Upsala J Med Sci 92: 277-286, 1987 Object Size Determination at Computed Tomography Anders Magnusson From the Department of Diagnostic Radiology, Uppsala University, Uppsala, Sweden A B S TRACT To investigate whether large differences in attenuation between an object and its background at CT examination alter the apparent size of the object on the image compared with the conditions with a small attenuation difference, phantom experiments were performed. A t the same time the influence of large attenuation differences on the geometric resolution was examined. The results show that neither the size reproduction of the object nor the geometric resolu- tion is altered by variations in attenuation differences. Especially with large differences in attenuation between the object and background, however, the window centre setting was of greatest importance for correct reproduction of the object. INTRODUCTION In computed tomography (CT) s h a r p borderlines between s t r u c t u r e s of different density a r e reproduced a s blurred contours. The degree of blurring depends, among other things, on the X-ray beam width, the dimensions and properties of the detectors, and the reconstruction algorithms employed. This blurring is of importance when measuring the size of s t r u c t u r e s in the image (1,3). When an object which differs markedly in attenuation from its background is examined by CT, its apparent size varies with the viewing conditions. A decisive factor in this context is the question of which p a r t of t h e CT-cal- culated attenuation profile is utilized to produce the image. This is determined by the setting of the window centre (WC) and window width ( W W ) (Fig. 1). When patients have undergone CT examination both before and after bipe- dal lymphography, the lymph nodes have appeared larger after the contrast filling, when the image has been viewed with the window setting normally used for evaluation of the retroperitoneal space. This enlargement has considerably exceeded that caused by the contrast filling in itself ( 2 , 4 ) . It is not clear whether the apparent enlargement is solely an effect of the W W and WC settings or whether it is due to an e r r o r of measurement or of 277 Image Fig.1. The i n f l u e n c e o f d i f f e r e n t window c e n t r e (WC) s e t t i n g s on t h e CT r e p r o d u c t i o n o f an o b j e c t when t h e window w i d t h ( W W ) i s c o n s i d e r a b l y s m a l l e r t h a n t h e a t t e n u a t i o n d i f f e r e n c e i n t h e image. a ) WC a t t h e l e v e l o f t h e background a t t e n u a t i o n and b ) WC a t t h e l e v e l o f t h e a t t e n u a t i o n o f t h e o b j e c t . Note t h a t i n t h i s image t h e grey s c a l e i s reversed, so t h a t h i g h a t t e n u a t i o n values are represented by b l a c k and low ones by w h i t e . I W C w w I Image reconstruction. An investigation was therefore undertaken to elucidate the way in which the interpretation of the CT image is influenced by the difference in attenuation between an object and its background, different reconstruction algorithms and different window settings. In addition, the geometric resolution with various attenuation differences and window settings was examined. MATERIAL A N D METHODS A Siemens Somatom D R 2 whole-body scanner with a 2 5 6 x 2 5 6 matrix was used for the investigation. The exposure data were 1 2 0 kV, 0 . 5 1 8 As, exposure time 4.5 s , 480 projections and a slice thickness of 8 m m . Reconstructions were performed with a zoom factor of 1 . 7 , corresponding t o a normal abdominal examination, giving a pixel s i z e of 1 . 2 7 mm, and a zoom factor of 6 , with a pixel size of 0.36 m m . For the reconstructions both the standard algorithm (Kernel 11, which is normally used for abdominal examinations, and an algorithm 278 with strong edge enhancement (Kernel 7) were employed. The images were analysed on the standard diagnostic console with different window settings (scale: air -1000 Hounsfield units (HU), water 0 H U ) . The term image is defined in this article as the image reproduced on the monitor screen. Areas of interest were magnified; measurements were performed directly on the image screen and on the attenuation profiles, with the cursor incorporated in the console. For investigation of the influence of the attenuation difference on the apparent size of the object, a polystyrene phantom with a diameter of 20 cm was used. In i t s centre there was a cylindrical hole 2 3 m m in diameter. The atte- nuation of the phantom was -10 H U . The hole was filled with Gastrografina ( 3 7 0 mg I/ml) in dilution of 0 . 