Neuroradiology (1984) 26:363-367

Springer-Verlag 1984

Neuroradinlogv

Extradural hematomas: 'Measurement of size by volume summation on CT scanning

i

O. F. Petersen a and J. O. Espersen 2

IDepartments of INeuroradiology and 2Neurosurgery, The University Hospital, Aarhus, Denmark

The method of volume summation (V= T(A1 + A2 ... An) was used to measure the size of extradural hematomas. The accuracy was tested on six different artificial silicone hematomas and the mean difference w a s - 2 . 7 ml, SD 3.7 ml. The reproducibility was tested on CT scans of clinical hematomas, SD was 2.1 ml. An empirical formula for volume estimation then found: 0.5xheightx length x depth was moderately reliable, while midline shift and "vesselfree space" were poor indicators of size. In conclusion, the volume summation with manual outlining was found to be highly accurate, but the problems of CT smoothing, spectral shift artifact, partial volume effect and separation of the hematoma from other structures must be considered.
Summary.

Materials

Key words: Extradural hematomas - volume measurement - volume summation - CT scanning

To test the reliability of the volume summation method a phantom was made compriming six hematomas of different sizes modelled in silicone (Bostik fugemasse ) inside a skull in different regions. To achieve comparability in attenuation differences between real hematomas, skull and brain, the phantom was placed in a plastic bag filled with 1.25% Conray 400 in water. The attenuation values of the "hematomas" and "brain" were 136 H U and 111 HU, respectively (Fig. 1). To compare the method to other ways of estimating size, 54 consecutive patients with completed preoperative CT scans were used. Four CT examinations were performed on a Siemens Somatom 2 N in a referring county hospital (Holstebro). All other examinations were carded out on an Ohio Nuclear Delta Scan 25, scan time 1.5 rain per double slice. In all patients the CT scans were performed with a 15 caudal tilt to the meato-orbital

The introduction of CT imaging has provided far more reliable diagnostic accuracy of intracranial hematomas [1]. The CT scanning probably never fails in the acute stage of bleeding if the haematoma is large enough for surgery [2]. Erichson and Hhkonsson [3] measured the size of extradural hematomas based on a mathematical formula applied to CT. Most work concerning these hematomas is based on surgical measurement or estimation from angiograms [4], often based only on "vesselfree space" [5]. The purpose of this study was to describe and evaluate the method of volume summation in extradural hematomas in order to use CT in the Study of their pathophysiology in man and to test the clinical significance of their size.

Fig.t. CT scanning of a phantom with 2 "extradural hematomas". The attenuationvalues are 136 HU in the outlined "hematoma" (I) and 111 HU in the "brain" square (II) Fig. 2. Outlining of an extradural hemat0ma. Inverted grey scale and special settings to visualize the complete outline

364 Table I. The accuracy of volume summation in six artificial extradural hematomas Measured at CT scannings Slice thickness mm Tilta Region Left frontal/vertical Left parietal/vertical Right parietal Right temporal Left frontal Left occipital Average difference 8 8 - 1 0 0 ml 0.0 11.7 36.6 61.1 73.2 108.9 - 1.6 ml 4.0 12.4 35.0 63.6 74.1 109.4 Measured by water 8 13 displace+15 +15 ment ml 0.0 0.0 31.1 60.4 71.8 99.4 ml ml

0.0 5.5 0.0 9.5 31.4 36.0 5 2 . 7 65.5 73.8 76.5 98.8 105.5 - 6,9

- 0.02 - 5.1

Mean difference + SD a Related to meato-orbital plane

1St, m e a s u r e m e r l t ml

- 213 _ 3.7

- 6,9 3.6

200

'

To reduce the artifact of partial volume effect in slices with marked gradual attenuation shifts, the marking chosen lay in the middle of the border zone (see Fig. 8). For comparison the volumes of the silicone "hematomas" were measured by water displacement. All distance measurements were obtained electronically at the display and calculated directly on the Evaluskop. Midline shift was measured at that part of the septum pellucidum which appeared to have the most pronounced displacement. The maximal depth (or "vesselfree space") of the hematoma was measured from bone to brain over the widest part of the hematoma - in some basal hematomas on reformatted coronal sections. The factors of the empirical formula of volume (0.5 height x depth x length) were based on distance measurements of depth and length on the C T slice having the largest area of extradural hematomal The height was calculated by multiplying the number of slices in which the hematoma could be seen by the slice thickness.

