Radiation dose and image quality in spiral computed tomography ...

2 downloads 0 Views 504KB Size Report
provided with four different types of high specification spiral CT scanners. ... on four different spiral CT scanners with a dedicated ionization chamber inserted ...
T he British Journal of Radiology, 71 (1998), 734–744

© 1998 The British Institute of Radiology

Radiation dose and image quality in spiral computed tomography: multicentre evaluation at six institutions 1R J SCHECK, MD, 1E M COPPENRATH, MD, 2M W KELLNER, MD, 3K J LEHMANN, MD, 1C ROCK, MD, 1J RIEGER, MD, 4L ROTHMEIER, MD, 5F SCHWEDEN, MD, PhD, 6A A BA¨UML, PhD and 1K HAHN, MD 1Department of Diagnostic Radiology, Klinikum Innenstadt, Ludwig-Maximilians-University of Munich, Ziemssenstraße 1, 80336 Munich, and Departments of Radiology, 2University of Wu¨rzburg, Josef-SchneiderStraße 2, 97080 Wu¨rzburg, 3University of Mannheim, Theodor-Kutzer-Ufer 1-3, 68167 Mannheim, 4TU University of Munich, Ismaninger Straße 22, 81675 Munich, 5University of Mainz, Langenbeckstraße 1, 55131 Mainz, and 6Institut fu¨r Strahlenhygiene (BfS), Mu¨nchen-Neuherberg, 85764 Oberschleißheim, Germany Abstract. The purpose of this study was to evaluate the correlation of radiation dose with image quality in spiral CT. Seven clinical protocols were measured in six different radiological departments provided with four different types of high specification spiral CT scanners. Central and surface absorbed doses were measured in acrylic. The practical CT dose index (PCTDI) was calculated for seven clinical examination protocols and one standardized protocol using identical parameters on four different spiral CT scanners with a dedicated ionization chamber inserted into PMMA phantoms. For low contrast measurements, a cylindrical three-dimensional (3D) phantom (different sized spheres of defined contrast) was used. Image noise was measured with a cylindrical water phantom and high contrast resolution with a Perspex hole phantom. Image quality phantoms were scanned using the parameters of the clinical protocols. Images were randomized, blinded and read by six radiologists (one from each institution). PCTDI values for four different scanners varied up to a factor between 1.5 (centre) and 2.2 (surface) for the standardized protocol. A greater degree of variation was observed for seven clinical examination protocols of the six radiological departments. For example, PCTDI varied up to a factor between 1.7 (cerebrum protocol ) and 8.3 (abdomen paediatric protocol ). Low contrast resolution correlates closely with dose. An improvement in detection from 8 mm to 4 mm sized spheres needs approximately a ten-fold increase in dose. Noise shows a moderate correlation with PCTDI. High contrast resolution of clinical protocols is independent of PCTDI within a certain range. Differences in modern CT scanner technology seem to be of less importance for radiation exposure than selection of protocol parameters in different radiological institutes. Future discussion on guidelines regarding optimal (patient adapted) tube current for clinical protocols is desirable.

Since its introduction in 1989 [1] the spiral technique has completely revolutionized CT as a result of its immense benefits. Spiral CT allows a whole CT examination to be carried out within a single breathhold [2]. Continuous data acquisition means small lesions can be detected which may be missed using conventional CT. Contrast media spiral CT offers the opportunity to examine the arterial system selectively. In addition, the advantage of a short acquisition time may be crucial for specific patients (e.g. children, restless patients). However, the efficiency of spiral CT may also have some adverse aspects. The short scanning time means there is a danger of uncritical use being Received 16 June 1997 and in final form 20 February 1998, accepted 11 March 1998. Supported in part by a grant from the Deutsche Ro¨ntgengesellschaft (DRG), PO Box 1336, 61283 Bad Homburg, Germany. 734

made of this technique. The major contribution to collective dose in diagnostic radiology results from CT, as reported recently [3, 4]. However, so far only draft guidelines exist in Europe concerning optimal image quality and radiation dose in CT [5]. Recommendations of CT manufacturers vary with regard to clinical protocols and cannot really be compared because of different scanner makes and models. Furthermore, important information concerning central and surface absorbed dose in correlation with image quality may be missing or may not be comparable. The surface dose, for example, would be relevant for estimating breast dose in chest CT. To our knowledge, there are no published data which correlate applied radiation dose with image quality for high specification modern spiral CT scanners. Consequently, in this multicentre trial we examined four different high specification spiral T he British Journal of Radiology, July 1998

Radiation dose and image quality in spiral CT

scanners in six radiological departments. The purpose of our study was four-fold: (a) to measure differences in radiation output between the scanners when standardized parameters are applied; ( b) to estimate the variability of single practical CT dose index (PCTDI) measurements in central and peripheral positions; (c) to measure PCTDI applied for clinical examinations in six radiological departments and (d) to determine image quality ( low contrast resolution, noise and high contrast resolution) in correlation with PCTDI for clinical protocols.