7 5 and 2 5 % , which gave the hole attenuation values of 75 and 2050 H U respectively. The phantom was placed in the centre of the gantry and the cylindrical hole passed perpendicularly through the entire CT slice. The width of the attenuation profile was measured at l o % , 50% and 90% of the profile height. To examine the effect of different attenuation differences and window settings on the geometric resolution, a phantom of parallel plexiglass rods with a square 5 x 5 mm cross-section was used. The distance between the rods was 1 m m . The plexiglass rods were examined both in air and when surrounded by water. The attenuation value for plexiglass was 120 H U , which means that the difference in attenuation between plexiglass and air was 1120 H U and between plexiglass and water 120 H U . The CT slice was perpendicular to the longitudinal axis of the rods. RESULTS The attenuation profiles for the cylindrical hole when it was filled with Gastrografin of different concentrations, are given in Fig. 2 . When the stan- dard algorithm was used for reconstruction, the width of the profile was inde- pendent of the difference in attenuation between the object and background (Table 1 ) . Neither did the zoom factor used in the reconstruction influence the apparent dimensions. Reconstruction with the edge-enhancing algorithm result- ed in a decrease in profile width by a maximum of 1 mm on examination of the high-density object. The impact of different WC settings on the apparent size of the object in the image when the object-background difference in attenuation was large (2060 H U ) , is illustrated in Figs 3 and 4 . When measured on an image of the object, reconstructed with a standard algorithm and a zoom factor of 6 at different WC settings (from 0 to 2300 H U ) and at a constant W W ( 4 0 0 HU), the diameter varied from 26 to 2 0 mm. The t r u e diameter ( 2 3 m m ) was reproduced at a WC setting of between 1000 and 1200 H U , thus at about half the attenuation differ- ence between the object and background. 279 100 76 52 28 HU 2400r - 100- - 76 - - 52 - - 28 - HU C d Fig.2. ( c and d), r e c o n s t r u c t e d w i t h s t a n d a r d ( a and c ) and edge-enhancing a l g o r i t h m s ( b and d ) . A t t e n u a t i o n p r o f i l e s f o r t h e 23 mrn o b j e c t w i t h a t t e n u a t i o n s o f 75 ( a and b ) and 2050 HU Table 1 The w i d t h o f t h e a t t e n u a t i o n p r o f i l e i n mm a t 10% ( d ), 50% ( d ) and 90% ( d ) o f i t s h e i g h t . 1 2 3 Attenuation Zoom 1 . 7 Zoom 6 difference Standard Standard Edge en- object- algorithm algorithm hancement background dl d 2 d 3 dl d 2 d 3 dl d2 d3 8 5 H U 25 2 3 21 2 5 2 3 21 2 5 2 3 2 1 2060 H U 25 2 3 21 2 5 2 3 2 1 2 4 2 3 21 280 Fig.3. s e t t i n g s and a c o n s t a n t window w i d t h (400 HU). The CT image o f an o b j e c t o f h i g h a t t e n u a t i o n (2050 HU) a t d i f f e r e n t window c e n t r e (WC) WC = 0 ( a ) , 1200 ( b ) and 2000 HU ( c ) . 26 - 24 - 22 20 - - > I I I I Fig.4. E f f e c t o f window c e n t r e on t h e o b j e c t diameter. The window w i d t h i s s e t a t 400 HU. The t r u e diameter (23 mm) i s reproduced a t a window c e n t r e o f between 1000 and 1200 HU (shaded column). Fig.5. The 23 nun o b j e c t o f h i g h a t t e n u a - t i o n (2050 HU) reproduced w i t h a window c e n t r e o f 1000 HU and a window w i d t h o f 4000 HU. A p p a r e n t diameter (mm) \. 24".-. . .......... '. \ W W 400 Fig.6. The i n f l u e n c e o f t h e window c e n t r e and window w i d t h ( W W ) s e t t i n g s on 2 0 t h e reproduced diameter o f an o b j e c t w i t h an a t t e n u a t i o n o f 75 HU. 2 2 '. "-.. w w 200 W W 30 W W 73 z 0 100 200 0 1 1 I I Window c e n t r e (HU) 18-878573 28 1 HU 130r a 8 - 20L 0 HU 180 [ C -201 0 8 16 m" d HU 180r Fig.7. A t t e n u a t i o n p r o f i l e s f o r two p l e x i g l a s s rods s i t u a t e d 1 mm from each o t h e r , examined i n water ( a and b ) and a i r ( c and d ) , w i t h t h e use o f standard ( a and c ) and edge-enhancing ( b and d ) a l g o r i t h m s f o r r e c o n s t r u c t i o n . When the high-density object was viewed with a large WW, so that the entire attenuation difference between the object and background was covered by W W , the border of the object against the background