is required and the demonstration of high resolution phantoms with a predominant orientation in the z-axis is not quite relevant to the day-to-day handling of human anatomy, unless one is looking at cylindrical objects. The patient radiation dose in the Pfizer/AS and E scanner is 2.2 rads on the skin on the anterior aspect of the patient, and is approximately 1.4 rads in the center of the patient. As an example of the high resolution achievable with the Pfizer/AS and E stationary detector array scanner, the demonstration of the anatomy in the orbital apex and parasellar region is most suitable. Figures 2a and 2b show the outline of a tumor of the orbit apex with bone destruction that could not be demonstrated by a largefocal-spot two-motion scanner (Fig. 2c and 2d). Figure 3 shows a coronal section through the sella and optic chiasm region. Of note is the outline of the chiasm and of the structures within the two cavernous sinuses.

Stability and Accuracy of CT Numbers: With the great tendency for quantitation of anatomic and physiologic information on the CT image, the need for stable CT

526

Third Panel Discussion. The Ideal Head CT Scanner

Fig. 2a. Pfizer/AS and E scan of the orbit showing a sharply marginated tumor in the apex of the right orbit (solid black arrow). In front of the tumor the superior ophthalmic vein is seen crossing the orbit forward and medially. The left superior ophthalmic vein is also well shown (white arrow)

Fig. 2b. Pfizer/AS and E scan on the same patient at a lower level than that Of Figure 2a. The tumor in the right orbital apex is again sharply demonstrated (black arrow). There is also evidence of bone erosion (hollow black arrow). In the left orbit the ophthalmic artery is clearly demonstrated (white arrow). Note that the artery is quite different in shape and course from the vein shown in Figure 2a

Figs. 2c, 2d. CT scan of the same patient as Figures 2a and 2b obtained with an advanced two-motion body scanner with a slice thickness of approximately 5 ram. The two levels of the section are the closest to those illustrated in Figures 2a and 2b. There is clearly loss of resolution. The sharply outlined tumor can hardly be appreciated, the vascular structures so clearly seen on the Pfizer/AS and E scans of Figures 2a and 2b cannot be adequately visualized, and, finally, the bone of the calvarium is quite thick. The resolution achievable by two-motion machines is limited compared to what can be achieved by a stationary detector array scanner, primarily because of the large X-ray source needed in the two-motion machines. Other factors are explained in the text

Fig. 2a-d. Comparison of two CT scans of the same orbit obtained with a stationary detector scanner and two-motion scanner. Figures 2a and 2b are obtained with the stationary detector array scanner (the prototype Pfizer/AS and E machine) and Figures 2c and 2d are obtained with an advanced two-motion scanner

numbers is rapidly increasing. The physiology of the cerebrospinaI fluid tagged with metrizamide is of great importance. Also, the use of the xenon inhalation techniques for regional cerebral perfusion requires the acquisition of accurate quantitative data from the CT image. In existing machines the CT numbers tend to drift as a function of object size, object composition, kVp of the X-ray beam as well as the mA. The instability of the CT numbers is due to a variety of nonlinear processes in the system. For one, the X-ray beam utilized is polychromatic and its absorption characteristics are highly nonlinear depending on object size and composition.

Another factor is the nonlinearity of the detectors over the large dynamic range needed to image the h u m a n body. The instability of the CT numbers is particularly disturbing in air machines which, unlike the initial Hounsfield scanner (EMI Mark 1), cannot be calibrated against the standard water box at the end of each traverse. The air machines have to contend with a bigger dynamic range of response and a more important polychromaticity effect. To achieve the stabi/ity of the CT numbers in the Pfizer/AS and E machine, two corrections were introduced:

Third Panel Discussion. The Ideal Head CT Scanner

527 Table 1. CT density of five pins of the water phantom illustrated .in Figure 4. Results before and after the use of the two correction parameters described in the text CT numbers for five pins Pin 1 2 CT number of each pin

-90

-20

3

4

5

95

100

120

Variation of phantom size 6-in. vs. 8-in. phantom: 50 mA, 10-mm slice thickness Pin

Fig. 3. Coronal section of the cranium at the level of the sella turcica. The section was obtained after contrast enhancement. Please note the visualization of the structures within the cavernous sinuses which are filled with the radiopaque medium. The multiple lucencies within the sinus probably represent the nerves crossing the sinus cavity. The carotid artery in the sinus is f'tUed with contrast and cannot be visualized. One can also see the optic chiasm transversely above the sella turcica forming the floor of the third ventricle

Uncorrected difference Corrected difference

1

2

3

4

5

Mean

18.5

19.4

25.7

20.8

23.8

21.6

-2.7

-1.8

4.5

-0.4

2.6

-0.4

Variation in mA 50 mA vs. 20 mA: 8-in. phantom, 10-mm slice thickness Pin Uncorrected difference Corrected difference