Statistics
100

To test trends, a linear regression model was used. Comparing reproducibility and accuracy the sign t-test was used. The standard deviation was calculated when appropriate.

100
2nd m e a s u r e m e n t

200

ml

Results

Fig.3. The reproducibility of volume summation. SD: 2.1 ml

Accuracy and precision of the method

The accuracy of the method appears from Table 1, in which a trend to underestimate the size is shown, but in practice the difference is negligeable. It appears that the plane of the CT scans in relation to the placement of the hematoma plays an important role, as well as the size of hematoma and the slice thickness. The reproducibility of the volume measurement appears from Figure 3. Six hematomas of different size were selected by JOE among 39 hematomas measured at least half a year before and measured again by OFP. The maximal difference was 3 ml, SD 2.1 ml, which indicates, that the method is very precise. There was a greater difference, as expected, between the CT determined and the surgically removed volume of coagulated hematomas [6] (Fig.4). The correlation coefficient, however, was 0.91. In bleeding hematomas [6] the difference between the surgically removed volume and the CT measured volume

plane and a slice thickness of 8 mm: matrix 256 256 (pixel size 1 x 1 x 8 mm). All CT examinations were analysed on a Siemens Evaluskop.

Methods The volume of the hematomas (clinical as well as artificial) was measured by volume summation: V = T (AI+A2 ... An), where V = v o l u m e (ml), T=slice thickness (actually 0.8 cm), AI~ A2, . . . An = areas of the h e m a t 0 ~ (9m2). The area was measured by the region-of-interest function with manual outlining of the hematoma (Fig. 2). When outlining near bone the center- and window width settings were fixed to 30 and 1000 HU, respectively, and when outlining near brain 30 and 100 HU. The center setting was raised to 80 HU, when outlining the silicone hematomas.

365
Measured during surgery ml 2OO Midline shift rnm
2O

100 ,

0
0 100 2 0 0 ml

-10
0

100

200 ml

Volume

Volume

Fig.4. The relation between extradural clots removed at surgery and the volume measured at CT. The regression line y ml = (0.84___ 0.10)x + 24.5 ___5.t ml, correlation coefficient 0.91. Time from CT scan to removal of clot: Range 0.5-94h, median 1.5 h

Fig.7. The midline shift related to volume (midline and f0ssa posterior hematomas are excluded). The regression line: yml=(0.11 +0.02)x+(2.8+1.5)ml. Correlation coefficient 0.66. P < 0.001

0.5x(Height x Length x Depth) on CT ml 200

//
2 0 0 ml

showed a higher variance (correlation coefficient

0.78).
Comparison to other methods of size-estimation
The empirical formula 0.5 x height x depth x length which is easy and fast to use. The first mentioned method - seems to be almost as precise as the method of volume-summation (Fig. 5). Comparison of the CT measured volume to the maximal "vesselfree space" shows that a relation exists, but that the variance could be large in some cases, especially with different curvatives of the skull (Fig.6). Similar comparison of the CT measured volume to the midline shift showed that a relation existed, but the variance was large, too, due to the influence o f accompanying brain lesions and the location of hematoma (Fig. 7).

10/ 0 0
0 100

Volume summation

Fig.5. The volume calculated from the empirical formula compared to the volume measured by volume summation in 54 patients. The regression line y m l = ( 1 . 0 5 _ 0 . 0 2 ) x - ( 4 . 9 _ 2 . 1 ) m l . Correlation coefficient 0.98

"Vesselfree space" on CT
mm 5O

Discussion To measure the size Of an intracranial structure or lesion on CT scans more methods are available. The linear measurements used by earlier investigators [7, 8] are very reproducible and have previously been used in estimation of the size of extradural hematomas [3]. In this study the value of this quick method is underlined. To achieve the most precise evaluation of the size, either statistical method [9-11] or volume summation [12] must be used. To achieve an observer independent measurement a statistical method should

o ~

0 0 100 200 ml

Volume

Fig. 6. The maximal "vessel free space" on CT compared to the volume. Regression line: yml=(0.014_+0.001)x+(1.5+0.1)ml. Correlation coefficient 0.78. P < 0.001