Methods and materials Six academic radiological departments with four different, modern spiral scanners (Table 1) joined the multicentre study, which took place between March 1996 and August 1996. A standardized protocol with nearly identical parameters (120 kV, 240–250 mA, scan time 1 s, section thickness 5 mm, reconstruction parameters as for liver examination) was determined to compare PCTDI measurements of the four different scanners centrally and peripherally. Each institute was asked for their clinical examination CT protocols for defined indications (Tables 2 and 3). These institute-specific protocols had been used routinely for at least 4 weeks before dose and image quality measurements took place in each institution. The extent to which these clinical protocols varied from those recommended by the manufacturer was not investigated. Dose was measured as absorbed dose in acrylic

(polymethyl methylacryl, PMMA) with a penciltype ionization chamber of 10 cm length (type PC4-P, Wellho¨fer Dosimetrie, Schwarzenbruck, Germany), a Capintec digital dosimeter WK 92 (Wellho¨fer Dosimetrie, Schwarzenbruck, Germany) and cylindrical PMMA phantoms ( body phantom of 32 cm diameter and head phantom of 16 cm diameter, length 14 cm; Mayo, Mc. Crohan, Conway, USA) according to the method described by Rothenberg et al [6]. These cylindrical phantoms possess drilled holes in the centre and in four symmetrical positions (3, 6, 9, 12 o’clock, each located 1 cm below the surface) to insert the chamber. The PMMA phantom was positioned in the centre of the gantry, for chest and abdomen protocols on the patient couch; for head, neck and inner ear measurements on the head support device. The ionization chamber was placed consecutively in each of the five holes provided by the phantom. According to Kalender and Polacin [7] a spiral CT scan of a volume with a pitch factor of 1 yields the same dose as that delivered by scanning the same volume with single contiguous sections and the same mAs value when the same collimation width is used. Therefore, measurements were performed with one single scan of 360° tube rotation. The number of measurements was 20 for each chamber position (100 measurements for each protocol ) to register scanner dependent variations of PCTDI measurement. The ionization chamber was also used to measure the dose free-in-air at the centre of rotation (Table 4). Dose was calculated as PCTDI or PCTDI/pitch and average dose (D ) or average dose/pitch (D ) according to AVE AVE* the following formula:

Table 1. Type and number of CT scanners included in the study

Practical CTDI [8]:

Type of scanner

Number

1 PCTDI= d

GE High Speed Advantage Philips Tomoscan AV Picker PQ 5000 V Siemens Somatom Plus 4 A

1 1 2 2

−50 mm where d is the nominal section thickness

Identical scanners were used at institutions A and D as well as at B and C.

Table 2. Indications for clinical protocols Protocol

Indication

Cerebrum Inner ear Neck Chest adult Abdomen adult Chest paediatric Abdomen paediatric

Exclusion of bleeding or infarction Exclusion of fracture Staging of tumour (e.g. lymphoma) Staging of tumour (e.g. lymphoma) Staging of tumour (e.g. lymphoma) Staging of tumour (e.g. lymphoma) Staging of tumour (e.g. lymphoma)

Clinical protocols are defined for a 70 kg adult or a 5-year-old child (20–25 kg body weight).

T he British Journal of Radiology, July 1998

+50 mm

P

D(x) dx

Average dose [8]: D =1/3 PCTDI central+2/3 PCTDI peripheral AVE In order to compare dose values of conventional and various spiral protocols with different pitch factors and to be able to estimate patient exposure we suggest the introduction of the axial scan corresponding average dose D , which is AVE* defined as: D

D = AVE AVE* pitch

To assess low contrast resolution a special 3D phantom was used [9] (Figure 1). The phantom design consists of a PMMA cylinder 16 cm in diameter which contains a water equivalent solution. Spheres are embedded, with a fixed low 735

R J Scheck, E M Coppenrath, M W Kellner et al Table 3. Parameters, D and image quality for clinical protocols: CT of the cerebrum, inner ear, neck, adult chest, AVE* adult abdomen, paediatric chest, paediatric abdomen Institution A

B

C

D

E

F

High contrast resolution (mm) in x–y plane

Axial 120 420 8 — 54.3 4.3 (±0.47) 1.8 (±0.05) 0.95 (±0.13)

