Evaluating the Complex Relationship of Automated … · abdominal CT protocol: 120kVp, 0.8s rotation time, 40mm beam width, 0.984 pitch, and 2.5 mm image thickness. A large scan field-of-view

Purpose To determine the influence of phantom size on automated

tube current modulation (ATCM) performance.

Introduction

Automated tube current modulation (ATCM) dynamically

optimizes radiation exposure by modulating x-ray tube

current during scanning based on a patient’s varying

anatomic attenuation, both in-plane and along the z axis.

ATCM plays a central role in controlling radiation dose and

maintaining image quality.

While the concept of ATCM is similar across

manufacturers, how ATCM is applied, and the parameters

used to control it vary (1).

In our study, we utilized a GE Healthcare HD750 CT

scanner (General Electric, Milwaukee). ATCM exposure

control parameters for this scanner include a quality

setting called “noise index” (NI) and minimum/maximum

tube current levels. The NI parameter is expected to have

a direct, linear relationship with noise and we expected

that, for a given NI value, the ATCM algorithm would result

in similar image noise regardless of patient size. However,

previous studies have reported varying correlation

between tube current modulation settings and resultant

image noise (2,3).

The purpose of this study was to evaluate the effect of NI

on a measure of image noise (standard deviation) across

a range of phantom sizes.

Materials & Methods

Four tissue-equivalent abdominal CT dose phantoms

(CIRS 007TE) of different sizes were scanned using a GE

HD750 scanner. (Figure 1). To simulate an “extra-large”

patient size, a 5th phantom was created by wrapping a fat-

equivalent ring around the large adult phantom.

We used scanning parameters similar to our routine

abdominal CT protocol: 120kVp, 0.8s rotation time, 40mm

beam width, 0.984 pitch, and 2.5 mm image thickness. A

large scan field-of-view was utilized. Auto-mA and Smart-

mA (General Electric, Milwaukee WI) were enabled and

Noise Index (NI) was varied. Images were reconstructed

using the standard reconstruction kernel.

For each phantom size/NI combination, three regions of

interest (ROI’s) were placed in an off-center, consistent

portion of 10 consecutive images. The standard deviation

of each ROI was recorded as our measure of noise. The

average of these 30 standard deviation values was

calculated and then plotted against NI for each phantom

size (Figure 2).

Using a constant NI value and the same scan protocol,

noise levels decreased with increasing phantom size. For

the small to medium phantom sizes (circumferences of 71,

86, and 96cm), the differences in slopes (1.26, 1.21, and

1.11) were relatively minor. The slopes (0.68 and 0.47) of

the large and extra-large phantoms (circumferences of 116

and 136cm) were substantially less compared to the small-

medium size phantoms, and also quite different from each

other, resulting in three distinct sets of lines on the noise

(standard deviation) vs NI plot (Figure 2).

In summary, for large and extra-large phantoms at a given

NI, image noise (standard deviation) is less than that seen

for small to medium phantoms.

Figure 2. Graph showing noise index (x axis) versus noise (standard deviation)

(y axis). Each line represents a different phantom size. For a given fixed

phantom size (e.g., 15 year old with a 71 cm circumference and 24 cm width),

image noise increased linearly with increasing noise index. However, use of the

same NI values did not produce the same noise level across phantoms of

different sizes. If a constant noise level is preferable, the NI values must be

increased for larger phantoms.

Conclusion Use of ATCM with a given noise index results in lower image

noise for larger phantoms than for smaller phantoms in the

CT scanner we studied.

To maintain a more consistent amount of image noise across

patient sizes, our data suggests utilizing larger NI values for

larger patients.

References

1. Lee CH, et al. Radiation Dose Modulation Techniques in the

Multidetector CT Era: From Basics to Practice. Radiographics 2008;

28:1451–1459.

2. Lee S, et al. Comparison of image quality and radiation dose

between combined automatic tube current modulation and fixed tube

current technique in CT of abdomen and pelvis. Acta Radiologica

2011; 52: 1101–1106

3. Daignault CP, et al. Evaluation of Dose Modulation Software Through

the Assessment of Body Mass Index, Radiation Dose and Image

Noise. J Comput Assist Tomogr 2013;37: 547–550.