Mean

1

2

3

4

5

14.5

15.1

19.2

17.7

16.3

16.6

-2.7

-2.1

2.0

0.5

-0.9

-0.6

Variation in slice thickness 10-mm vs. 6-ram slice thickness: 8-in. phantom, 50 mA Pin Uncorrected difference Corrected difference

Fig. 4. Calibration phantom with five pins of different densities covering a range from 120 CT numbers to -90 CT numbers. The background of the phantom is water a. A correction for beam hardening introduced on the raw data before image reconstruction. This correction assured a fiat field regardless o f object size. b. The use of a water p h a n t o m in the field o f reconstruction and the recalibration of the CT image after reconstruction, so that water is always zero and air is always - 1 0 0 0 units. To test the results we imaged t i e phantom shown in Figure 4 which contains five pins varying from +120

Mean

1

2

3

4

5

18.7

18.3

16.9

21.7

23.8

19.9

-2.6

-3.0

-4.4

0.4

2.5

-1.4

Hounsfield numbers to - 9 0 Hounsfield numbers (Table 1). This phantom was imaged within water bottles of different sizes measuring 6, 8, 11, and 14 inches in diameter. The mA and the slice thickness were also varied. Before any correction is used, the mean error in the CT numbers of all the pins could read as much as 21.6 Hounsfield numbers. After the correction was introduced the largest mean error was - I .4 CT numbers. In conclusion, the Pfizer/AS and E scanner offers images of probably the highest resolution today because it provides 2-mm slices with a high resolution in the xand y-axes as well. The increased resolution is inherent in the design of the stationary detector array machines. It also provides unusually stable CT numbers because of the use of several polychromaticity corrections as well as a correction for a number of nonlinear effects. A mean error of - 1 . 4 CT numbers is achieved over a wide range. One additional advantage for the neuroradiologist is the ability to tilt the bed of the Pf'tzer/AS and E scanner for imaging of the spine. One can move a radiopaque medium either away from, or to the area of interest. CT myelography is greatly facilitated.

528

Artronix Scanners (Artronix, Inc.): Coronal-Sagittal Reconstruction and Source-Detector Geometry Maravflla: We have now performed approximately one thousand examinations in select patients using the sagittal-coronal computer reconstruction techniques to evaluate CT scans in multiple projections. We have found this technique of extreme value in providing us with additional information which we often find critical in diagnosis and management of these patients. I could show dozens of cases illustrating the usefulness of this technique. Briefly, our method varies slightly from the method used in most other scanners in that we obtain thin-section axial plane scans of 3-mm slice thickness. These 3-ram scans are taken contiguously with no overlap between scans. By using this technique we can decrease partial volume effects and increase resolution; we can decrease computer reconstruction time by eliminating the need to reconstruct thin sections from the overlapping scans; and we can decrease radiation dose. The radiation dose has been measured in our unit for a complete study of the entire head with 3-ram sections and maximum skin exposure is 3.4 rads. The resolution is quite good, as you can see from this illustration (Fig. 5). This is a mid-

Fig. 5

Third Panel DiscusSion.The Ideal Head CT Scanner line sagittal projection which has been reconstructed from the 23 thin slices which are illustrated directly above it. Notice that you get good definition of the deep venous drainage system and the fourth ventricle. Even the anterior cerebral artery together with its bifurcation into the pericallosal and callosomarginal arteries can be seen. We have found the sagittal¢oronal computer reconstruction techniques to be most useful in the sella and parasellar areas, in the region around the base of the skull, in the posterior fossa, especially when evaluating the brain stem, and in the foramen magnum region. I would like now to discuss one feature of the Artronix CT Scanner Systems which Dr. Di Chiro has asked me to comment on, i.e., the new geometrical configuration of the Artronix Total Body Scanner System. This has generated some interest and it is a somewhat confusing concept. I will briefly try to explain this configuration and indicate its advantages. We see in these slides a graphic representation of the system which consists of an X-ray tube which rotates around the patient (Fig. 6). The patient aperture is located in the inner circle shown in the sketch. A fanshaped X-ray beam of approximately 120 ° encompasses the entire patient. The detector system consists of a 360 ° ring of xenon detectors. The detectors do not

