Radiation Physics
Anthropomorphic Test Objects for CT Scanners 1 Gary D. Fullerton, Ph.D.,2 and David R. White, Ph.D.
Numerous types of artifacts can result in clinical misinterpretations in computed tomography. The authors have evaluated phantom materials and techniques of construction which would produce objects similar enough to the human body to test the accuracy of CT in vivo. The object used for this purpose is a thoracic section with five soft-tissue materials, representing adipose tissue, muscle, heart, esophagus, and lung, and three bone materials representing cortical bone, average rib, and "inner" (trabecular) bone. The accuracy of CT numbers for studies of different organs was assessed and sources of error in five commercial scanners were identified. INDEX TERMS:
Computed tomography, instrumentation. Computed tomography. thorax, 6 [0 J. 1211
Radiology 133:217-222, October 1979
HE RAPID
development of diagnostic and therapeutic
numbers in thoracic scanning in order to determine whether we would be justified in using in vivo CT measures of lung density for inhomogeneity correction in radiation therapy. The thorax was selected for this purpose because it includes both low-density lung and large high-density bones and therefore would be expected to produce the largest errors due to shape and composition perturbations: thus data obtained from the thorax would allow us to determine an upper limit for expected errors in other body sections (with the possible exception of the pelvis). Evaluation of five commercial scanners showed a maximum error of -50 H for soft tissues adjacent to bone (assuming air to be -1,000 and water 0) and a maximum error of +45 H in measuring lung densities due to improper calibration. Typical shape-related errors in measurement for soft tissue were less than 10 H in most locations.
Tprocedures based upon computed tomography (CT) and variations in CT number urgently requires an understanding of the dependability of this information. Early reports indicate several potential sources of error (1, 2), some of which result from the irregular shapes and densities of the human body. These errors vary with different manufacturers according to the reconstruction algorithm and other design considerations. The apparent high quality of CT images. combined with lack of experience, has resulted in some identifiable clinical misinterpretations involving these artifacts, raising the question of whether there are other as yet unidentified shape-related artifactual problems. Recently several investigators have proposed direct use of CT numbers for diagnostic purposes, while others have suggested using CT numbers for inhomogeneity correction in radiation therapy (3-6), all of which makes evaluation of the accuracy of CT numbers in vivo even more necessary. The first purpose of the present study was to develop anthropomorphic test objects which would be adequate to evaluate computed tomography. The primary problems-formulation of materials that would precisely simulate human tissues within the contrast resolution capacity of CT scanners and formation of appropriate organ shapes-were already nearly resolved by the work of White et al. (7-9). Materials were optimized and evaluated by us for the range of x-ray energies typical in transmission CT. The phantoms and materials described in this paper were prepared at St. Bartholomew's Hospital in London by one of us (OW.), while experimental evaluations were done at various U. S. centers by the co-author (G.F.). The second purpose of our work was to determine the extent of shape-related errors in soft-tissue organ-specific CT
MATERIALS AND METHODS
Tissue Substitutes
Three principal factors governed the choice of base materials used in the tissue substitutes which made up the body phantom: (a) The need for a homogeneous mixture of base material and corrective particulate fillers meant that wax-based substitutes, despite their good casting and molding properties. were unacceptable due to poor filler dispersion and air entrapment (7). (b) An ongoing program of phantom design and manufacture requires reproducibility, necessitating a base material with well-defined elemental composition. (c) The choice was limited further by the need to readily simulate a wide range of tissues so that a high degree of realism could be maintained.
1 From the Therapeutic Radiology Department, Health Sciences Center, University of Minnesota, Minneapolis, Minn. (G.D.F.), and the Radiation Physics Department, St. Bartholomew's Hospital, London, England (D.RW.). Received July 20, 1978: accepted and revision requested Oct. 26; revision received March 12, 1979. This investigation was supported in part by Public Health Service research grant CA-15548 from the National Cancer Institute. 2 Present address: Department of Radiology, Medical Physics Division, University of Texas Health Science Center at San Antonio. 7703 Floyd Curl Drive, San Antonio, Texas 78284. sjh
217
218
GARY D. FULLERTON AND DAVID
b
2P~~©~ \ (~~"\ \\
IN (b\d\~ ~ LNF4 ~"b-AP6 r
I
(a-RB2
I
I
~
\)
-_~~
......_.---
c-WTl )
\
WHITE
October 1979
added. (Details of these products will be reported shortly.)