4. Israel GM, et al. Patient Size and Radiation Exposure in Thoracic,

Pelvic, and Abdominal CT Examinations Performed With Automatic

Exposure Control. AJR 2010; 195:1342–1346.

5. Schindera et al. Effect of Patient Size on Radiation Dose for

Abdominal MDCT with Automatic Tube Current Modulation: Phantom

Study. AJR 2008; 190:100–105.

Evaluating the Complex Relationship of Automated Tube Current Modulation, Noise Index, Image Noise and Phantom Size

CT Jensen MD, XJ Rong PhD, EP Tamm MD, V Gershan PhD, DD Cody PhD, X Liu PhD, EK Paulson MD and V Kundra MD, PhD

Figure 1. The five different phantoms used in this project are shown above with their

corresponding scanned CT image. The dimensions of the 5 phantoms respectively (top to

bottom) were 18.5 x 24 cm (15 year old equivalent), 22 x 30 cm (small adult equivalent) , 25

x 32.5 cm (medium adult equivalent), 31 x 38.9 cm (large adult equivalent), and 38 x 47 cm

(X-large adult equivalent). The CT images show the degree to which NI must be altered to

keep image noise (standard deviation) constant over the range of phantom sizes. To keep

standard deviation, our measure for noise, at a value of 22-24, NI varied from 18 for the

smallest phantom to 42 for the largest.

0

5

10

15

20

25

30

35

40

45

50

6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

No

ise

(ST

D)

NI

15-yr old

Small Adult

Medium Adult

Large Adult

xLarge Adult

Linear (15-yr old)

Linear (Small Adult)

Linear (Medium Adult)

Linear (Large Adult)

Linear (xLarge Adult)

Discussion ATCM plays a central role in controlling image quality and

radiation exposure. Our study shows ATCM is complex. We

expected that, for a given fixed NI, the resulting images

would have a similar level of noise regardless of phantom

size. This was not the case. Our study showed a major

difference in ATCM behavior between small-medium and

large-to-extra-large phantom sizes, as illustrated by the

diverging linear pots between phantoms in Figure 2. This

figure illustrates that, for a constant level of noise across

images of different phantom sizes, much larger NI values are

needed for large-to-extra-large phantoms. It is notable that

the findings in the literature for ATCM, patient size, and noise

have been mixed (2,3). We believe one explanation is that

our extra-large phantom is larger than some phantoms and

patients previously evaluated, which may explain the

discrepancy between our study and another study that

showed constant noise across varying patient sizes (2).

Our findings have significant implications for protocol design,

particularly given the increasing number of large patients

being scanned. Israel et al. showed that patients in the 90th

percentile for weight received 3 - 4 times the organ doses of

patients in the 10th percentile during ATCM use (4,5).

Results For a given phantom size, noise increased linearly as NI

value increased (R2 = 0.989-0.999). However, the slopes

(ranged 0.47-1.26) of these plots differed between

phantom sizes (Figure 2).

Our evaluation suggests it may be possible to scan large

patients with much higher NI values than probably used

previously while preserving diagnostic image quality. We

have already used these results to create a “Noise Index

Look-up Table” for our low dose surveillance

abdomen/pelvis protocol which also utilized Model-Based

Iterative Reconstruction (MBIR).

[Refer to separate poster titled, Veo (MBIR)

Implementation: Process Description and Lessons

Learned]

Further clinical evaluations are needed to determine

optimal scan parameters for larger patients. Authors have

reported higher subjective image quality with increasing

patient size, even with constant image noise (5). This,

coupled with our results, may suggest the possibility of

significant further dose reduction related to scanning of

large patients.

We would note that this study has several limitations. First,

we studied only one CT scanner model from a single

manufacturer. While the results cannot be directly

extrapolated to other scanners, our approach may be

useful for assessing the behavior of ATCM for other

manufacturers. Secondly, the use of five phantoms does

not fully demonstrate the range of patient sizes and shapes

encountered clinically. Thirdly, our extra-large phantom

was created by the addition of a fat-equivalent layer to the

next largest size phantom. The other phantoms, in contrast,

differed as a result of scaling. Finally, our study was based

on phantoms. We therefore advise caution in implementing

our findings when scanning patients. For instance, we used

a process of using progressively larger noise indices for

large patients rather than going directly to large NI values.

71 (24)

86 (30)

136 (47)

116 (38.9)

96 (32.5)