Third Panel Discussion. The Ideal Head CT Scanner

1

1

540~180 A

540~'180

360 1

B

360 1

4000

180

540 C

529

360

360

Fig. 6

sult of such a configuration is a stationary detecto r system with X-rays that are emitted radially from the center of the detector circle in a fashion which is shown in this slide (Fig. 7). Since the X-rays are emitted radially, all of the individual ray projections emanating from the focal spot are normal to the circumference of the detector circle. Because of this configuration, deep well detectors, such as xenon detectors, can be used and individual detector' collimation is possible. This type of collimation would not be possible with ray projections coming into a detector window at various angles. The advantages of this system include those of the stationary ring detector systems. In addition, since deep-set xenon detectors can be used, the cost of building a scanner with a 360 ° detector ring can be reduced. Individual detector collimation to reduce scatter radiation and improve spatial and density resolution is possible with this type of system, and a slip ring is employed so that the X-ray tube can continuously rotate about the patient, permitting the use of rapid sequential scanning, or gated cardiac scanning. The potential for measuring other physiologic events or synchronizing various physiologic parameters, thus, exists.

ND 8000 Scanner (CGR Medical Corporation): High Resolution OrbitolOcular Studies

540 ~

180 360

Fig. 7

rotate, but they do move or 'slicle' in such a fashion as to maintain the edge of the detector ring tangent to the circle formed by the rotating X-ray tube. The point of tangency remains directly opposite the X-ray tube at all times during the scanning motion. You will notice that as the X-ray tube rotates in the direction indicated by the circular arrows (drawn on inner ckcle), the detector system is sliding in a direction roughly indicated by the large upper arrow (drawn outside the outer circle). Reference detectors are numbered in order to show that the system is not rotating. On these graphic slides you will notice that the circular patient aperture remains in the center and does not move, but that the detector ring is moving eccentrically around the patient aperture. The geometrical equivalent of this system is one in which the X-ray tube always remains at the isocenter of the detector ring. The net re-

Cabanis: In the last few months we have been using the ND 8000 scanner strictly in the field of radiographic analysis of the orbits and optic pathways. We are paying particular attention to the recognition of the fine anatomic detail of the retro-ocular structures. For this purpose symmetrical scans of the two orbits are indispensable. The resolution capabilities of the ND 8000 allow us to study with high accuracy not only the optic nerves but also the retro-ocular vessels. We are able to visualize the superior ophthalmic veins consistently and quite frequently we succeed in demonstrating the various segments of the ophthalmic artery. We may choose between a 180° 80-s scan and a 360 ° 80-s scan. The zoom mode allows us to significantly increase the resolution. We frequently carry out the dynamic tests intended to demonstrate the changing position of the optic nerves and the extraocular muscles with gaze shifts according to the method described by Di Chiro. We make extensive use, as a landmark for our analysis, of the so-called neuro-ocular plane which includes three elements: the lens, the papilla, and the optic nerve up to the orbital opening of the optic canal. Comparative histograms of the content of the two orbits are easily obtained with the ND 8000. The densitometric observations of the optic nerve are important. We should note that the attenuation values in the central part of the optic nerve are higher than at the periphery; this may create problems in the evaluation of the precise diameter of the optic nerves. In this regard the partial volume effect is also important. Obtaining coronal images with the ND 8000 is quite easy.

530

Third Panel Discussion. The Ideal Head CT Scanner

700 Scanner (Elscint, Inc.): Scan Zoom and Beam Hardening Correction Seheehter: The unique features of the Elscint Neuroscanner will be dealt with in this talk. The Elscint CT is a translate-rotate scanner with a 5.8-s scan, 13-s reconstruction, and 52 BGO detectors. The machine offers high quality images for both head and body. This is the only neuroscanner with true neuroscanning capability, that is, the entire spine can be scanned. Scan zoom, reconstructed zoom, and pre-zoom are unique features of the Elsci~.t Neuroscanner. The high resolution restricted scanning field :settings are helpful in obtaining the greatest degree of ~flefinition at the least expense to the patient in radiation ~cIose. The Swivel Laser feature enables the operator ea~t/y to sdign the ~9~Iterrt~shead by swivelling ,~e couch ~fiitm~t manipulating the patient. The laser beam d e ~ . ~ ~ scan phase so ~ a t 'when the patient's external ~ea~thi are aligned with ~he ,beam, almost perfect ~ ~is guaranteed. A unique 5:earn hardening correVli~m ~. ~stem makes it possible to,ob:tain~undershoot-overs~'oot~fa~ee ~ n - b o n e interface. This .neuroscanner p~.ovides full diagnostic quality iimages .reconstructed .~n line during the scan. Direct ~ o s i t i o r ~ ) coronal ,scan of the head as well as computer a e ~ c o ~ c t e d coronal or sagittal scans are available.

High ResoLution Sector Scans Port(a[ {&20 ram)

/ / Catib:

....

