Characterization and Quantiﬁcation of Image Quality in CT Imaging

Systems: A Phantom Study

Camilla Scapicchio

1,2 a

, Manuela Imbriani

1,3

, Francesca Lizzi

, Mariagrazia Quattrocchi

Alessandra Retico

, Sara Saponaro

1,2

, Maria Irene Tenerani

1,2

, Alessandro Tofani

Arman Zafaranchi

1,2

and Maria Evelina Fantacci

1,2

University of Pisa, Pisa, Italy

National Institute for Nuclear Physics, Pisa, Italy

Medical Physics Department, Azienda Toscana Nord Ovest Area Nord, Lucca, Italy

Keywords:

Computed Tomography, Image Quality, Detectability Index, CatPhan Phantom.

Abstract:

Computed Tomography (CT) is a widely used imaging technique in lung cancer screening programs. To ad-

dress the problem of exposing potentially healthy patients to ionizing radiation, Iterative Reconstruction (IR)

algorithms can be employed. Indeed, traditional Filtered Back Projection reconstruction does not deliver ad-

equate image quality with reduced dose levels. IR instead is prone to preserve diagnostic information and

resolution while reducing noise and radiation dose. We characterized image quality for two CT scanners

equipped with different iterative algorithms by using a quantitative metric, the detectability index. We com-

pared the dependence of the image quality on the dose and the iterative level when the human visual perception

is considered or not in the detectability index deﬁnition. It has been found that similar image quality can be

obtained by using different scanners and different combinations of dose and iterative levels. This allows us

to extrapolate the protocols corresponding to a lower dose while preserving as much as possible the imaging

properties.

1 INTRODUCTION

Chest Computed Tomography (CT) is the best and

most used imaging modality to detect small pul-

monary nodules that are the early radiological signs of

lung cancer (Raju et al., 2017; Kennedy et al., 2022).

It is recommended as a screening tool in high-risk

populations, thus exposing potentially healthy peo-

ple to ionizing radiation (Yeh et al., 2016; Cao et al.,

2022). For this reason, in recent years a wide effort

has been dedicated to the development of new strate-

gies to reduce the radiation dose delivered to the pa-

tient during the CT acquisition complying with the

ALARA principle (radiation doses should be kept As

Low As Reasonably Achievable) while maintaining

the same imaging properties (Smith-Bindman et al.,

2019). In particular, the advent of Iterative Recon-

struction (IR) algorithms has been one of the main

advances in CT technology introduced to reduce the

image noise in diagnostic images (Beister et al., 2012;

Caramella et al., 2018), thereby allowing for the re-

https://orcid.org/0000-0001-5984-0408

duction of the dose required for routine imaging.

However, the use of IR algorithms is limited by the

necessity of preserving the image quality to capture

diagnostic information. The intrinsic non-linear na-

ture of the IR reconstruction algorithms compared to

conventional Filtered Back-Projection (FBP) has been

highlighted (Dodge et al., 2016). Therefore the tra-

ditional metrics, such as the Contrast-To-Noise ratio

or the Modulation Transfer Function (MTF), appear

to be inadequate to comprehensively describe clinical

imaging performance when IR techniques are used for

image reconstruction (Samei et al., 2019). This is be-

cause IR algorithms result in object-dependent resolu-

tion and noise performances. An accurate evaluation

and optimization of the best IR blending levels for the

different possible applications is needed to deﬁne new

low-dose diagnostic protocols (Barca et al., 2018a).

Nowadays powerful instruments based on Artiﬁ-

cial Intelligence (AI) have also been developed to ex-

tract signiﬁcant diagnostic information from chest CT

images acquired with a lower radiation dose. This of-

fers new perspectives on lung cancer diagnosis and

Fantacci, M.

Characterization and Quantiﬁcation of Image Quality in CT Imaging Systems: A Phantom Study.

DOI: 10.5220/0012400400003657

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2024) - Volume 1, pages 289-296

ISBN: 978-989-758-688-0; ISSN: 2184-4305

289

other lung pathologies. However, a more speciﬁc

analysis of image quality has to be implemented when

the CT scans are aimed to be provided to an AI algo-

rithm.

