A method to obtain the same levels of CT image noise for patients of various sizes, to minimize radiation dose

A method to obtain the same levels of CT image noise for patients of various sizes, to minimize radiation dose

G Starck, PhD ¹^,2 L Lönn, MD, PhD ³ Å Cederblad, MSc ¹ E Forssell-Aronsson, PhD ² L Sjöström, MD, PhD ⁴ and M Alpsten, PhD ¹^,2

¹Division of Medical Physics and Medical Engineering and Departments of ²Radiation Physics, ³Radiology and ⁴Medicine, Göteborg University, Sahlgrenska University Hospital, Bruna Stråket 13, SE-413 45 Göteborg, Sweden

View larger version (37K):

[in a new window]

Figure 1. Scanning geometry used in the derivation of noise in the ideal CT concept. This figure illustrates the geometry of one parallel projection with the sampling interval R. The air kerma length product (KLP_tube) is measured at the centre of rotation using a pencil-shaped ionization chamber.

View larger version (14K):

[in a new window]

Figure 2. (a) CT noise figure vs photon incidence on the detector (KLP_det). The noise figure data are plotted separately for each phantom for CT_a ( ${circ}$ — ${circ}$ ) and CT_b ( x --- x ). The theoretical relation between KLP_det and CT noise figure (Equation 7

) (—) was plotted at a level where it coincided with the data of the best performing CT system. Error bars representing the maximum and minimum values obtained from six scans are smaller than the spatial extent of the data symbols and are therefore not shown. The horizontal separation of the plots for CT_a and CT_b indicates the difference in the number of photons required by the two systems to achieve the same noise figure level. CT_a yielded erroneous attenuation values at the four lowest levels of KLP_det (see Results). Vertical dashed lines show the values of photon incidence required for the noise level specified for tissue volume determination with CT ( $<=$ 30 HU). (b) When the attenuation of 1 cm water was added to the attenuation of the phantoms that were positioned on the tabletop (to account for the attenuation in the tabletop), the curves of these phantoms merged with the curves of the other phantoms (which were attached to the phantom holder) and formed one distinct line for both CT_a ( ${circ}$ — ${circ}$ ) and CT_b ( x --- x ).

View larger version (15K):

[in a new window]

Figure 3. Profiles of CT image noise from the CT_a system (48 cm field of view, 256 x 256 pixel matrix) along the horizontal diameter of the 48.4 cm polyethylene phantom. Noise profiles of scans with 250 mAs, 10 mm ( ${circ}$ ) (10 Hounsfield units (HU) in the centre) and 100 mAs, 3 mm ( ${square}$ ) (30 HU in the centre) are shown. Standard deviations (SD) were obtained from circular regions ofinterest (40 mmdiameter) and the plotted profiles were normalized to unity at the centre of the phantom. Error bars represent the maximum and minimum values of six measurements.

View larger version (16K):

[in a new window]

Figure 4. Noise in transaxial CT images obtained at the top of the iliac crest in 11 volunteers. Images were acquired with reduced radiation dose to the volunteer with the CT_a system using patient-specific scan parameters (Table 3

). The dashed line indicates the highest noise level that would be obtained centrally in the image of a circular water phantom using these scan parameters. Standard deviations were obtained in four regions of interest in homogeneous tissue (central in the left and right psoas muscles ( ${circ}$ ) and peripheral in subcutaneous fat in the left and right dorsal waist areas( x )).

View larger version (84K):

[in a new window]

Figure 5. Transaxial CT images obtained at the top of the iliac crest in two single volunteers. The scans were performed with reduced radiation dose to each volunteer with the CT_a system using patient-specific scan parameters (Table 3

). (a) Image of a slim volunteer(31 cm greater diameter) obtained with the optimum radiation dose. The image noise is high but does not exceed the specified level of 30 Hounsfield units (HU). (b) Image of a corpulent volunteer (47 cm greater diameter). The image noise (approximately 10 HU) is much lower than the required level of 30 HU, indicating that a further substantial dose reduction is possible.