CaLib:

lUlllllllllllllllllllltlllllllllIIIIIIIIIIIIIIItllllllllllllllllflIIIIUllllltllllIIIlllllllllltllIlitlllUlillllllUlll ~ 0.6 m m

I

~0.6 m m

i

~0.6 mm

NR

q2diib: ~---Data "i Catib: I llllllll[llll IIHIIIIII1111111111111111 i[IllVUUlIInlIIII~IIItlIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIHIIIIIIIIIIII I III II Ifl II Ill Ill tllllllllllllllllllllll HR ~ 0.6 mm I l ~~6Se[ecta, clcd(at~mm J ~03 mm SeLected data I

k

Fig. ,8

CT 5005 (EMI Medical Systems): High Resolution Scanning

Table 2. High resolution sector scans

New: I have chosen to be quite selective concerning the material I will present this afternoon. I have decided that of the greatest potential importance with respect to the various EMI scanners we have used is a modification developed by Mr. David King, with the advice of Dr. Romeo Ethier at the Montreal Neurological Institute and currently under clinical evaluation by Dr. Ethier. This is the modification of the CT 5005 general purpose scanner, termed high resolution sector scanning. It is important to note that in this system the traverse speed and distance of the scanner are unchanged from normal operation. The total radiation dose involved in sector scanning is therefore unchanged and is received by the entire body section scanned, rather than being concentrated in the area or organ of interest. I believe that in future there will be a significant place for the option of increasing the dose to the area of interest. Section thickness options are not yet finally determined, but a range of section thickness of 2, 4, 8, or 12 mm is envisaged. Pixel widths as a function of resolution in normal operation in this unit, are 1 nun for a 13-in. scan field and 0.75 mm for a 10-in. scan field. With the high resolution mode, pixel widths are halved, to 0.5 nun and 0.37 mm respectively. The normal scan fields utilized in the CT 5005 are 320 mm and 240 mm, and with high resolution sector scanning, 160 and 120 mm. The total measurements in a translation are normally 761 for the larger field and 711 readings for the 10-in. field. These

Section thickness2,4,F8,,or 12 mm

Traverse speed and distance tunchanged Pixel w.idth s: ?No~rm~l 1.0/0.75 mm ,t~R 0.5/0.37 mm Fields: Normal 320/240 (13 or 10 in.) HR 160/120 (6.5 or 5 in.) Total pulses N 7611711 Total pulses N 761/711

are unchanged in the high resolution sector scanning mode. The geometry of the system is summarized in Figure 8 and the operating parameters are shown in Table 2. Assuming a basic 13-in. scan field which just encompasses the trunk, in the normal scan mode, the outer segment beyond the body is used for calibration and data acquisition across the body is also at approximately 0.6-mm intervals. In the sector scanning mode, the measurements outside the region of interest are still obtained at 0.6-mm intervals. These measurements are used for selection of data sufficient to measure the total attenuation along a beam path, so that there is the possibility of applying corrections to the measured data within the region of interest. Without such information beyond the region of interest, the measured data of the reconstruction could not provide accurate quantitative information.

Third Panel Discussion. The Ideal Head CT Scanner

531

The potential for more accurate attenuation measurements within the region of interest requires at least selected data processing outside of the area. Within the region of interest, measurements are obtained at 0.3-ram intervals. To date, the sector scan field used has been 6.5 in. in diameter. Sector scans of 5.5-in. diameter are to be available almost immediately and smaller fields are envisaged. Another modification which has, I think, significant bearing, particularly with respect to parsimony of dose by the reduction of radiological examinations providing very similar information, is the use of thinner than usual CT sections to improve resolution in the thickness of the slice, in addition to improved resolution in the plane of section. With high contrast tissues such as bone, good quality detail can be obtained without significant increase in CT scan dose with section thickness of 5 mm or less. One can envisage improvements that will permit good quality scans at 2 or 3 mm in the z-axis. Such a capability should reduce significantly our present dependence u p o n polydirectional laminography. Clearly, however, CT scanning can never achieve the exquisite bone detail of radiography and we shall continue to depend upon the latter for assessment of fine textural bone changes. CT/T 8800 Scanner (General Electric Company): Matching CT Images With Plain Radiograms Haughton: I wish to illustrate briefly the CT/T 8800 scanner and describe in more detail the 'scout view,' a localizing feature that provides a plain frontal or lateral image to which the computed tomograms may be referred. The CT/T 8800 scanner has a large aperture, with a rotating tube and a xenon detector containing 511 elements. Five and 10-mm slices are available and the speed of scanning is 5 to 10 s. To obtain the scout view the detector and the tube in the scanner are held stationary, the beam is pulsed and collimated to 3 mm an d the patient is moved the chosen distance through the gantry. Promptly after the image has been obtained, this is shown on the display console. The radiation exposure for the scout view is between 100 and 200 millirads. Because of the accuracy in selecting the levels for CT scanning, the examination time is reduced, the radiation exposure per patient is diminished, and the scans can be obtained with high reproducibility from patient to patient and time to time. One of the most important applications of the scout view is the spine. If one is going to be able to diagnose herniated discs accurately with CT scanning it will probably be necessary to find the exact levels for imaging and also to tilt the gantry so that the plane of cut is perpendicular to the spinal canal. Figure 9a illustrates a lateral view of the lower trunk with a cursor line tilted to intercept the superior vertebral endplate of the body of L-5. This is the bottom edge of a 5-mm-thick CT slice. In Figure 9b