In this study, we characterize two CT scanners

in terms of image quality across different acquisition

and reconstruction parameters by means of phantom

image acquisitions. It is not so easy to deﬁne a com-

prehensive metric that quantiﬁes image quality, be-

cause different objective measures can impact it, such

as resolution, noise, and contrast. Typically it is use-

ful to consider these measures separately. However,

for some applications, it is useful to have a unique

deﬁnition to quantify image quality as a whole. We

adopted a new comprehensive more adequate metric

to quantitatively deﬁne the image quality, which is the

detectability index (d’) (Samei et al., 2019). It allows

us to evaluate the joint effect of all the factors that can

impact image quality. This new index is also more

suitable to describe image quality when IR algorithms

are used.

The main objective of this study is to evaluate the

imaging properties of chest CT reconstructed with

different Iterative Reconstruction blending levels by

leveraging this new detectability index in order to ﬁnd

a strategy to reduce the radiation dose delivered in

CT acquisitions without degrading the imaging prop-

erties. The possibility of using IR algorithms in con-

junction with CAD (Computer Aided Detection) sys-

tems has already been investigated as useful to bring

an improvement in image quality when very low radi-

ation exposure levels are employed (Caramella et al.,

2018; Barca et al., 2018b). In addition, nowadays

these CAD systems are being updated with new Deep

Learning-based tools (Manickavasagam et al., 2022;

Forte et al., 2022). Thus the image quality charac-

terization here proposed is ﬁnally aimed at applying

Deep Learning- or Radiomics-based image analysis

systems.

2 MATERIALS

2.1 Phantom and CT Scanners

The phantom under consideration is the model

Catphan-500 ® (The Phantom Laboratory, NY, USA).

It is a commercially available phantom commonly uti-

lized in clinical procedures for quality control aimed

at ensuring ideal imaging performance for a CT scan-

ner. It has a cylindrical shape and it is composed of

4 modules (Figure 1). Full technical speciﬁcations of

the Catphan-500 phantom can be found in the manual

(Mail, 2013).

Figure 1: Illustration of the Catphan-500 phantom model

(Mail, 2013).

In this study, the CTP401 and CTP486 modules

of the phantom were considered for the computation

of the detectability index. The CTP486 module is a

homogeneous water-equivalent module. The CTP401

module includes seven cylindrical inserts of differ-

ent materials, with the nominal CT Hounsﬁeld Units

(HU) reported in the manual (Mail, 2013), in a uni-

form water-equivalent background.

The CT scans of the Catphan phantom have been

acquired at San Luca Hospital-Azienda Toscana Nord

Ovest (ATNO)-Lucca, Italy. We used the two CT

scanners from two different vendors available in this

hospital: Revolution Evo 64 Slice (GE Healthcare),

and Aquilon CX 128 Slice CT (Toshiba).

2.2 Image Data Acquisition

The phantom was scanned in helical modality. We

aimed to characterize image quality in conditions as

close as possible to those of lung imaging. There-

fore, we started from the institutional clinical CT pro-

tocols for diagnostic tasks in chest imaging used in the

Hospital for the two scanners ( CTDI

vol

∼ 7mGy for

the GE scanner, and CTDI

vol

∼ 8mGy for the Toshiba

scanner). Then we explored three other dose radiation

levels: high (twice the standard value, i.e. the one of

the institutional protocol), reduced (60% of the stan-

dard value), and low (30% of the standard value). We

also used four percentages of the IR blending level

(FBP alone and three increasing strength levels of the

IR algorithm available for the system, which is the

ASIR for GE and the AIDR 3D for Toshiba). A sharp

reconstruction kernel has been used in both systems:

LUNG for GE, and FC56 (equivalent to LUNG) for

Toshiba. The complete list of parameters set in the

two scanners is reported in Table 1. Each acquisition

with the same set of parameters has been repeated

3 times, each time removing and repositioning the

BIOIMAGING 2024 - 11th International Conference on Bioimaging

290

phantom. A total of 96 CT scans of the phantom were

made available at the end of the acquisition process.

3 METHODS

3.1 Image Quality Quantiﬁcation

The new comprehensive more adequate task-based

metric that we used to quantify image quality is the

detectability index d’. It is based on model observer

methodologies that try to combine the system im-

age performance, the task characteristics, and the de-

gree of clinician image perception (Vennart, 1997).

The procedure to compute the detectability index is

reported in the American Association of Physicists

in Medicine task group report-23 (AAPM TG-233)

(Samei et al., 2019). We adopted the d’ deﬁnition pre-

sented in this report:

′2

R R

T TF

dudv

R R

T T F

NPS

dudv

(1)

where W is the Task function, i.e., the Fourier

transform of a synthesized ideal image of a signal to

be detected. In fact, the d’ index quantiﬁes the degree

of separation for signal present/signal absent distribu-

tions on the image. We considered a signal with a cir-

cular shape and a designer contrast proﬁle. The T TF

is the Task Transfer Function that is a representative

metric of spatial resolution. It replaces the traditional

MT F that may not represent the imaging system’s re-

sponse to an arbitrary input object because the system

resolution becomes dependent on the object contrast

and background noise level. Instead, when reporting

T T F, the background noise, the object’s contrast, and

the object’s radial location are included. This is the

reason why d’ is more suitable when IR algorithms

are used. NPS is the Noise Power Spectrum that sum-

marizes the noise texture in the spatial frequency do-

main. E is a function that simulates the eye ﬁlter mod-

eling the human visual system sensitivity to different

spatial frequencies. The arrays should have the same

shape, typically representing frequency values along

the u and v dimensions.

Hence, the detectability index, depending on spa-

tial resolution, contrast and noise, in turn, depends

on the dose, tube potential, tube current modulation

setting, phantom size, task size, task contrast, image

thickness, reconstruction algorithm, and reconstruc-

tion kernel. It is therefore evident that this metric al-

lows us to evaluate the joint effect of all the factors

that can impact image quality.

To compute the detectability index on our phan-

tom CT images with a semi-automatic procedure,

we used the imQuest open-source software (Duke

University, Durham, NC, USA) (Solomon, 2018).

Once each CT scan is uploaded, the imQuest soft-

ware requires the ﬁrst step of T T F computation (Fig-

ure 2). An insert has to be chosen as a reference

and we placed a circular ROI with a radius about

twice that of the chosen insert (∼ 60 pixels) on it.

In this manuscript, we will show the results related

to polystyrene (with a nominal object-to-background

contrast (|∆HU|≈100) adequate for low-contrast di-

agnostic tasks) and air inserts. However, we repeated

the d’ computation by also considering all the other

inserts. The second step is the NPS computation. Five

64x64 pixels ROIs in the CTP486 homogenous mod-

ule (Figure 2) have been drawn. The 2D −NPS is then

computed as the area-normalized Fourier transform of

the ROI.

Figure 2: Manually ROI placement for TTF (on the left)

and NPS (on the right) computation on imQuest.

Then, the W parameters have to be set in imQuest.

We wanted to simulate a 5-mm diameter lesion with

the contrast automatically taken from the insert con-

sidered for the TTF computation. Therefore the pa-

rameters set in imQuest were: designer contrast pro-

ﬁle, signal diameter of 5 mm, proﬁle exponent of 1.

For the eye ﬁlter E, we used the NPWE model ob-

server obtained with the non-prewhitening matched

ﬁlter by adding the eye ﬁlter correction (Burgess,

1994), selecting the radial noise generation mode and

the Saunders visual response function. To repro-

duce the working scenario of radiologists, we also set

the following interpretation conditions: display pixel

pitch of 0.2 mm, zoom factor of 1.74, viewing dis-

tance of 500 mm, and a ﬁeld of view of 380 mm. Once

all these steps are performed, the software automati-

cally outputs the detectability index value, by using

the equation (1).

Characterization and Quantiﬁcation of Image Quality in CT Imaging Systems: A Phantom Study

291

Table 1: Acquisition and reconstruction parameters used for the two scanners. The CTDI

vol

is the weighted average measure-

ment of the dose in a reference phantom, expressed in mGy.

REVOLUTION GE AQUILON TOSHIBA

CTDI

vol

[mGy] (Tube current [mA])

High 13.52 (160) 16.50 (300)

Standard 6.76 (80) 8.30 (150)

Reduced 4.06 (50) 5.00 (90)

Low 2.03 (25) 2.49 (45)

DATA ACQUISITION

Tube potential (kVp) 120 120

Pitch 0.984 0.938

IMAGE RECONSTRUCTION

Display ﬁeld of view (mm) 210 219

Pixel Spacing (mm) 0.406 0.427

Slice thickness (mm) 1.25 1.00

Kernel LUNG FC56

Reconstruction algorithm FBP, ASIR FBP, AIDR 3D

Iterative level 0%, 10%, 40%, 70% 0%, mild, standard, strong

3.2 Characterization of CT Protocols in

Terms of Image Quality

In order to characterize the CT protocols in terms of

image quality, we ﬁrst plotted the mean detectabil-

ity index value, computed on the three identical ac-

quisitions, as a function of the CTDI

vol

, highlighting

the different combinations of the changed parameters

(scanner, dose level, and IR blending level). Then we

also visualized how the d’, i.e., image quality varies

with the CTDI

vol

alone and, by means of a color map,

how it varies simultaneously with the CTDI

vol

and the

IR level, for each scanner separately.

3.3 Image Quality with and Without the

Eye Filter

To make the d’computation more automatic and have

full control over each step, we have also developed a

Python script that takes as input the CT image, the co-

ordinate of the center and the radius of the ROIs for

the TTF computation and automatically returns the

detectability index value, besides the 2D TTF and the

2D NPS. The Python code will be made available af-

ter the study is completed. Since there is no available

documentation on the mathematical form of the eye

ﬁlters available in imQuest, we adopted the following

deﬁnition (Solomon et al., 2015):

E(ρ) = |ηρ

1.5

· e

−0.98·ρ

0.68

(2)

We computed the d’ values for the images ac-

quired on the phantom also with this automated pro-

gram, obtaining the d

′

NPWE

with the previous equation

(1), but considering this deﬁnition of the eye ﬁlter.

However, it is also possible to deﬁne the de-

tectability index without the addition of the eye ﬁlter,

with the deﬁnition 3. In this case, we quantify a sort

of ”objective” image quality not related to the human

vision perception.

′

NPW

R R

T T F

dudv

R R

T T F

NPS

dudv

(3)

We leveraged the automated code to re-compute

this detectability index without adding the eye ﬁlter.

We ﬁnally compared the dependence of d’ on

the acquisition/reconstruction parameters in the two

cases, when image quality comprises how the image

appears to the human visual system and when it is not

based on visual perception.

4 RESULTS

In Figure 3, the plots of the mean detectability in-

dex value, computed on the three identical acquisi-

tions, referred to both the polystyrene and air inserts,

as a function of the CTDI

vol

are shown. Howevere,

the three parameters that we changed in the acquisi-

tion/reconstruction phase, scanner, CTDI

vol

and iter-

ative blending level, were highlighted for each point.

As evident from the plots, the range assumed by

the d’ value varies. This was expected given the dif-

ferent contrast of the two inserts. It is also possible to

identify in the plots different protocols, i.e., different

combinations of scanner-dose-iterative level param-

eters that produce a similar value of the detectabil-

BIOIMAGING 2024 - 11th International Conference on Bioimaging

292

Figure 3: Plots of the mean detectability index as a function of the CTDI

vol

for both the polystyrene and air inserts. The

circles are used for the GE scanner and the stars for the Toshiba scanner. In the legend, the colors are corresponding to the

iterative levels for the two scanners.

Table 2: Set of acquisition/reconstruction parameters corre-

sponding to the protocols providing a similar detectability

index. STD:STANDARD.

Scanner CTDI

vol

Iterative level

GE 4 ASIR 70%

GE 7 ASIR 0%

GE 7 ASIR 10%

GE 7 ASIR 40%

GE 7 ASIR 70%

GE 14 ASIR 0%

GE 14 ASIR 10%

Toshiba 2 AIDR3D MILD

Toshiba 2 AIDR3D STD

Toshiba 2 AIDR3D STRONG

Toshiba 5 FBP

Toshiba 5 AIDR3D MILD

Toshiba 5 AIDR3D STD

Toshiba 5 AIDR3D STRONG

ity index, meaning that they produce a similar image

quality. This was equally observed for the d’ from the

other ﬁve inserts in the CTP401 module of the phan-

tom. Hence this observation is not restricted to a spe-

ciﬁc insert or material, but having a similar trend, it

is generalizable to all inserts under examination. It

is interesting to observe that these ”equivalent” proto-

cols reported in Table 2 are heterogeneous in terms of

scanner, CTDI

vol

and iterative levels.

We visually inspected the images obtained with

protocols that produce a similar d’ and we compared

them to images obtained with protocols that produce

a very different d’, always extrapolated from the plots

in Figure 3. It was observed that a similarity in image

appearance can be perceived for the protocols cor-

responding to a similar detectability index. Instead,

when we inspect images obtained with protocols pro-

ducing a different d’, they present a quite different

appearance, especially in terms of noise. Therefore,

the detectability index could be a reasonable index to

”quantify” image quality.

As for the plot of d’ as a function of the dose alone

and the colormaps of d’ as a function of both dose and

iterative level, they are represented for the polystyrene

insert in Figure 4, for both GE and Toshiba scanners.

In Figure 5, the same plot of Figure 3 for the

polystyrene insert is represented, with the difference

that the d’ value has been obtained with the automated

program, and thus by using the eye ﬁlter deﬁned in

(2). Therefore, the trend is the same but the range of

values assumed by d’ is different (also because the le-

sion diameter simulated for the task function W was

set to 12 mm in the script and is different from the

one of 5 mm set on imQuest). It can be compared to

the plot obtained for the detectability index as deﬁned

by equation (3), namely without considering the eye

ﬁlter, shown in Figure 6.

From these two graphs, it can be observed that in

both cases, the dependence of d’ on the protocol is

similar. Also, in these cases, some protocols that pro-

duce a similar detectability index and thus a similar

image quality can be identiﬁed. However, when we

do not consider the eye ﬁlter it is evident that by in-

creasing the iterative level, there is not an evident im-

provement in image quality as in the case in which

we consider the eye ﬁlter. This is particularly true for

the ASIR algorithm in the GE scanner. It seems that

when we do not evaluate the quality based on visual

perception, the image quality at lower doses contin-

ues to improve as ASIR increases. Still, it appears to

worsen at higher doses.

Characterization and Quantiﬁcation of Image Quality in CT Imaging Systems: A Phantom Study

293

Figure 4: Upper row: Mean detectability index of polystyrene insert as a function of the dose level. Lower row: color maps

of the mean detectability index of polystyrene insert as a function of both dose level and iterative level. The color intensity

corresponds to the d’ value.

Figure 5: Plot of the mean detectability index as a function of the protocol for polystyrene. d’ values are output from the

Python script by using the deﬁnition in 1 and the expression 2 for the eye ﬁlter.

BIOIMAGING 2024 - 11th International Conference on Bioimaging

294

Figure 6: Plot of the mean detectability index as a function of the protocol for polystyrene. d’ values are output from the

Python script by using the deﬁnition in 3 without the eye ﬁlter.

5 DISCUSSION

In this study, we characterized two CT scanners in

terms of image quality, which was quantiﬁed by

means of a new metric, the detectability index, more

suitable for images with iterative algorithm recon-

struction. We found that it is possible to identify

“equivalent” protocols, meaning protocols that pro-

duce a similar d’, i.e., a similar image quality. From

these “equivalent” protocols, it would be possible

to extrapolate those that correspond to a lower dose

level. In other words, it is possible to ﬁnd the opti-

mal protocol to reduce the radiation dose delivered in

CT acquisitions without degrading the imaging prop-

erties. These results conﬁrm the ones obtained in an-

other similar study (Muti et al., 2023).

The conclusions of our study also pave the way

for new perspectives in radiomics study. We have

indeed started to analyze if these “equivalent” proto-

cols can be leveraged to obtain more robust extracted

radiomics features, in order to deﬁne a new possi-

ble harmonization strategy based on matching image

quality. The analysis on using or not the eye ﬁlter

while evaluating image quality poses another impor-

tant question. In fact, the obtained results demonstrate

that a further exploration of the iterative algorithms

effect is needed. We hypothesized that the worsen-

ing of image quality at higher iterative levels while

maintaining the dose constant could be related to the

blurring that ASIR causes in the image. However, this

should be further investigated. The interesting note

is that increasing the iterative level of some iterative

algorithms like ASIR can improve the image quality

as it is perceived by human eyes. Still, it can deter-

mine a worsening when we consider a more ’objec-

tive’ quality not related to the human visual system.

This should be further investigated when the images

have to be ﬁnally provided to an AI algorithm, which

is naturally not equipped with a visual ﬁlter.

A possible limitation to the analysis presented in

this paper is that d’ is a task-based image quality met-

ric, therefore it depends on the speciﬁc task function

chosen to represent a task in clinical practice. We sim-

ulated a 5-mm diameter lesion with a speciﬁc contrast

but this is not the only one that could be considered.

Moreover, this study is preliminary as in order

to claim the consistency of our results the analysis

should be extended to other clinical centers and scan-

ners, possibly equipped with different types of iter-

ative algorithms, and exploring other protocols and

a wider range of parameters. Therefore, the future

perspectives are in parallel reﬁning the analysis by

introducing more variables and investigating a pos-

sible harmonization strategy based on this character-

ization of image quality to improve the robustness of

radiomics studies.

Characterization and Quantiﬁcation of Image Quality in CT Imaging Systems: A Phantom Study

295

ACKNOWLEDGEMENTS

This research has been developed as part of a

Ph.D. program in collaboration with the Univer-

sity of Pisa and the National Institute for Nu-

clear Physics (INFN). Research partly supported

by: Artiﬁcial Intelligence in Medicine (next AIM,

https://www.pi.infn.it/aim) project, funded by INFN-

CSN5; FAIR-AIM project funded by Tuscany Gov-

ernment (POR FSE 2014-2020); PNRR - M4C2 -

Partenariato Esteso ”FAIR - Future Artiﬁcial Intel-

ligence Research” - Spoke 8, and PNRR - M4C2

- Centro Nazionale ”ICSC – Centro Nazionale di

Ricerca in High Performance Computing, Big Data

and Quantum Computing” - Spoke 8, funded by

the European Commission under the NextGeneration

EU programme; the Italian Ministry of Health Grant

RC and 5×1000 Health Research; AIMS2-Trials,

http://aims-2-trials.eu; the European Union NextGen-

erationEU through the Italian Ministry of University

and Research under PNRR M4C2-I1.3 Project PE

00000019 ”HEAL ITALIA” to Maria Evelina Fan-

tacci and Arman Zafaranchi CUP I53C22001440006.

The views and opinions expressed are those of the au-

thors only and do not necessarily reﬂect those of the

European Union or the European Commission Nei-

ther the European Union nor the European Commis-

sion can be held responsible for them.

REFERENCES

Barca, P., Giannelli, M., Fantacci, M. E., and Caramella,

D. (2018a). Computed tomography imaging with the

adaptive statistical iterative reconstruction (asir) algo-

rithm: dependence of image quality on the blending

level of reconstruction. Australasian physical & engi-

neering sciences in medicine, 41:463–473.

Barca, P., Palmas, F., Fantacci, M. E., and Caramella, D.

(2018b). Evaluation of the adaptive statistical iter-

ative reconstruction algorithm in chest ct (computed

tomography).

Beister, M., Kolditz, D., and Kalender, W. A. (2012). It-

erative reconstruction methods in x-ray ct. Physica

medica, 28(2):94–108.

Burgess, A. (1994). Statistically deﬁned backgrounds:

performance of a modiﬁed nonprewhitening observer

model. JOSA A, 11(4):1237–1242.

Cao, C.-F., Ma, K.-L., Shan, H., Liu, T.-F., Zhao, S.-Q.,

Wan, Y., and Wang, H.-Q. (2022). Ct scans and cancer

risks: A systematic review and dose-response meta-

analysis. BMC cancer, 22(1):1238.

Caramella, D., Fantacci, M. E., Palmas, F., Barca, P., et al.

(2018). Evaluation of the adaptive statistical itera-

tive reconstruction algorithm in chest ct (computed

tomography)-a preliminary study toward its employ-

ment in low dose applications, also in conjunction

with cad (computer aided detection). In Proceedings

of the 11th International Joint Conference on Biomed-

ical Engineering Systems and Technologies-Volume 5:

AI4Health, volume 5, pages 688–694.

Dodge, C. T., Tamm, E. P., Cody, D. D., Liu, X., Jensen,

C. T., Wei, W., Kundra, V., and Rong, X. J. (2016).

Performance evaluation of iterative reconstruction al-

gorithms for achieving ct radiation dose reduction—a

phantom study. Journal of applied clinical medical

physics, 17(2):511–531.

Forte, G. C., Altmayer, S., Silva, R. F., Stefani, M. T., Liber-

mann, L. L., Cavion, C. C., Youssef, A., Forghani, R.,

King, J., Mohamed, T.-L., et al. (2022). Deep learning

algorithms for diagnosis of lung cancer: a systematic

review and meta-analysis. Cancers, 14(16):3856.

Kennedy, K., Hulbert, A., Pasquinelli, M., and Feldman,

L. E. (2022). Impact of ct screening in lung cancer:

Scientiﬁc evidence and literature review. In Seminars

in Oncology. Elsevier.

Mail, T. B. (2013). Catphan® 500 and 600 m anual. The

Phantom Laboratory.

Manickavasagam, R., Selvan, S., and Selvan, M. (2022).

Cad system for lung nodule detection using deep

learning with cnn. Medical & Biological Engineering

& Computing, 60(1):221–228.

Muti, G., Riga, S., Berta, L., Curto, D., De Mattia, C.,

Felisi, M., Rizzetto, F., Torresin, A., Vanzulli, A.,

and Colombo, P. E. (2023). Performance of three

model-based iterative reconstruction algorithms using

a ct task-based image quality metric. arXiv preprint

arXiv:2301.08691.

Raju, S., Ghosh, S., and Mehta, A. C. (2017). Chest ct

signs in pulmonary disease: a pictorial review. Chest,

151(6):1356–1374.

Samei, E., Bakalyar, D., Boedeker, K. L., Brady, S., Fan, J.,

Leng, S., Myers, K. J., Popescu, L. M., Ramirez Gi-

raldo, J. C., Ranallo, F., et al. (2019). Performance

evaluation of computed tomography systems: sum-

mary of aapm task group 233. Medical physics,

46(11):e735–e756.

Smith-Bindman, R., Wang, Y., Chu, P., Chung, R., Einstein,

A. J., Balcombe, J., Cocker, M., Das, M., Delman,

B. N., Flynn, M., et al. (2019). International variation

in radiation dose for computed tomography examina-

tions: prospective cohort study. Bmj, 364.

Solomon, J., Wilson, J., and Samei, E. (2015). Character-

istic image quality of a third generation dual-source

mdct scanner: noise, resolution, and detectability.

Medical physics, 42(8):4941–4953.

Solomon, J. B. (2018). Performance evaluation of com-

puted tomography systems. resources: Aapm tg-233.

Vennart, W. (1997). Icru report 54: Medical imaging-the as-

sessment of image quality-isbn 0-913394-53-x. april

1996, maryland, usa. Radiography, 3(3):243–244.

Yeh, D.-M., Tsai, H.-Y., Tyan, Y.-S., Chang, Y.-C., Pan, L.-

K., and Chen, T.-R. (2016). The population effective

dose of medical computed tomography examinations

in taiwan for 2013. PLoS One, 11(10):e0165526.

BIOIMAGING 2024 - 11th International Conference on Bioimaging

296