366
=/g deviation

30

20

10

I
w Q

50

100 Volume ml

150

200

Fig.& Partial volume effect. The levels of slices I and H on a schematic transverse section. The dotted line on the CT scans I and H shows the measuring points. The profiles show the stepwise increase in attenuation values when moving from brain to hematoma. It is seen that in s l i c e / t h e shift from 37 to 71 H U occurs over 4 pixels, but in slice H an identical shift is found Over 14 pixels

Fig.9. Relative deviations of volume. The points indicate the relative deviation of volume of the single hematoma, if one pixel is added along the surface calculated f r o m t h e formula: AV= 0.5 (height x (length + 2 mm) x (depth + 2 mm)) - V. 1 shows the best regression line for these points. H shows one standard deviation related to volume found in this study

be preferred, but in practice these methods are impossible to perform for most extradural hematomas. 1. The spectral shift artifact [13, 14]; 2. "Smoothing" [6], (which is both particularly pronounced near bone) 3. The partial volume effect (Fig. 8), (which is pronounced both near bone and brain [15]) 4. Separation of an extradural hematoma from another hemorrhagic lesion in the region, which in many cases would require manual outlining; 5. Different attenuation values within the hematoma according to ongoing bleeding [6, 16-18]. We were led t o perform the measurements by manual outlining and summation, because of the above mentioned problems. In this method only partial volume effect occurring in the most basal and vertical slices of the hematoma was pronounced. 'This error is minimal in large hematomas, but could ~be important in small ones, and might explain that deviations in two measurements (Fig. 3) are numerically equal and independent of the size. Observer error in measuring is reduced by the introduction of electronic devices [19] and avoiding electronic and viewing pitfalls [15, 20] the method u s e d is very accurate. The precision and accuracy of the volume summation in artificial hematomas were very near the ideal as found also by Breiman et al. [12] applying t h e method to both organs and phantoms.

The calculated volume was in all cases lower than real volume, but most pronouncedly in 13 mm slices, as predicted in a simulation study of Baxter and Sorenson [15]. The standard deviation was equal in all sizes of hematoma, which means that the relative standard deviation is dependent on the size of the lesion (Fig. 9). As shown in Figure 9 the addition of only one pixel along the entire surface would increase the volu m e by from 8% in hematomas of 200ml to 15% in hematomas of 30 ml and much more in smaller hematomas. As a consequence of the increasing relative inexactitude in size measurements - the combination of zero deviation and standard deviation - a CT scanh e r has a lower limit in the detection of extradural hem~tomas. Based on the results our CT scanner would not be able "to see" 5% and 20% of extradural hematomas with a size of 10 ml when, respectively, 8 mm and 13 mm slices were used. In this context it m u s t be emphasized that small hematomas placed perpendicularly are Visualized better than hematomas placed parallell to the CT plane. Applying the method to clinical hematomas a islight inexactitude was found in relation to surgically removed clots. The intercept of 25 ml was probably due to a slight mixing with circulating blood during surgery - a problem which lead Habash et al. to divide their material into only large, medium and small t extradural hematomas [4].
i

367

In our clinical material of extradural hematomas : consisting of 56 patients only two had to be excluded: one, because the hematoma was at the vertex and another, because the hematoma could not be separated from an accompanying subdural one. These are the main limitations in applying th e method, the vertically placed hematoma, in which measuring could be very inexact due to pronounced partial volume effect in several slices and other hematomas closely adjacent, which do not allow a clear distinction. Mathematically the simple formula used (0.5 x height x depth x length) differs by only 5% from the formula of a rotational ellipsoid proposed by Erichson and Hhkonsson (% x Jr x height x depth x length [8]) [3]. The volume of the extremely spherical hematoma would in our formula be 0.5 x 2R 2R = 4R 3 or approximately the volume of a sphere. At the other extreme, the cube, the formula would; give results that could be out by a factor of two, but that extreme does not exist within the curved walls of the skull. The deviation from the ideal line was less than 5% in 2~ of the 54clinical hematomas and did not exceed 30%, which shows the reliability of the method. The midline shift and the "vesselfree space" seems, however, to be relatively imprecise, as recognized by others [4] and this is obviously due to the many different regions in which the hematomas are localized as well as accompanying lesions, which appear to be very common in extradural hematomas [3, 21] and much more common than previously presumed.