Fig. 9a and b

the tomogram obtained at the level of the line cursor is shown in the magnification mode. The scout view is also valuable for head CT studies. Structures identified in the CT image are referred accurately to a lateral or a frontal skull X-ray: reference lines such as the Taylor-Haughton line can be projected on the CT image. Series 100 Scanners (Ohio-Nuclear, Inc.): L o w Cost Scanning Weinstein: The other speakers today will be showing you the Mercedes of CT scanning. What I'll be showing you I like to think of as the Volkswagen of CT scanning. Specifically, what I'm talking about is a lower cost tomograph. This is a simplified scanner with a lower purchase price, easier maintenance and hopefully fewer, more easily repairable breakdowns. In Figure 10 the X-ray cooling and control units of the Delta 25 (Fig. 10a) and the Delta 100 (Fig. 10b) are compared. Note the marked re-

532

Third Panel Discussion. The Ideal Head CT Scanner

Fig. lOa and b

Fig. 11

Fig. 13a

Fig. 12

Fig. 13b

Third Panel Discussion. The Ideal Head CT Scanner duction in size of the 100 units. In the 100 series, many simplifications have been made. Obviously, one does not have all the advantages of the Mercedes scanner. In the 100 series the scan time is 2 min; it is a single-slice scanner and the slice thickness is uniformly 10 ram. But, for me,. the important thing is the quality of the images one can generate on the scanner, and Ill just show you some examples with both high and low contrast resolution. Here we see the differentiation of the white and gray matter which the scanner is capable of(Fig. 11), an orbital study (Fig. 12), and a coronal scan (Fig. 13a) with detail magnification (Fig. 13b) of the skull base at the level of the anterior clinoids-tuberculum sellae.

0200FS and 0450 Scanners (Pfizer Medical Systems, Inc.): Operator-Computer Direct Communication Rothman: The Pftzer 0200FS and 0450 scanners have been designed to maximize the usefulness of the very expensive computer hardware built into the system. It is this feature of the scanning system which I will discuss this afternoon. The ideal CT scanner must allow operator control over the entire scanner operation. Unlike many CT scanners, the 0200FS and 0450 allow the operator to communicate directly with the computer. A wide variety of software subroutines are available to aid the operator in obtaining information from the scan: (1) area of interest; (2) statistical analysis; (3) direct area and linear measurement capability; (4) on-line multiplanar reconstruction; and (5) radiation therapy planning programs. Aside from these supplemental programs which may be found on different CT systems the scanner is unique in that the operator can directly interact and change the scanner program to his own specifications. In a program called SET UP all of the more than 200 variable switches and 75 mathematical constants are displayed. Each can be changed by typing new values into the computer, creating new operating programs. Several of these possible changes are important enough to illustrate. The 0200FS is a rotation-translation scanner. The standard 20-s head scan is performed by completing eight rotations and nine translations. The maximum radiation dose to the skin on the tube side of a single slice is approximately 2 rads. By changing the switches in the set-up file one can have the scanner translate twice for each rotation, doubling the dose. It is also possible for the operator to slow the speed of the translation down by a factor of 2 and double the dose in that manner. Of equal importance is the matter of convolutional filtration. The mathematical algorithm used in the program is called a filtered convolution back projection algorithm. This means that a mathematical filter is superimposed on the convolved data to improve the resolution of the system. In most scanning systems these filters are modified by adding fixed amounts of smoothing or edge enhancement. The 0200FS and the 0450 are desig-