- - - - - -__
·1
R.
Phantom Construction
'
':
\
1/\\
jV\
Iii I
Ij~/
cV -----------834{
-_._--
Fig. 1. Simplified outlines of thoracic section 26 from A Cross-Section Anatomy (7) as simulated in the antlYopomorphic CT (ACT) phantom.
Since epoxy-resin-based substitutes have already been used in the production of realistic body phantoms (8), they were chosen as the most convenient base material. There is a wide range of such tissue substitutes whose end products are homogeneous, air-free, and reproducible in both elemental composition and mass density (9); moreover, their casting characteristics are ideal for fabrication of small, intricate segments representing such structures as the ribs, esophagus, and vertebrae. The particulate fillers added to the liquid epoxy systems must not only modify the attenuation properties (i.e., J.1/ ,) and I)) of the basic materials to those of the tissues being simulated but also maintain sufficient viscosity that settlement and flotation are negligible and trapped air is efficiently removed. Compounds such as aluminum oxide, calcium carbonate, magnesium oxide, and Teflon powder are used as attenuation modifiers, phenolic microspheres as density modifiers, and, when required, polyethylene powder as a viscosity modifier. Particle sizes were restricted to 1-50 J.1m. Whenever possible, compounds containing elements normally found in body tissues are employed. Only products with elemental atomic numbers below 25 are used, thus avoiding problems with photoelectric absorption edges which occur below 10 keV for these elements; in addition, difficulty in maintaining adequate filler dispersal, a severe risk when minute quantities of high-atomicnumber compounds are used, was minimized.
Forming Techniques Techniques of manufacturing non-foam substitutes have been described in detail by White et al. (9). Essentially, the liquid resin system and particulate fillers are accurately weighed into a Pyrex reaction vessel and mixed mechanically under vacuum conditions. The mixture is then poured into a suitable mold and left to cure and harden. Similar techniques are used for foam substitutes, except that in addition to the liquid resin system and corrective fillers. small quantities of a foaming agent and surfactant are
The thoracic section chosen for replication was Section 26 from A Cross-Section Anatomy (10) (Fig. 1). It was specified to be 4-5 cm thick, with a constant cross section throughout (i.e., no attempt was made to replicate the three-dimensional structure of the section at this stage) and was roughly 20 X 33 cm in maximum sagittal and coronal dimensions. The basic manufacturing procedure comprises four stages. (a) A cavity representing the shape of the tissue or organ is cut from a 6-cm-thick block of expanded polystyrene, using a hot wire. (b) An alloy with a low melting point (Lipowitz's metal) is poured into the cavity, producing a solid replica of the segment. (c) After the metal block has been cleaned and polished, a silicon rubber mold is made. (d) When the rubber mold has cured, the appropriate epoxy-resin-based substitute is poured into it, yielding a replica with a constant geometry. The four operations are repeated for each of the main components of the thoracic section. In this case, the outer adipose strip, both lungs. heart, ten ribs, sternum, two cartilage segments. esophagus, spinal cord, and vertebrae are manufactured first. These components are then oriented on the baseplate and an epoxy muscle substitute poured into the spaces between them to bond the segments together. When the muscle resin has cured, the section is faced on a milling machine. The resulting phantom contains materials simulating eight types of tissue: Adipose (AP6) Bone Cortical (S83) Inner (186) Average rib (R82) Lung (LNF4) Muscle Low (WT1/L) Medium (WT 1) High (WT 1/H) The ICRP formulation for adipose tissue ( 11) was used as the basis of the outer adipose strip (AP6). Bone was divided into three categories as indicated above. The cortical bone substitute, S83, simulates the composition reported by Woodard (12); "inner bone" (186) is a mixture of 22.4 % cortical bone and 77.6 % soft tissue and represents an average trabecular bone/red marrow component. The ribs and sternum (RB6) are simulated by a component representing an average of 46.5 % cortical bone and 53.5 % soft tissue (13). Lung substitutes (LN4F) are based on the elemental and specific gravity data given by the ICRP (11).
Vol. 133
219
ANTHROPOMORPHIC TEST OBJECTS FOR CT SCANNERS
Radiation Physics
2a,b
2e,d
Fig. 2. a-d. Scans of the ACT phantom indicating regions in which CT numbers were evaluated for WT1/L (a), WT1/H (b), LNF4 (e), and WTlIL in the area with the most bone artifacts (d).