Axial 120 375 8 — 37.2 4.2 (±0.37) 2.5 (±0.05) 0.68 (±0.07)

Spiral 120 375 5 1.25 40.7 4.2 (±0.37) 2.5 (±0.05) 0.67 (±0.05)

Axial 140 300 8 — 56.9 3.7 (±0.47) 1.6 (±0.00) 0.90 (±0.00)

Axial 120 500 5 — 64.7 4.0 (±0.00) 2.8 (±0.06) 0.62 (±0.04)

Axial 120 420 5 — 43.4 4.0 (±0.00) 2.3 (±0.00) 0.72 (±0.04)

Inner ear protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch (mGy) D AVE* Low contrast resolution (mm) Noise High contrast resolution (mm) in x–y plane High contrast resolution (mm) in x–z plane

Spiral 120 420 1 1 91.5 — — 0.50 (±0.00) 1.0 (±0.00)

Axial 140 150 1 — 23.6 — — 0.50 (±0.00) 1.0 (±0.00)

Spiral 120 100 1 1 15.0 — — 0.50 (±0.00) 0.92 (±0.10)

Axial 120 150 1 — 64.9 — — 0.50 (±0.00) 1.0 (±0.00)

Axial 120 500 1 — 82.3 — — 0.50 (±0.00) 1.27 (±0.13)

Axial 140 160 1 — 26.5 — — 0.50 (±0.00) 1.0 (±0.00)

Spiral 140 240 5 1.5 29.2 4.7 (±0.47) 6.7 (±0.14) 5.2 (±0.39) 0.62 (±0.05)

Spiral 120 150 5 1.25 15.3 5.3 (±0.47) 6.1 (±0.10) 5.2 (±0.25) 0.63 (±0.05)

Spiral 120 200 5 1.5 24.0 4.8 (±0.37) 5.7 (±0.12) 5.1 (±0.29) 0.55 (±0.08)

Spiral 140 160 5 1.5 20.5 5.7 (±0.47) 7.1 (±0.08) 5.8 (±0.45) 0.63 (±0.06)

Spiral 120 250 3 1 32.6 5.3 (±0.47) 6.5 (±0.20) 5.1 (±0.17) 0.65 (±0.05)

Spiral 120 210 5 1.5 14.4 4.8 (±0.37) 3.6 (±0.12) 3.7 (±0.23) 0.72 (±0.04)

Spiral 140 180 8 1.5 12.7 5.3 (±0.47) 11.0 (±0.12) 7.2 (±0.7) 0.70 (±0.06)

Spiral 120 100 8 1.5 4.4 5.8 (±0.37) 23.8 (±0.13) 20.0 (±2.20) 0.65 (±0.05)

Spiral 120 200 8 1.5 13.1 5.0 (±0.00) 18.8 (±0.13) 14.8 (±1.80) 0.62 (±0.04)

Spiral 120 125 8 1.5 6.2 5.8 (±0.37) 22.0 (±0.12) 12.0 (±1.41) 0.67 (±0.05)

Spiral 120 200 7 1 13.4 5.2 (±0.37) 19.5 (±0.25) 10.7 (±0.40) 0.63 (±0.05)

Spiral 120 200 7 1.4 6.5 5.5 (±0.50) 14.1 (±0.17) 12.0 (±1.48) 0.73 (±0.05)

Cerebrum protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch (Central) (mGy) D AVE* Low contrast resolution (mm) Noise (central)

Neck protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch (mGy) D AVE* Low contrast resolution (mm) Noise (central) Noise (peripheral ) High contrast resolution (mm) in x–y plane Chest adult protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch D (mGy) AVE* Low contrast resolution (mm) Noise (central) Noise (peripheral ) High contrast resolution (mm) in x–y plane

Continued on next page.

736

T he British Journal of Radiology, July 1998

Radiation dose and image quality in spiral CT Table 3. (Continued) Institution

Abdomen adult protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch (mGy) D AVE* Low contrast resolution (mm) Noise (central) Noise (peripheral ) High contrast resolution (mm) in x–y plane Chest paediatric protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch D (mGy) AVE* Low contrast resolution (mm) Noise (central) Noise (peripheral ) High contrast resolution (mm) in x–y plane Abdomen paediatric protocol Technique Voltage ( kV) Current (mAs) Slice thickness (mm) Pitch D (mGy) AVE* Low contrast resolution (mm) Noise (central) Noise (peripheral ) High contrast resolution (mm) in x–y plane