References
1. French BN, Dublin AB (1977) The value of computerized tomography in the management of 1000 consecutive head injuries. Surg Neurol 7:171-183 2. Espersen JO, Petersen OF (1981) Computerized tomography in patients with head injuries. Relation between CT scans and clinical findings in 96 patients. Acta Neurochir 56:201-217 3. Erichson K, H~tkonsson S (1981) Computed tomography of epidural hematomas. Association with intracranial lesions and clinical correlation. Acta Radiol (Stockh) 22: 513-519 4. Habash AH, Sortland O, Zwetnow NN (1982) Epidural hematoma. Pathophysiological significance of extravasation and arteriovenous shunting. Acta Neurochir 60:7-27 5. Mendelow AD, Karmi MZ, Poul KS, Fuller GAG, Gillingham

FJ (1979) Extradural haematoma: effect of delayed treatment. Br Med J 1240-1242 6. Petersen OF, Espersen JO (1984) How to distinguish between bleeding and coagulated hematomas on the plain CT scanning. Neuroradiology 26 (in press) 7. Gyldensted C (1977) Measurements of the normal ventricular system and hemispheric sulci of 100 adults with computed tomography. Neuroradiology 14:183-192 8. Haug C (1977) Age and sex dependence of the size of normal ventricles on CT. Neuroradiology 14:201-204 9. Pentlow KS, Rottenberg DA, Deck MDF (1978) Partial volume summation: a simple approach to ventricular volume determination from CT. Neuroradiology 16:130-138 10. Gado M, Huges CP, Danziger N, Chi D, Jost G, Berg L (t982) Volumetric measurements of the cerebrospinal fluid spaces in demented subjects and controls. Radiology 144:535-538 11. Zatz LM, Jerigan TL, Alumada AJ (1982) Intracranial fluid volume. AJNR 3: 1-11 12. Breimann RS, Beck JN, Korobkin M, Glenny R, Akwari OE, Heaston DK, Moore AV, Ram PC (1982) Volume determinations using computed tomography. AJR 138:329-333 13. Zatz LM, Alvarez RE (1977) An inaccuracy in computed tomography: the energy dependence of CT values. Radiology 124:91-97 14. Di Chiro G, Brooks RA, Dubol L, Chwe E (1978) The apical artifact: elevated attenuation values towards the apex of the skull. J Comput Assist Tomogr 2: 65-70 15. Baxter BS, Sorenson JA (1981) Factors affecting the measurement of size and CT number in computed tomography. Invest Radiot 16:337-341 16. Norman D, Price D, Boyd D, Fishman R, Newton TH (1977) Quantitative aspects of computed tomography of the blood and cerebrospinal fluid. Radiology 123:335-338 17. New PEJ, Aronow S (1976) Attenuation measurements of whole blood and bloodfractions in computed tomography. Radiology 121: 635-640 18. Zimmerman RA, Bilanuik LT (1982) Computed tomographic staging of traumatic epidural bleeding. Radiology 144: 809-812 19. Sabattini Z (1982) Evaluation and measurement of the normal ventricular and subarachnoidal spaces by CT. Neuroradiology 23: 1-5 20. Koehler PR, Anderson RE, Baxter BS (1979) The effect of computed tomography viewer controls on anatomical measurements. Radiology 130:189-194 21. Cordob~s F, Lobato RD, Rivas JJ, Mufioz MJ, Chillou D, Portillo JM, Lamas E (1981) Observations on 82 patients with extradural hematoma. J Ne~rosurg 54:179-186 Received: 27 June 1983 in revised form: 7 November 1983 Dr. J.O. Espersen Department of Neurosurgery Aarhus Kommunehospital DK-8000 Aarhus Denmark