became diffuse and it was difficult to define any exact measurement points for the diameter in the image (Fig. 5 ) . The WC setting also influenced the reproduction of an object which differ- ed relatively little in attenuation from its background (by 85 H U ) , but only when W W was narrow (Fig. 6). When W W was smaller than the object-back- ground attenuation difference, the apparent size of the object varied with different settings of WC, and this size variation increased with decreasing W W . When, on the other hand, WW was larger than the attenuation difference, the size was not affected by changes in WC as long as the entire attenuation profile lay within the W W limits. On examination of the plexiglass phantom with parallel rods with a square cross-section, in air and in water, parts of the attenuation profiles of the two objects coincided under all tested conditions (Fig. 7 ) . The lowest attenuation value for the "space" between the rods, when the images were reconstructed 282 F i g . 8 . The image o f t h e two c l o s e l y a d j a c e n t p l e x i g l a s s rods surrounded by a i r , a t a constant window width (400 HU) and w i t h d i f f e r e n t window c e n t r e s : -1000 ( a ) , -850 ( b ) , -700 ( c ) , -550 ( d ) , -400 ( e ) and -250 HU ( f ) . with a standard algorithm, was found to be higher than t h e attenuation value of the background by 30% of the object-background attenuation difference (Fig. 7 a and c ) . This finding was independent of the magnitude of the attenuation difference and of the zoom factors employed. On examination both in air and in water, too low values ( 2 0 and 110 H U ) were obtained for the maxima of the profiles, t h u s 1 0 0 and 1 0 H U , respectively, below the t r u e attenuation value for plexiglass ( 1 2 0 H U ) . On reconstruction with an edge-enhancing algorithm, the attenuation values for the "space" were lower than when a standard algorithm was used; t h u s it was now higher than the background value by only 15% of the object-background attenuation difference, both when the examination was performed in air and in water (Fig. 7 b and d ) . The position of WC was of importance for the geometric resolution, espe- cially on reconstruction with a standard algorithm. When the plexiglass r o d s , surrounded b y a i r , were viewed with a WC of -1000 HU and a W W of 400 H U , they appeared to form one figure (Fig. 8 a ) ; when WC was increased stepwise while W W was kept constant, the rods w e r e reproduced a s one unit (Fig. 8 b) until the upper limit of W W exceeded the lowest attenuation value for the "spa- 283 ce" between the rods. Thereafter the space began to appear in grey tones. The space increased in width with increasing WC up to about -500 HU, and parflllel with the increase in width of the space the size of the rods decreased (Fig. 8 c and d ) . When WC was so high that the entire W W lay above the lowest attenua- tion value for the space, this appeared black (Fig. 8 e ) . The same effect was observed when the image of the plexiglass rods surrounded b y water was viewed with varying WC. As long as WC lay within the values for the attenua- tion profile, 0-110 H U (Fig. 7 a ) , however, W W had to be relatively small - no greater than 60 H U at a WC of 0 HU, for the two rods to converge. However, with the window settings that are normally used at abdominal examinations, 400 H U for W W and 100 H U for WC, the rods appeared clearly separated. DISCUSSION Variation of the attenuation difference between the object and background did not alter the shape of the attenuation profile. Similarly, the width of the profile was influenced to only a minor extent b y the use of different algorithms in the reconstruction. The setting of WC, on the other hand, was of great importance for the reproduced size of the object. Thus the apparent diameter of the object of high attenuation ( 2 0 5 0 H U ) varied by 6 m m when the image was viewed with different WC settings. Baxter & Sorensen (1) found that when WC was set at half the attenuation difference between the object and background, the size of the object was reproduced correctly, a finding in accordance with the present results (Fig. 4 ) . When WC is set at the level of the background attenuation on examination of an object of high attenuation, the apparent size of the object is greater than its real size, and its greatest part appears completely white, as this part of the attenuation profile lies above the upper