Three types of "muscle" tissue (low, medium, and high density), used for cartilage, heart, spinal cord, and general muscle tissue, are represented by "water-like" substitutes (WT1Il, WT 1, WT1IH), with an overall spread of CT numbers within ±5 % of the data for water. Materials Testing
Samples of AP6, WT 1Il, WT 1, and WT 1IH were cast in the form of small cylinders 3 or 5 cm in diameter, while samples of RB2 were formed into bars 1 or 2 cm square in cross section. Density: The densities of all samples were measured to an accuracy of ±0.002 q/crn", using Archimedes' principle in distilled, air-free water and/or methanol with a Mettler Model B balance. CT Numbers (Energy Dependence): Three samples of WT 1 and AP6 from different batches and single samples of WT1/l, WT1/H, RB2, water, and air were scanned under identical conditions with an ACTA 0200FS scanner (140 kVp, 30 mA) at the center of a cylindrical water bath 20 cm in diameter to give shape-independent CT numbers under "good scanning conditions." Scans were then repeated with 6 or 12 mm AI additional filtration to demonstrate energy-dependent variations in CT number due to the atomic composition of the tissue substitutes. Samples of AP6, WT 1Il, WT1, and WT 1IH were also scanned on an ACTA 0100 (120 kVp, 20 mA) and an EMI head unit (120 kVp, 30 mA) as well as the ACTA 0200FS to further demonstrate the spectral dependence of CT numbers in a variety of scanner configurations.
Phantom Testing
The ACT phantom was scanned on ACTA 0200FS, GE CT IT, Ohio Nuclear ~-50, EMI 5000, and EMI 5005 units. Each object was positioned and control parameters adjusted by the technologist to optimize the scan for a patient having the same dimensions as the phantom. Mean CT numbers for the components of the phantom were evaluated in roughly identical locations for all scanners as indicated in Figure 2. In each instance a minimum of 100 pixels was averaged to reduce the standard error of the mean to less than 0.1 % in uniform regions. CT numbers of water and air were also measured in a water bath 20 cm in diameter -equivalent to the ACT phantom in size--with settings identical to those used for the ACT phantom. RESULTS
Materials
The measured densities of the tissue substitutes are given in TABLE I along with CT numbers (H m ) calculated with 0,6. and 12 mm AI additional filtration with the ACTA 0200FS scanner. All reported CT numbers represent the mean over at least 100 pixels. The spread was independent of the material, and one standard deviation was typically 6 H or 0.6 %. The CT numbers for air and water allowed us to adjust for the small inaccuracies in scanner calibration observed with changes in total filtration (e.g., phantom size). The 6- and 12-mm AI filtration configurations therefore permitted us to isolate the effect of the
GARY D. FULLERTON AND DAVID R. WHITE
220
October 1979
TABLE I: MEASURED (Hm ) AND CALCULATED "IDEAL" (H) HOUNSFIELD VALUES' No Added Filtration
Density Material Water Air WT1/L WT1 WT1/H AP6 RB2
(g/cm 3 )
Hm
0.998 0.001 1.015 1.045 1.072 0.971 1.298
-0.2 -971.8 1.1 30.7 56.9 -80.5 514
6mmAI
12 mm AI
Hm
H
-22.7 -980.0 -22.8 6.4 28.3 -100.3 453.2
0 -1,000 1.2 31.7 58.6 -82.7 529
Hm
H
13.8 -939 15.8 43.7 67.1 -61.6 478.8
0 -1,000 -0.1 30.4 53.3 -81.1 497.3
H
0 -1,000 2.1 31.4 55.9 -79.1 488.0
• Measurements were made in a 20-cm water bath, using the head scan mode and a software linearity correction. TABLE II: EVALUATION OF ACT PHANTOM MATERIAL SAMPLES ON THREE TYPES O~ SCANNERS
100 Liver Muscle (steer) Muscle(monkey) poncreosfmonkey)--..A
80
.r:
60 ~
"c ::l
-0 Q)
c
(monkey)
~~s~;~H
Heart _ .
20
::l
0
E
A-c.