533 ned to allow the operator to choose from any of three mathematical filters, fifteen different smoothing f u n c tions, and five edge-enhancement variables. After evaluation of the various possible falter, smoothing, and edge. enhancement combinations the radiologist can decide which image he thinks is best and create a scanning pro. gram to fit his own preference. By changing one flag in the program it is possible to have the actual raw data stored on disk or magnetic tape. This allows the operator the flexibility of recomputing the scan using any of the filters, smoothers and edge enhancers without radiating the patient more than once. Scans may be reconstructed using the standard Shepp and Logan, or Ramachandran and Lakshminarayanaia filters with no added smoothing. They are almost indistinguishable. By changing one constant from 0.0 to 1.0 it is possible to add smoothing to the Shepp and Logan filter. Two other variable falters are available, one called W, which is an edge-enhanced function, and another, G, a smoothing function. Because the computer is available to us through the teletype we are limited only by our own ingenuity as to what other scanner-related programs can be written. Other research programs have been written to calculate bone mineral, to detect edge position, scan subtraction, and others. All of these call for the manufacturer to open the system to the radiologist. By providing such things as a FORTRAN or basic compiler with the scanner the 0200FS and 0450 have become a major resource to the department. Not all of the advantages need to be scientific. When the computer is accessible to us it allows us to fill in our late night and early morning hours when patier/fs are asleep with a good chess game against a sophisticated and tireless opponent. Tomoscan 200 (Philips Medical Systems): True Partial Scan Technique Chase: The trends in the geometry of the computed tomographic devices reminds me of the Paris fashion industry: always interesting, sometimes exciting, never immutable. The feature I will discuss is the use of the partial scanning technique, which was developed with the support of the Philips Medical Systems and in cooperation with Dr. Manlio Abele in the research laboratories of the Department of Radiology of New York University. This incorporates a true partial scanning technique unlike those that, to the best of my knowledge, have been described by other manufacturers. Only the region of interest is scanned and there is no overscanning. Most of the scans are of a 3-in. diameter: only 3 in. are scanned and 3 in. are reconstructed. The advantage and unique feature of this approach is that it concentrates the photons in the 3-in. circle; the area outside the 34n. circle is not of any importance. A new algorithm is used where differential densities are

534 utilized to reconstruct the image. The deficiency of our approach is that in the regional scans we cannot get accurate CT numbers. The regional scans, to the best o f our knowledge, are probably of greatest use when studying bone or areas as the orbit where there are high contrast structures such as fat against muscle and fat against nerve. The new approach allows for a great deal of versatility whether one takes advantage of the potential spatial resolution or of the potential density resolution. I sincerely believe that one way of obtaining fine detail and fine differences in density is the 'true partial scans technique.'

T-R 120 and S-D Scanners (Picker Corporation): High Speed Scanning

Cook: In designing the ideal head scanner, scan speed is an important consideration. In our high volume practice, the single most common artifact encountered is caused by motion. Because of motion artifacts, scans may have to be repeated. Frequently we have to resort to drug therapy to eliminate motion and, occasionally, we have to enlist the aid of an anesthesiologist. The single most common clinical condition encountered in our practice associated with motion artifacts is uncooperative patients due to cerebrovascular accidents. The ability to obtain a fast scan will hopefully result in a better examination and a decreased use of drags in critically ill patients. Scan speed may be defined in one of several ways, that is, the time needed to complete one section, serial sections, or an entire study. The Picker stationary detector or generation 4 machine is capable of obtaining sections in 1 s. Assuming eight sections of the head, a scan section time of 3 s and a recycle time of approximately 4 s, we can complete a study in slightly less than 1 rain. The' usefulness of scan speed will depend upon its definition and upon its possible future applications. In • conclusion, I would like to challenge the manufacturers to give serious consideration to scan speed, both as it applies to image generation and to the acquisition of physiologic data.

Pho/Trax 4000 Scanner (Searle CT): Low Dosimetry Scanning

Churchill: The Pho/Trax 4000 is a third-generation, rotational scanner which utilizes a fixed anode X,ray source and 504 xenon detectors. Some of the advantages which I feel are included in this system are the selectable number of views or profiles per scan, variable slice thickness which can be varied from 3 to 12 mm in l-ram increments, and the ability to reconstruct any area of interest in the original image on a smaller size circle from 5 to 50 cm in centimeter increments. Direct coronal imaging is facilitated by the tiltable gantry which has a 60-cm aperture.

Third Panel Discussion. The Ideal Head CT Scanner

Maximum skin dose Central dose

Single 1-cm slice

10 contiguous 1-cm slices

0.54 rad 0.1 rad

0.67 rad 0.4 rad

Fig. 14

Figure 14 shows the dosimetry which was measured on our prototype scanner. This figure represents what I feel to be the most important advantage with this scanner, namely, the low radiation dose delivered to the patient. For a 1-cm-thick slice at 130 kVp and 30 mA taken with the 5-s scan mode, the maximum skin dose is 0.54 rad and the central dose is 0.1 rad. For ten contiguous 1-cm-thick slices the skin dose is 0.67 rad and the central dose is 0.4 rad. You can see that there is approximately a 25% increase in skin dose with contiguous slices. The' dosage is doubled with each doubling of the scan time. The dosage delivered with a production model of the Pho/Trax scanner is slightly higher than on the prototype. The dose at 130 kVp and 30 mA using a 1-cm beam width and a 5-s scan is 0.8 rad at the brow. Remember that at 10 s the dose is 1.6 tad and at 20 s the dose is 3.2 rad. Likewise the exposure will be 25% higher with ten contiguous 1-cm-thick slices. We perform head scans with the 20-s scan time so the skin dose is approximately 2.2 rad per slice. Some institutions use the 10-s mode with the production model and the skin dose at the brow is 1.6 rad per slice.