A

B

C

D

E

F

Spiral 120 210 8 1.5 10.5 5.5 (±0.50) 9.8 (±0.05) 6.6 (±0.71) 0.72 (±0.04)

Spiral 120 150 8 1.5 6.7 5.8 (±0.37) 16.9 (±0.06) 13.8 (±1.59) 0.70 (±0.06)

Spiral 120 200 8 1.5 13.1 4.8 (±0.37) 14.3 (±0.13) 11.1 (±1.22) 0.65 (±0.05)

Spiral 120 175 10 1.2 10.8 5.3 (±0.94) 13.2 (±0.29) 8.3 (±1.20) 0.72 (±0.04)

Spiral 140 225 7 1 21.2 4.2 (±0.37) 9.5 (±0.07) 6.0 (±0.24) 0.76 (±0.04)

Spiral 120 240 7 1.4 7.9 5.0 (±0.00) 13.3 (±0.19) 11.0 (±1.22) 0.73 (±0.05)

Spiral 120 37 5 1.5 3.1 8.0 (±0.00) 12.7 (±0.33) 10.5 (±0.86) 0.72 (±0.04)

Spiral 120 30 5 1 5.7 7.7 (±0.75) 14.1 (±0.50) 11.5 (±0.75) 0.65 (±0.05)

Spiral 120 100 5 1.5 9.1 6.0 (±0.00) 19.1 (±0.10) 17.5 (±2.0) 0.50 (±0.00)

Spiral 120 50 5 1 6.6 6.0 (±0.00) 10.5 (±0.22) 8.62 (±0.60) 0.70 (±0.08)

Spiral 120 150 5 1 19.7 5.2 (±0.37) 6.7 (±0.32) 5.1 (±0.17) 0.65 (±0.05)

Spiral 120 100 5 1 10.3 6.0 (±0.00) 5.6 (±0.33) 5.27 (±0.33) 0.75 (±0.05)

Spiral 120 37 5 1.5 3.1 8.2 (±0.37) 9.8 (±0.17) 8.4 (±0.66) 0.78 (±0.09)

Spiral 120 30 5 1 5.7 6.0 (±0.00) 10.2 (±0.05) 8.9 (±0.54) 0.65 (±0.05)

Spiral 120 100 5 1.5 9.1 5.3 (±0.47) 8.6 (±0.07) 8.2 (±0.48) 0.50 (±0.00)

Spiral 120 50 5 1 6.6 6.2 (±0.90) 8.3 (±0.26) 7.0 (±0.46) 0.73 (±0.07)

Spiral 120 200 5 1 26.0 4.8 (±0.37) 3.50 (±0.12) 3.0 (±0.09) 0.75 (±0.05)

Spiral 120 100 5 1 10.3 5.3 (±0.47) 4.5 (±0.16) 4.3 (±0.29) 0.80 (±0.06)

Standard deviations for low contrast resolution, noise and high contrast resolution are written in parentheses.

Table 4. Kerma free in air (centre of rotation) for 120 kV and a section of 10 mm Type of scanner

Kerma free in air (mGy/100 mAs)

GE High Speed Advantage Philips Tomoscan AV Picker PQ 5000 V Siemens Somatom Plus 4 A

20 20 21 18

T he British Journal of Radiology, July 1998

contrast (5.5 HU) and different diameter (8, 6, 5, 4, 3 mm) assembled in the x–y and in the x–z plane (Figure 2). For high contrast resolution, Perspex hole patterns were used (Institute of Medical Physics, University of Erlangen, Germany). In the x–y plane there are 11 arrays of five drilled holes each with diameters and separations of 0.5–4 mm (Figure 3a). In the x–z plane, the hole diameter and separation varies from 0.8 to 8 mm (Figure 3b). The phantoms were positioned centrally on the patient couch (abdomen and chest protocols) or on the head support device (cerebrum neck and 737

R J Scheck, E M Coppenrath, M W Kellner et al

Figure 1. 3D low and high contrast phantom. Rows of spheres with a fixed low contrast and different diameters as well as different sized holes in a Perspex test pattern are assembled in both the x–y and the x–z planes.

(a)

For estimation of high and low contrast resolution, the six best reconstructed images of each examination were chosen by one radiologist (RJS) and printed on X-ray film by the laser printer of each institute. For this purpose window width and level settings (Figures 2 and 3) were defined as follows. For evaluation of low contrast resolution, window width was fixed at 60 HU, window level was set 2 HU above the mean density value of the fluid surrounding the spheres. For high contrast resolution, window width was fixed at 1200 HU and window level at −450 HU. Two series of images, each consisting of six consecutive images (reconstructed every 2 mm), were printed on a film sheet, which resulted in 12 images per film. The films were randomized, blinded with tapes, and evaluated by six radiologists of the different institutes using a standardized interview schedule.