l i m i t of W W (Fig. 1 a ) . This occurs in clinical examinations when contrast-filled retro- peritoneal lymph nodes a r e being evaluated with settings of WC (100 H U ) and W W ( 4 0 0 H U ) that are normal for abdominal examinations, on account of the fact that the attenuation value of the contrast-filled lymph nodes (1000-2000 H U ) is considerably higher than the upper limit of W W . If, on the other hand, WC is chosen according to the attenuation of the contrast-filled lymph nodes, with W W unchanged, only the maximum of the attenuation profile will give an image, which will mean that the apparent size will be smaller than the real one, at the same time as the image will appear i n grey tones (Fig. 1 b ) . For t h e object of low attenuation, the W W setting also influenced the appa- rent s i z e in those cases where W W was smaller than the object-background attenuation difference (Fig. 6 ) . With decreasing W W the object became more and more distinctly outlined in the image, but in parallel with this the size repro- duction became more sensitive to changes in WC. When closely adjacent objects of high attenuation were examined and WC was 284 F i o . 9 . CT image of c o n t r a s t - f i l l e d lymph nodes. a ) Window s e t t i n g normal f o r abdominal examinations. Window c e n t r e 70 HU, window w i d t h 400 HU. b ) Window c e n t r e 1000 HU, window w i d t h unchanged. set at a level with the attenuation of the background, at the same time as the selected W W (400 H U ) was considerably smaller than the object-background attenuation difference (1110 HU), the geometric resolution was poor. On CT examination of closely adjacent contrast-filled lymph nodes, these nodes, with WC and W W settings normal for abdominal examinations ( 1 0 0 and 400 HU respec- tively), may converge and be interpreted as one large single node (Fig. 9 ) . On reconstruction with the edge-enhancing algorithm, t h e geometric resolution was improved at low WC values. This means that in clinical examinations of s t r u c t u r e s differing largely in attenuation, an edge-enhancing algorithm should be used. When WC was set at a level corresponding to half the attenuation difference, good resolution was obtained. The maximal attenuation values for the plexiglass r o d s , when examined in air and water and with use of a standard algorithm for the reconstruction, lay 100 and 10 HU, respectively, below the real attenuation value for plexiglass. This is explained by the fact that the spatial resolution for the algorithm used is insufficient for a correct repro- duction of the attenuation values of such small objects. CONCLUSIONS The difference i n attenuation between an object and i t s background has v e r y little influence on the reproduced size of the object and on the geo- metric resolution with the CT scanner used for this investigation. For correct reproduction of the size of an object and for good geometric resolution in the monitor image, t h e WC setting is of the greatest import- ance. A correct reproduction of the object of interest was achieved when WC was set at a value in the middle of the attenuation values of the object cand t h e background. 285 3. With a W W that is smaller than the object-background attenuation differ- ence, good delineating of the reproduced object is achieved. The edge will become sharper the smaller the W W setting used. Parallel with a decreasing WW, however, t h e sensitivity t o the setting of WC increases. 4. In the present experiments ideal conditions prevailed, with objects of similar thickness which passes perpendicularly through the CT slice. In practice, however, consideration must also be paid to partial volume effects when measuring the sizes of structures. REFERENCES 1. Baxter, B.S. and Sorensen, J . A . : Factors affecting the measurement of size and CT number in computed tomography. Invest. Radiol. 16:337-341, 1981. 2 . Koehler, P.R., Potchen, E . J . , Cole, W.R. & Studer, R . : Experimental studies of intralymphatic administration of radiotherapy. Radiology 90 : 3. Koehler, P.R., Anderson, R . E . 8 Baxter, B . : The effect of computed tomography viewer controls on anatomical measurements. Radiology 130: 189- 4. Steckel, R . J . & Cameron, T.P.: Changes in lymph node size induced by 495-501, 1968. -194, 1979. lymphangiography. Radiology 87: 753-755, 1966. Address for reprints D r Anders Magnusson Department of Diagnostic Radiology Akademiska sjukhuset S-751 85 Uppsala, Sweden 286