Siretom 2000 Scanner (Siemens Corporation): Operational Ease Archer: My experience with the Siretom 2000 includes 2100 examinations performed during an 8-month period. Approximately 10 to 12 patients, 80% of whom had both a noncontrast and contrast study, were examined in an 8-h work day. I would like to emphasize two features: (1) the ease, speed, and reliability of operation; (2) the consistently high quality images in the transaxial and coronal planes. This has assured high quality sagittal and coronal reconstructions, which are in great demand in our hospital. To consider the first point: the equipment can be made ready to begin an examination in less than 3 min, because the dialogue between the operator and the computer has been reduced to a minimum. Further, no warmup period is required. When standard operating procedure is used, computer commands including patient information and choice of program can be delivered in about 1 to 1.5 min. When free choice is exercised, the time is increased to about 2 to 2.5 min. The positioning of the patient is accomplished with ease and efficiency; the fact that neither a water bag nor bolus material are needed, as well as the fact that light localizers are provid-

Third Panel Discussion. The Ideal Head CT Scanner ed, facilitates proper and rapid head positioning. Subsequent operation is controlled from the console. Under most circumstances we use the 69-s scanning time which provides us with two contiguous slices of 5 or 10-ram thickness. In my experience, movement artifact has been no more noticeable with the l-rain scanning time than with a 13-s scanning time. This is not true when the time is reduced to 5 s. At the end of the scanning time two contiguous slices are immediately available and are displayed in a one-to-one representation on a TV monitor. During the l-rain scanning time, the console is available for recording and viewing images. Thus a single examination can be performed and recorded on floppy disc and film in approximately 10 to 11 rain. Our down time has been 6%. In summary, the Siretom 2000 has offered ease, speed, and reliability of operation. I would like to have further improvement in terms of slices thinner than 5 mm and faster scanning times. Access to CT numbers in print form were not available during this 8-month period, but will become available shortly.

V-360-3 CT Scanner (Varian): Direct (Positional) Sagittal Scanning

Anderson: We have been using the Varian scanner particularly for body scanning, in view of the fact that a dedicated head tomograph is also available to us. In the interest of brevity, let me point two outstanding characteristics of the Varian scanner, one of which is advantageous and the other unique. The advantageous feature is the capability of 3-s scanning, which is one of the fastest among the available commercial devices. The unique feature of the Varian scanner is its very large gantry opening. This large gantry opening was not made for neuroradiologic use, but since it was available to me I attempted to see what advantage it might have for neuroradiology. One advantage relates to coronal scanning. The patient can lie in a comfortable position on his side with very little extension of the neck necessary to get a coronal view. Another unique application of this large gantry is the capability of scanning directly in the sagittal plane. This can be done quite easily. We have found direct sagittal scanning to be of value. The direct positional sagittal scan has certain advantages compared to computer-generated sagittal scans. The whole head can be imaged, resolution approaches maximum, the dose does not exceed that used for an axial scan, and no special computer capacity is needed. Current disadvantages include artifacts created by the interference of the patient's shoulders and some discomfort in positioning for this view. In conclusion, perhaps the ideal head scanner of the future should have a large gantry Opening to allow us to position patients for direct coronal and sagittal scanning so that we can work with multiple views in a manner similar to other X-ray examinations.