A_ _ po;~os --Brain O"---Plasma o~SF
40
;;:
Anthropomorphic Test Objects for CT Scanners 1 Gary D. Fullerton, Ph.D.,2 and David R. White, Ph.D.
Numerous types of artifacts can result in clinical misinterpretations in computed tomography. The authors have evaluated phantom materials and techniques of construction which would produce objects similar enough to the human body to test the accuracy of CT in vivo. The object used for this purpose is a thoracic section with five soft-tissue materials, representing adipose tissue, muscle, heart, esophagus, and lung, and three bone materials representing cortical bone, average rib, and "inner" (trabecular) bone. The accuracy of CT numbers for studies of different organs was assessed and sources of error in five commercial scanners were identified. INDEX TERMS:
Computed tomography, instrumentation. Computed tomography. thorax, 6 [0 J. 1211
Radiology 133:217-222, October 1979
HE RAPID
development of diagnostic and therapeutic
numbers in thoracic scanning in order to determine whether we would be justified in using in vivo CT measures of lung density for inhomogeneity correction in radiation therapy. The thorax was selected for this purpose because it includes both low-density lung and large high-density bones and therefore would be expected to produce the largest errors due to shape and composition perturbations: thus data obtained from the thorax would allow us to determine an upper limit for expected errors in other body sections (with the possible exception of the pelvis). Evaluation of five commercial scanners showed a maximum error of -50 H for soft tissues adjacent to bone (assuming air to be -1,000 and water 0) and a maximum error of +45 H in measuring lung densities due to improper calibration. Typical shape-related errors in measurement for soft tissue were less than 10 H in most locations.
Tprocedures based upon computed tomography (CT) and variations in CT number urgently requires an understanding of the dependability of this information. Early reports indicate several potential sources of error (1, 2), some of which result from the irregular shapes and densities of the human body. These errors vary with different manufacturers according to the reconstruction algorithm and other design considerations. The apparent high quality of CT images. combined with lack of experience, has resulted in some identifiable clinical misinterpretations involving these artifacts, raising the question of whether there are other as yet unidentified shape-related artifactual problems. Recently several investigators have proposed direct use of CT numbers for diagnostic purposes, while others have suggested using CT numbers for inhomogeneity correction in radiation therapy (3-6), all of which makes evaluation of the accuracy of CT numbers in vivo even more necessary. The first purpose of the present study was to develop anthropomorphic test objects which would be adequate to evaluate computed tomography. The primary problems-formulation of materials that would precisely simulate human tissues within the contrast resolution capacity of CT scanners and formation of appropriate organ shapes-were already nearly resolved by the work of White et al. (7-9). Materials were optimized and evaluated by us for the range of x-ray energies typical in transmission CT. The phantoms and materials described in this paper were prepared at St. Bartholomew's Hospital in London by one of us (OW.), while experimental evaluations were done at various U. S. centers by the co-author (G.F.). The second purpose of our work was to determine the extent of shape-related errors in soft-tissue organ-specific CT
MATERIALS AND METHODS
Tissue Substitutes
Three principal factors governed the choice of base materials used in the tissue substitutes which made up the body phantom: (a) The need for a homogeneous mixture of base material and corrective particulate fillers meant that wax-based substitutes, despite their good casting and molding properties. were unacceptable due to poor filler dispersion and air entrapment (7). (b) An ongoing program of phantom design and manufacture requires reproducibility, necessitating a base material with well-defined elemental composition. (c) The choice was limited further by the need to readily simulate a wide range of tissues so that a high degree of realism could be maintained.
1 From the Therapeutic Radiology Department, Health Sciences Center, University of Minnesota, Minneapolis, Minn. (G.D.F.), and the Radiation Physics Department, St. Bartholomew's Hospital, London, England (D.RW.). Received July 20, 1978: accepted and revision requested Oct. 26; revision received March 12, 1979. This investigation was supported in part by Public Health Service research grant CA-15548 from the National Cancer Institute. 2 Present address: Department of Radiology, Medical Physics Division, University of Texas Health Science Center at San Antonio. 7703 Floyd Curl Drive, San Antonio, Texas 78284. sjh
217
218
GARY D. FULLERTON AND DAVID
b
2P~~©~ \ (~~"\ \\
IN (b\d\~ ~ LNF4 ~"b-AP6 r
I
(a-RB2
I
I
~
\)
-_~~
......_.---
c-WTl )
\
WHITE
October 1979
added. (Details of these products will be reported shortly.)