(b)

Figure 2. Low contrast resolution in the scanner’s isocentre measured with a 3D phantom in the x–y plane (a). Multiplanar reformation demonstrates z-axis low contrast resolution which is limited by high image noise (b).

(a)

(b)

Figure 3. Spatial resolution in the scanner’s isocentre measured with a Perspex hole phantom. Even the smallest holes are separable in the x–y plane (a). Multiplanar reformation demonstrates z-axis spatial resolution ( b), which allows separation of holes up to the ninth row.

inner ear protocols) and scans were performed for clinical protocols. Reconstruction interval in the x–y plane was 2 mm for all protocols except for the inner ear protocol, for which 1 mm was used. Multiplanar reconstructions in the x–z plane were done with an interval of 1 mm. 738

Detectability of the row of smallest-diameter spheres for a series consisting of six consecutive images determined low contrast resolution. For high contrast resolution, the last row had to be assessed in which the holes were just separable. Noise was evaluated by a cylindrical water T he British Journal of Radiology, July 1998

Radiation dose and image quality in spiral CT

phantom (16 and 32 cm in diameter). The water phantom was positioned in the gantry and scanned with the clinical protocols. Standard deviation of density within a region of interest (ROI) was calculated to determine the degree of noise. ROIs were evaluated in five (one central and four peripheral ) image positions (area 500/2000 mm2 in 16/32 cm phantoms, respectively). Measurements were performed in three consecutive images, reconstructed by an interval of 2 mm. Although the low and high resolution parts of the 3D phantom do not offer sub-millimetre increments, mean values and standard deviations for all the observers were calculated from the diameters of spheres and holes which could be separated. Linear correlation coefficients (r2) were calculated by using double-logarithmic numbers of corresponding variables (PCTDI or D , detecAVE* tion of spheres, image noise). Statistical significance was assessed with a two-tailed Student’s t test.

Results Absorbed dose: standardized protocol Using comparable scanning parameters (120 kV, 240–250 mA, scan time 1 s, section thickness 5 mm, reconstruction parameters as for liver examination, body phantom placed on the patient couch) PCTDI differences in the phantom centre were moderate, and in the four surface positions between the scanners were considerable (Table 5, Figure 4). Two of the six scanners (institutions B and C) are provided with an additional filter (3.5 mm aluminium equivalent) which can be activated deliberately and is used for some of the clinical protocols measured in this study. The values shown in Table 5 and Figure 4 were measured with this filter activated. Without this additional filter, PCTDI rises about 30% centrally and 50% in the surface position. PCTDI/100 mAs between the scanners was different by a factor of 1.3 (1.5) for the central position and a factor of 1.6 (2.2) for the

surface position (values in parentheses without additional filter). Surface PCTDI in the 6 o’clock position (measurement position 1 cm above the patient couch) is between 2.6% and 13.7% lower than the mean value of the other surface positions (Table 5), which may be explained by X-ray attenuation due to the patient support device. Standard deviation of PCTDI values varied depending on the measuring position and scanner. Standard deviation in the central position was always low (0.3–1.4%), but differed significantly in the surface positions (0.5–8.5%, Table 5). Kerma measured in the centre of rotation freein-air for 10 mm slices ranged from 18 to 21 mGy/100 mAs (Table 4).

Absorbed dose: clinical protocols For routine CT examinations of the neck, chest and abdomen all the institutes use the spiral technique. Cerebrum and inner ear protocols are mainly scanned in conventional mode (Table 3). Routinely applied doses calculated as PCTDI or D vary considerably in the six radiological AVE* departments for most of the seven protocols that were investigated (Table 3). Between different institutions D varies by a AVE* factor of 1.7 (cerebrum protocol ), 6.1 (inner ear protocol ), 2.3 (neck protocol ), 3.0 (chest adult protocol ), 3.2 (abdomen adult protocol ), 6.3 (chest paediatric protocol ) and 8.3 (abdomen paediatric protocol ), respectively.

L ow contrast resolution Table 3 shows the low contrast resolution (diameter of smallest spherical lesion recognized) for clinical examination protocols. In the transverse plane (x–y plane) spheres with a diameter of about 4 mm were recognized by using cerebrum protocols; lesions of 5–6 mm (neck protocols), 5–6 mm (chest adult protocols), 4–6 mm (abdomen adult protocols), 5–8 mm (chest and abdomen paediatric protocols) were detected, respectively. Low contrast resolution proved to be dependent on D (r2=0.93, p