535

Summarizing Comments lsherwood: I feel very privileged to have been asked to summarize this session on the ideal head scanner. It is important I should mention at the beginning that my own experience is based only on EMI scanning equipment. I have no experience on other scanners, but objectivity is the name of the game and I will try to be objective. The definition of ideal, I think, is important. I know this is stepping slightly outside our brief but, nevertheless, I do feel that we should recognize the difficulties pointed out earlier in this meeting by our former President. These, of course, concern the costs of some of the newer technologies in health care systems with differing social and political priorities. In this Symposium and in this Panel we are talking about the Rolls Royce CT Scanner, 'but there are colleagues in this room now who do not have the facility at all and their definition of 'ideal' would, I guess, be quite different. Perhaps the most important criterion for a CT scanner in any circumstances should be: does it exist, does it work, will it continue to work and can it be obtained at low cost? It is a truism that we only know what we want when we are made aware what is possible. There are enormous commercial pressures at the moment on the profession, and these are influenced b y the commercial interpretation of what we as a profession want, rather, perhaps, than what we really need. I think it is necessary to examine, critically, some of the newer developments and options and apply both clinical judgement andlateral thinking. Dr. Di Chiro has asked me to speak to the 'aristocracy' about an 'ideal' machine, and that is what I shall try and do. There are some things which stand out clearly from what we have heard. Firstly, accuracy, precision, and stability of attenuation values are important requirements. Dr. Kricheff, earlier in the Symposium, noted the precision and stability of the attenuation values in a variety of scanners, but many of these values were, in fact, inaccurate. Historically, the water bath scanner was extremely accurate, but the scan time was too long without a totally stationary object. If we are to investigate tissue characterisation, dynamic tumour responses, periventricular lucencies, and many of the other things which academic centres wish to do, then accuracy, precision, and stability are all absolutely vital. This is not to deny that much routine diagnostic work will not continue to be carried out by the simple morphological description. I think the second most important group of interrelated features are the identification, relocation, and reproducibility of sections. We suggested in Miami some eighteen months ago a very simple radiographic technique developed in Manchester for obtaining reproducible CT scans. There are undoubted advantages to be obtained from the present scanogram facility available in several units, particularly increased speed of operation and precision of slice selection. This is an option that we really all require, though even with it reproducibility remains

536 one of the biggest problems in CT. If we are going to exploit dual energy scanning for atomic number studies, subtraction for cerebral blood volume, physiological studies of neural tissue and bone mineral in the spine, we certainly need very accurate reproducibility. The means not only being able to select a section with precision but being able to get back to it again with absolute accuracy. This requirement becomes increasingly important the thinner the section, and the higher the resolution. It also then, of course, becomes increasingly difficult. The scanogram is an important step forward, though there is still much work to be done. I think with high resolution scanograms we are looking towards some sort of general purpose radiographic machine incorporating a variety of diagnostic facilities and options. This is an area then, where we may have to rethink our total philosophy about information retrieval, dose efficiency, and radiological strategies. There seems to be a clear pointer to what I would like to see, and that is a universal imaging machine. Higher resolution, as one of the speakers referred to it, rather than high resolution in CT images, is clearly going to be an important facility, particularly if it can be achieved with low dose. Dose becomes an increasingly important factor when it is confined to a small area and where the whole of a sensitive structure is contained within the slice. If we are to recover density discrimination as well as achieve higher resolution then, of course, we are going to have to face a significant dose penalty. It may be possible with edge reconstruction techniques to reduce this, but only then by sacrificing attenuation values. Higher resolution for limited area reconstruction in the spine and, perhaps, the petrous bone, is an important objective to pursue. A machine incorporating this facility might then eliminate conventional tomography. The spinal cord is a longitudinal structure and we need to know the anatomical level with accuracy for any detailed study. If a good method of longitudinal imaging were available, particularly if that, too, were sectional, together with the option of thin transverse higher resolution sections at selected levels with a low dose penalty,

Third Panel Discussion. The Ideal Head CT Scanner then such a unit really would be ideal. We must, never-. theless, at present investigate the trade-offs associated with lower resolution smoothing and spatial f'fltering. One cannot emphasise too strongly the fixed relationship of dose, resolution, and density discrimination. Quantitative information provided by receiver operating characteristic curve studies are really necessary to ensure that image quality is being retained. Reconstruction or positioning to obtain other planes of section have been stressed. There are undoubted advantages in both methods especially for lesions near the skull base. Those options becoming available with advancing software for producing oblique sections and rotating images certainly give some conceptual information, but their real clinical value might be doubted. We need to be selective in the use of new options and handle them with good clinical sense. Speed of scanning has been mentioned. This seems to be a question of stability versus cost. One would like to think that only a small number of patients require very short scan times to avoid movement artefact but, of course, even vascular pulsation in the body can be a problem. If intravenous CT angiography is going to play a significant role in the future, and it seems that it might, and especially if we can look towards a universal machine, then the shorter the scan time, the better. Perhaps flexibility in scan times should be possible. Certainly, for most conventional brain scans longer times would be acceptable. Even at one second, physiological movement in the body cannot be eliminated and, in some circumstances, gating techniques need to be considered. Console options are very much a matter for the individual, but simple things like clear alphanumerics and separate controls with flexible options are important features. The comment about computer access is well taken. Many users would like to see immediate access to computer, in order to exercise local software options. I have really enjoyed trying to bring this discussion together and I hope I have achieved some degree of success.