- - - - - -__
·1
R.
Phantom Construction
'
':
\
1/\\
jV\
Iii I
Ij~/
cV -----------834{
-_._--
Fig. 1. Simplified outlines of thoracic section 26 from A Cross-Section Anatomy (7) as simulated in the antlYopomorphic CT (ACT) phantom.
Since epoxy-resin-based substitutes have already been used in the production of realistic body phantoms (8), they were chosen as the most convenient base material. There is a wide range of such tissue substitutes whose end products are homogeneous, air-free, and reproducible in both elemental composition and mass density (9); moreover, their casting characteristics are ideal for fabrication of small, intricate segments representing such structures as the ribs, esophagus, and vertebrae. The particulate fillers added to the liquid epoxy systems must not only modify the attenuation properties (i.e., J.1/ ,) and I)) of the basic materials to those of the tissues being simulated but also maintain sufficient viscosity that settlement and flotation are negligible and trapped air is efficiently removed. Compounds such as aluminum oxide, calcium carbonate, magnesium oxide, and Teflon powder are used as attenuation modifiers, phenolic microspheres as density modifiers, and, when required, polyethylene powder as a viscosity modifier. Particle sizes were restricted to 1-50 J.1m. Whenever possible, compounds containing elements normally found in body tissues are employed. Only products with elemental atomic numbers below 25 are used, thus avoiding problems with photoelectric absorption edges which occur below 10 keV for these elements; in addition, difficulty in maintaining adequate filler dispersal, a severe risk when minute quantities of high-atomicnumber compounds are used, was minimized.
Forming Techniques Techniques of manufacturing non-foam substitutes have been described in detail by White et al. (9). Essentially, the liquid resin system and particulate fillers are accurately weighed into a Pyrex reaction vessel and mixed mechanically under vacuum conditions. The mixture is then poured into a suitable mold and left to cure and harden. Similar techniques are used for foam substitutes, except that in addition to the liquid resin system and corrective fillers. small quantities of a foaming agent and surfactant are
The thoracic section chosen for replication was Section 26 from A Cross-Section Anatomy (10) (Fig. 1). It was specified to be 4-5 cm thick, with a constant cross section throughout (i.e., no attempt was made to replicate the three-dimensional structure of the section at this stage) and was roughly 20 X 33 cm in maximum sagittal and coronal dimensions. The basic manufacturing procedure comprises four stages. (a) A cavity representing the shape of the tissue or organ is cut from a 6-cm-thick block of expanded polystyrene, using a hot wire. (b) An alloy with a low melting point (Lipowitz's metal) is poured into the cavity, producing a solid replica of the segment. (c) After the metal block has been cleaned and polished, a silicon rubber mold is made. (d) When the rubber mold has cured, the appropriate epoxy-resin-based substitute is poured into it, yielding a replica with a constant geometry. The four operations are repeated for each of the main components of the thoracic section. In this case, the outer adipose strip, both lungs. heart, ten ribs, sternum, two cartilage segments. esophagus, spinal cord, and vertebrae are manufactured first. These components are then oriented on the baseplate and an epoxy muscle substitute poured into the spaces between them to bond the segments together. When the muscle resin has cured, the section is faced on a milling machine. The resulting phantom contains materials simulating eight types of tissue: Adipose (AP6) Bone Cortical (S83) Inner (186) Average rib (R82) Lung (LNF4) Muscle Low (WT1/L) Medium (WT 1) High (WT 1/H) The ICRP formulation for adipose tissue ( 11) was used as the basis of the outer adipose strip (AP6). Bone was divided into three categories as indicated above. The cortical bone substitute, S83, simulates the composition reported by Woodard (12); "inner bone" (186) is a mixture of 22.4 % cortical bone and 77.6 % soft tissue and represents an average trabecular bone/red marrow component. The ribs and sternum (RB6) are simulated by a component representing an average of 46.5 % cortical bone and 53.5 % soft tissue (13). Lung substitutes (LN4F) are based on the elemental and specific gravity data given by the ICRP (11).
Vol. 133
219
ANTHROPOMORPHIC TEST OBJECTS FOR CT SCANNERS
Radiation Physics
2a,b
2e,d
Fig. 2. a-d. Scans of the ACT phantom indicating regions in which CT numbers were evaluated for WT1/L (a), WT1/H (b), LNF4 (e), and WTlIL in the area with the most bone artifacts (d).
