T
The Peripheral Zone of Increased Density in Cranial Computed Tomography1
•
Neuroradiology
Mokhtar Gado, M.D., and Michael Phelps, Ph.D. The band of increased attenuation seen on in vivo CCT images of adults, frequently designated as the cerebral cortex, has been proved clinically and experimentally to be an artifact in the reconstructed image. Recognition of this artifact may be of help in identifying intracranial lesions, such as shallow subdural hematomas. Attenuation values of isolated cortex and white matter were also studied and do not account for the band of high attenuation. INDEX TERMS:
Computed Tomography, cranial
Radiology 117:71-74, October 1975
• tion to another artifact, hitherto unrecognized, that leads to a universal erroneous description of the cerebral cortex on CCT. In analyzing a CCT image, one recognizes at the periphery of the brain a band of higher attenuation ranging from 20-28 (Fig. 1, A). Its width ranges from a few millimeters to approximately 1.8 cm. In earlier reports, this band was referred to as the cerebral cortex (1, 3). It is best demonstrated by setting the display controls to the smallest window width of 1 EMI unit (M) and a level of 20-25 EMI scale units (Fig. 1, B).
the first report on cranial computed tomography (CCT), there have been several attempts to define the anatomic correlates of the different constituents of the CCT image. However, pitfalls are bound to occur in the early experiences of any new method of imaging. One of the first was the narrow zone of decreased attenuation on the inner side of the skull. It was soon recognized (1) that this "band" was not totally due to the low attenuation values of the subarachnoid space, but in addition, there was an artifact in the algorithm which produced an undershoot due to the large differential in attenuation between the skull and tissue. This feature was most conspicuous in the earlier EMI scanners, which employed an algebraic reconstruction algorithm. This artifact has been reduced to a large degree in the new EMI scanners, which employ a convolution algorithm to reconstruct the image (2). The purpose of this communication is to draw atten-
S
INCE
MATERIAL AND METHODS
(a) Four fresh brains were placed in a water phantom and studied in vitro by CCT using an EMI scanner with the convolution algorithm and 160 X 160 display matrix. Each brain was examined within 6 hours of refrigeration after removal from the body. The CCT scan
Fig. 1. Normal EMI scan. A. Image displayed with a window width of 20 and window level of 13. B. Same image displayed with a window width M (measure) and window level of 22. Note the band of apparent increased attenuation at the periphery of the brain. 1 From the Neuroradiology Section (M. G., Associate Professor and Chief), and Division of Radiation Sciences (M.P.), Mallinckrodt Institute of shan Radiology, Washington University School of Medicine, St. Louis, Mo. Accepted for publication in May 1975.
71
72
MOKHTAR GADO AND MICHAEL PHELPS
October 1975
Fig. 2. EMI scans of brain in vitro. A and B. Brain scanned without skullcap. A. Image displayed with window width of 20 and window level of 12. B. Image displayed with window width M (measure) and window level of 21. There is no demonstrable peripheral band of increased attenuation. C and D. Same brain scanned with overlying skullcap. Note appearance of the peripheral band of increased attenuation.
was repeated after covering the cerebral hemispheres by a skullcap. (b) The same procedure was repeated after fixation of the brain by formalin infusion for seven days. (c) The skullcap used was then scanned by CCT in a water phantom without a brain. (d) A living patient with a large defect in the skull vault was scanned. There were thus eight pairs of in vitro CCT scans, four fresh and four fixed. Each pair consisted of a CCT scan both with and without a skullcap. There was one GeT scan of the skullcap alone in a water phantom and one of a living patient with a skull defect. All GGT scans were done using a 13mm collimator at 120 kVp and 33 rnA. RESULTS
All GCT scans were analyzed for the presence of a
peripheral "band" of increased attenuation. It was found that no such band was visible in the GGT images of brains without the skullcap (Fig. 2). On the other hand, GCT images obtained from brains covered with the skullcap consistently showed the peripheral "band" often referred to as the cerebral cortex on in vivo CCT images. The attenuation values of this band were 20-28 EMI units. Furthermore, the CGT image of the skullcap alone showed an artifactual band on the inner side of the skull, with attenuation values of 7-10 EMI scale units (Fig. 3). DISCUSSION
A composite graphic representation of profiles across approximately corresponding levels of the EMI image of brain, skull, and brain-skull combination is shown in Figure 4. These profiles are from scans corresponding to 7.5 cm above the orbital meatal lines. It is
Vol. 117
73
THE PERIPHERAL ZONE OF INCREASED DENSITY IN eeT
1000
Neuroradiology
PROFILES
500
........
SKULL AND WATER
---
SKULL AND BRAIN
-
BRAIN
200 100
lr
50
ILl
lD ~
:::l
2
20
-
~ ILl
10
5
2
,
"
0
24
36
48
. ............
60
DISTANCE
Fig 3. EMI scan of the skull cap without brain. Note the artifactu.al peripheral band of attenuation to the inner side of the inner table. (Window width M (measure), window level 8.)
seen that the attenuation values at the peripheral zone of the image of the brain-skull combination is the algebraic sum of the values of the artifactual band on the skull-only scan (7-10 EMI units), and the values (15-18 units) shown on the scan of the brain only. Our results demonstrate beyond doubt that the band of increased attenuation seen on in vivo eeT images of adults, and frequently referred to as the cerebral cortex, is an artifact in the reconstructed image. A similar band was also present in the reconstructed image from an EMI scan of a ring phantom of epoxy resin and calcium salts used to simulate the skull (4). One can postulate four explanations for this artifact: (a) The curvature of the skullcap within the 13mm thickness of the slice leads to the inclusion of a progressively smaller thickness of bone in relation to brain in the matrix elements at the transition from bone to brain. This effect increases toward the top of the head due to the increasing curvature of the skull. (b) Spatial averaging due to finite width of the x-ray beam ("-'3 mm). (c) Spatial smoothing (2) is typically employed in the reconstruction algorithm to reduce statistical noise and other artifacts which are present in eeT scanning (e.g., the oscillations which occur close to sharp discontinuities such as the skull-tissue interface, data sampling inaccuracies, etc.) by spatially averaging the values of neighboring image elements (5). However, spatial averaging also has another effect. It turns the sharp change at the skull-tissue interface into a more gradual one resulting in long tails from the skull which run out into the brain with decreasing magnitude with distance away from the skull. This effect artificially increases the values for tissue in close proximity to the skull and to a lesser degree at some distance from the skull. It can be seen in Figure 5 that the largest in-
72
84
96
108
120
132
144
(rn m l
Fig. 4. Composite graph of profiles from numerical printout of in vitro EMI scans of brain, skull, and brain-skull combination in a water phantom. Level of scan corresponds to the level of 7.5 cm above orbital meatal line (note that the ordinate is a log scale).
creases are near the skull, but there is also a small increase ("-'1-2 EMI units) in the values in the center of the brain in the anterior and middle portions of the slice where brain is fully covered by the skullcap. In the posterior portion of the slice where no covering skullcap is included, there is no observed increase. The importance and magnitude of this effect depends upon the size of the difference in the attenuation coefficients of skull and tissue (which is large) and also the degree of spatial smoothing employed in the algorithm. (d) Artifacts due to the variable attenuation of the polychromatic x-ray beam by the skull. This artifact also tends to increase the attenuation values around the interior region of the skull (6, 7). The largest increase is close to the skull, and then diminishes with distance. Further support for the artifactual nature of the peripheral "band of higher attenuation" may be obtained from eeT images of patients who had undergone a large craniectomy. One such case is shown in Figure 6. The band is seen only in the part of the brain underlying the skullcap and is not visualized under the bone defect. The recognition of this artifact may be of help in understanding the problems encountered in peripheral intracranial lesions, such as shallow subdural hematomas (1,8). We have studied the attenuation values of the isolated cortex and white matter in vitro. The results of this work will be the subject of a separate communication. The difference between the attenuation values of the cortex and white matter was found to be on the order of 2-3 EMI units, and does not account for the peripheral band of high attenuation discussed in the present communication.
REFERENCES 1.
Ambrose JAE:
Computerized transverse axial scanning (to-
74
MOKHT AR GADO AND MICHAEL PHELPS
October1975
Fig. 5. EMI scan of brain with skullcap. At this level, the bone covers the anterior and middle portions of the slice. The posterior (occipital) portion has no overlying bone. There is a difference in the attenuation values between the covered and uncovered portions of the slice. A. Image displayed with window width of 20 and window level of 20. B. Image displayed with window width M (measure) and window level of 26. Fig. 6. EMI scan of a patient with a large defect in the skull vault on the right side. A. Image displayed with window width of 20 and window level of 10. B. Same image displayed with window width M (measure) and window level of 18. Note that the peripheral band of apparent increased attenuation is not visualized in the part of the brain underlying the bone defect.
mography). 2. Clinical application. Br J Radiol 46:1023-1047, Dec 1973 2. Shepp LA, Logan BF: The Fourier reconstruction of a head section. IEEE NS-21, 1974, pp 21-42 3. New PFJ, Scott WR, Schnur JA, et al: Computerized axial tomography with the EMI scanner. Radiology 110: 109-123, Jan 1974 4. Rutherford RL, Pullen BR, Goddard MG, et al: Personal communication 5. Phelps ME, Hoffman EJ, Gado M, et al: Computerized transaxial transmission reconstruction tomography. [In] The Past, Present and Future of Non-Invasive Brain Imaging. DeBlanc H, Sorensen JA, eds. New York, Society of Nuclear Medicine, Inc (in press) 6. Shepp LA, Stein JA: Simulated artifacts in computed x-ray
tomography. [In] Workshop on Reconstruction Tomography in Diagnostic Radiology and Nuclear Medicine, San Juan, Puerto Rico, Apr 1975 7. Hounsfield GN: Some practical problems in CT scanning. [In] Workshop on Reconstruction Tomography in Diagnostic Radiology and Nuclear Medicine, San Juan, Puerto Rico, Apr 1975 8. Davis DO, Pressman BD: Computerized tomography of the brain. Radiol Clin North Am 12:297-313, Aug 1974
Mallinckrodt Institute of Radiology Washington University School of Medicine 510 South Kingshighway St. Louis, Mo. 63110
The Peripheral Zone of Increased Density in Cranial Computed Tomography1
•
Neuroradiology
Mokhtar Gado, M.D., and Michael Phelps, Ph.D. The band of increased attenuation seen on in vivo CCT images of adults, frequently designated as the cerebral cortex, has been proved clinically and experimentally to be an artifact in the reconstructed image. Recognition of this artifact may be of help in identifying intracranial lesions, such as shallow subdural hematomas. Attenuation values of isolated cortex and white matter were also studied and do not account for the band of high attenuation. INDEX TERMS:
Computed Tomography, cranial
Radiology 117:71-74, October 1975
• tion to another artifact, hitherto unrecognized, that leads to a universal erroneous description of the cerebral cortex on CCT. In analyzing a CCT image, one recognizes at the periphery of the brain a band of higher attenuation ranging from 20-28 (Fig. 1, A). Its width ranges from a few millimeters to approximately 1.8 cm. In earlier reports, this band was referred to as the cerebral cortex (1, 3). It is best demonstrated by setting the display controls to the smallest window width of 1 EMI unit (M) and a level of 20-25 EMI scale units (Fig. 1, B).
the first report on cranial computed tomography (CCT), there have been several attempts to define the anatomic correlates of the different constituents of the CCT image. However, pitfalls are bound to occur in the early experiences of any new method of imaging. One of the first was the narrow zone of decreased attenuation on the inner side of the skull. It was soon recognized (1) that this "band" was not totally due to the low attenuation values of the subarachnoid space, but in addition, there was an artifact in the algorithm which produced an undershoot due to the large differential in attenuation between the skull and tissue. This feature was most conspicuous in the earlier EMI scanners, which employed an algebraic reconstruction algorithm. This artifact has been reduced to a large degree in the new EMI scanners, which employ a convolution algorithm to reconstruct the image (2). The purpose of this communication is to draw atten-
S
INCE
MATERIAL AND METHODS
(a) Four fresh brains were placed in a water phantom and studied in vitro by CCT using an EMI scanner with the convolution algorithm and 160 X 160 display matrix. Each brain was examined within 6 hours of refrigeration after removal from the body. The CCT scan
Fig. 1. Normal EMI scan. A. Image displayed with a window width of 20 and window level of 13. B. Same image displayed with a window width M (measure) and window level of 22. Note the band of apparent increased attenuation at the periphery of the brain. 1 From the Neuroradiology Section (M. G., Associate Professor and Chief), and Division of Radiation Sciences (M.P.), Mallinckrodt Institute of shan Radiology, Washington University School of Medicine, St. Louis, Mo. Accepted for publication in May 1975.
71
72
MOKHTAR GADO AND MICHAEL PHELPS
October 1975
Fig. 2. EMI scans of brain in vitro. A and B. Brain scanned without skullcap. A. Image displayed with window width of 20 and window level of 12. B. Image displayed with window width M (measure) and window level of 21. There is no demonstrable peripheral band of increased attenuation. C and D. Same brain scanned with overlying skullcap. Note appearance of the peripheral band of increased attenuation.
was repeated after covering the cerebral hemispheres by a skullcap. (b) The same procedure was repeated after fixation of the brain by formalin infusion for seven days. (c) The skullcap used was then scanned by CCT in a water phantom without a brain. (d) A living patient with a large defect in the skull vault was scanned. There were thus eight pairs of in vitro CCT scans, four fresh and four fixed. Each pair consisted of a CCT scan both with and without a skullcap. There was one GeT scan of the skullcap alone in a water phantom and one of a living patient with a skull defect. All GGT scans were done using a 13mm collimator at 120 kVp and 33 rnA. RESULTS
All GCT scans were analyzed for the presence of a
peripheral "band" of increased attenuation. It was found that no such band was visible in the GGT images of brains without the skullcap (Fig. 2). On the other hand, GCT images obtained from brains covered with the skullcap consistently showed the peripheral "band" often referred to as the cerebral cortex on in vivo CCT images. The attenuation values of this band were 20-28 EMI units. Furthermore, the CGT image of the skullcap alone showed an artifactual band on the inner side of the skull, with attenuation values of 7-10 EMI scale units (Fig. 3). DISCUSSION
A composite graphic representation of profiles across approximately corresponding levels of the EMI image of brain, skull, and brain-skull combination is shown in Figure 4. These profiles are from scans corresponding to 7.5 cm above the orbital meatal lines. It is
Vol. 117
73
THE PERIPHERAL ZONE OF INCREASED DENSITY IN eeT
1000
Neuroradiology
PROFILES
500
........
SKULL AND WATER
---
SKULL AND BRAIN
-
BRAIN
200 100
lr
50
ILl
lD ~
:::l
2
20
-
~ ILl
10
5
2
,
"
0
24
36
48
. ............
60
DISTANCE
Fig 3. EMI scan of the skull cap without brain. Note the artifactu.al peripheral band of attenuation to the inner side of the inner table. (Window width M (measure), window level 8.)
seen that the attenuation values at the peripheral zone of the image of the brain-skull combination is the algebraic sum of the values of the artifactual band on the skull-only scan (7-10 EMI units), and the values (15-18 units) shown on the scan of the brain only. Our results demonstrate beyond doubt that the band of increased attenuation seen on in vivo eeT images of adults, and frequently referred to as the cerebral cortex, is an artifact in the reconstructed image. A similar band was also present in the reconstructed image from an EMI scan of a ring phantom of epoxy resin and calcium salts used to simulate the skull (4). One can postulate four explanations for this artifact: (a) The curvature of the skullcap within the 13mm thickness of the slice leads to the inclusion of a progressively smaller thickness of bone in relation to brain in the matrix elements at the transition from bone to brain. This effect increases toward the top of the head due to the increasing curvature of the skull. (b) Spatial averaging due to finite width of the x-ray beam ("-'3 mm). (c) Spatial smoothing (2) is typically employed in the reconstruction algorithm to reduce statistical noise and other artifacts which are present in eeT scanning (e.g., the oscillations which occur close to sharp discontinuities such as the skull-tissue interface, data sampling inaccuracies, etc.) by spatially averaging the values of neighboring image elements (5). However, spatial averaging also has another effect. It turns the sharp change at the skull-tissue interface into a more gradual one resulting in long tails from the skull which run out into the brain with decreasing magnitude with distance away from the skull. This effect artificially increases the values for tissue in close proximity to the skull and to a lesser degree at some distance from the skull. It can be seen in Figure 5 that the largest in-
72
84
96
108
120
132
144
(rn m l
Fig. 4. Composite graph of profiles from numerical printout of in vitro EMI scans of brain, skull, and brain-skull combination in a water phantom. Level of scan corresponds to the level of 7.5 cm above orbital meatal line (note that the ordinate is a log scale).
creases are near the skull, but there is also a small increase ("-'1-2 EMI units) in the values in the center of the brain in the anterior and middle portions of the slice where brain is fully covered by the skullcap. In the posterior portion of the slice where no covering skullcap is included, there is no observed increase. The importance and magnitude of this effect depends upon the size of the difference in the attenuation coefficients of skull and tissue (which is large) and also the degree of spatial smoothing employed in the algorithm. (d) Artifacts due to the variable attenuation of the polychromatic x-ray beam by the skull. This artifact also tends to increase the attenuation values around the interior region of the skull (6, 7). The largest increase is close to the skull, and then diminishes with distance. Further support for the artifactual nature of the peripheral "band of higher attenuation" may be obtained from eeT images of patients who had undergone a large craniectomy. One such case is shown in Figure 6. The band is seen only in the part of the brain underlying the skullcap and is not visualized under the bone defect. The recognition of this artifact may be of help in understanding the problems encountered in peripheral intracranial lesions, such as shallow subdural hematomas (1,8). We have studied the attenuation values of the isolated cortex and white matter in vitro. The results of this work will be the subject of a separate communication. The difference between the attenuation values of the cortex and white matter was found to be on the order of 2-3 EMI units, and does not account for the peripheral band of high attenuation discussed in the present communication.
REFERENCES 1.
Ambrose JAE:
Computerized transverse axial scanning (to-
74
MOKHT AR GADO AND MICHAEL PHELPS
October1975
Fig. 5. EMI scan of brain with skullcap. At this level, the bone covers the anterior and middle portions of the slice. The posterior (occipital) portion has no overlying bone. There is a difference in the attenuation values between the covered and uncovered portions of the slice. A. Image displayed with window width of 20 and window level of 20. B. Image displayed with window width M (measure) and window level of 26. Fig. 6. EMI scan of a patient with a large defect in the skull vault on the right side. A. Image displayed with window width of 20 and window level of 10. B. Same image displayed with window width M (measure) and window level of 18. Note that the peripheral band of apparent increased attenuation is not visualized in the part of the brain underlying the bone defect.
mography). 2. Clinical application. Br J Radiol 46:1023-1047, Dec 1973 2. Shepp LA, Logan BF: The Fourier reconstruction of a head section. IEEE NS-21, 1974, pp 21-42 3. New PFJ, Scott WR, Schnur JA, et al: Computerized axial tomography with the EMI scanner. Radiology 110: 109-123, Jan 1974 4. Rutherford RL, Pullen BR, Goddard MG, et al: Personal communication 5. Phelps ME, Hoffman EJ, Gado M, et al: Computerized transaxial transmission reconstruction tomography. [In] The Past, Present and Future of Non-Invasive Brain Imaging. DeBlanc H, Sorensen JA, eds. New York, Society of Nuclear Medicine, Inc (in press) 6. Shepp LA, Stein JA: Simulated artifacts in computed x-ray
tomography. [In] Workshop on Reconstruction Tomography in Diagnostic Radiology and Nuclear Medicine, San Juan, Puerto Rico, Apr 1975 7. Hounsfield GN: Some practical problems in CT scanning. [In] Workshop on Reconstruction Tomography in Diagnostic Radiology and Nuclear Medicine, San Juan, Puerto Rico, Apr 1975 8. Davis DO, Pressman BD: Computerized tomography of the brain. Radiol Clin North Am 12:297-313, Aug 1974
Mallinckrodt Institute of Radiology Washington University School of Medicine 510 South Kingshighway St. Louis, Mo. 63110