Three types of "muscle" tissue (low, medium, and high density), used for cartilage, heart, spinal cord, and general muscle tissue, are represented by "water-like" substitutes (WT1Il, WT 1, WT1IH), with an overall spread of CT numbers within ±5 % of the data for water. Materials Testing
Samples of AP6, WT 1Il, WT 1, and WT 1IH were cast in the form of small cylinders 3 or 5 cm in diameter, while samples of RB2 were formed into bars 1 or 2 cm square in cross section. Density: The densities of all samples were measured to an accuracy of ±0.002 q/crn", using Archimedes' principle in distilled, air-free water and/or methanol with a Mettler Model B balance. CT Numbers (Energy Dependence): Three samples of WT 1 and AP6 from different batches and single samples of WT1/l, WT1/H, RB2, water, and air were scanned under identical conditions with an ACTA 0200FS scanner (140 kVp, 30 mA) at the center of a cylindrical water bath 20 cm in diameter to give shape-independent CT numbers under "good scanning conditions." Scans were then repeated with 6 or 12 mm AI additional filtration to demonstrate energy-dependent variations in CT number due to the atomic composition of the tissue substitutes. Samples of AP6, WT 1Il, WT1, and WT 1IH were also scanned on an ACTA 0100 (120 kVp, 20 mA) and an EMI head unit (120 kVp, 30 mA) as well as the ACTA 0200FS to further demonstrate the spectral dependence of CT numbers in a variety of scanner configurations.
Phantom Testing
The ACT phantom was scanned on ACTA 0200FS, GE CT IT, Ohio Nuclear ~-50, EMI 5000, and EMI 5005 units. Each object was positioned and control parameters adjusted by the technologist to optimize the scan for a patient having the same dimensions as the phantom. Mean CT numbers for the components of the phantom were evaluated in roughly identical locations for all scanners as indicated in Figure 2. In each instance a minimum of 100 pixels was averaged to reduce the standard error of the mean to less than 0.1 % in uniform regions. CT numbers of water and air were also measured in a water bath 20 cm in diameter -equivalent to the ACT phantom in size--with settings identical to those used for the ACT phantom. RESULTS
Materials
The measured densities of the tissue substitutes are given in TABLE I along with CT numbers (H m ) calculated with 0,6. and 12 mm AI additional filtration with the ACTA 0200FS scanner. All reported CT numbers represent the mean over at least 100 pixels. The spread was independent of the material, and one standard deviation was typically 6 H or 0.6 %. The CT numbers for air and water allowed us to adjust for the small inaccuracies in scanner calibration observed with changes in total filtration (e.g., phantom size). The 6- and 12-mm AI filtration configurations therefore permitted us to isolate the effect of the
GARY D. FULLERTON AND DAVID R. WHITE
220
October 1979
TABLE I: MEASURED (Hm ) AND CALCULATED "IDEAL" (H) HOUNSFIELD VALUES' No Added Filtration
Density Material Water Air WT1/L WT1 WT1/H AP6 RB2
(g/cm 3 )
Hm
0.998 0.001 1.015 1.045 1.072 0.971 1.298
-0.2 -971.8 1.1 30.7 56.9 -80.5 514
6mmAI
12 mm AI
Hm
H
-22.7 -980.0 -22.8 6.4 28.3 -100.3 453.2
0 -1,000 1.2 31.7 58.6 -82.7 529
Hm
H
13.8 -939 15.8 43.7 67.1 -61.6 478.8
0 -1,000 -0.1 30.4 53.3 -81.1 497.3
H
0 -1,000 2.1 31.4 55.9 -79.1 488.0
• Measurements were made in a 20-cm water bath, using the head scan mode and a software linearity correction. TABLE II: EVALUATION OF ACT PHANTOM MATERIAL SAMPLES ON THREE TYPES O~ SCANNERS
100 Liver Muscle (steer) Muscle(monkey) poncreosfmonkey)--..A
80
.r:
60 ~
"c ::l
-0 Q)
c
(monkey)
~~s~;~H
Heart _ .
20
::l
0
E
A-c.
A_ _ po;~os --Brain O"---Plasma o~SF
40
;;:
