Radiation dosage and image quality with
deep learning reconstruction techniques
in CT imaging - A Scoping Review
ABSTRACT
Purpose: Deep learning reconstruction
(DLR) is a new Computed Tomography
(CT) dose reduction strategy to optimise
CT image reconstruction. We compared
the impact of DLR on dose reduction,
diagnostic accuracy and image quality
with existing reconstruction techniques
(hybrid IR, MBIR and FBP) in CT ima-ging
. Methods: A scoping review of litera-ture
between April 2016 and April 2020
was conducted, following the PRIS-MA-
ScR guidelines, with data searched
from MEDLINE/PubMed, Semantic
Scholar, Google Scholar and Springer-
Link databases. Review includes 9 artic-
les.
Results: It seems that DLR tech-niques
yield better image quality, with
reduced image noise when done with re-duced
radiation dosages as compared to
FBP, hybrid IR and MBIR techniques. The
existing clinical CT studies also showed
the superiority of using DLR techniques
4 Kliininen Radiografiatiede 2021
over other reconstruction techniques
in terms of having significant improve-ments
in both qualitative and quanti-tative
aspects of CT image quality, by
having reduced noise and improved di-agnostic
capabilities, allowing for better
lesion detections.
Conclusion: DLR in CT have demon-strated
potential in further dose man-agement,
yielding better CT image quali-
ty with reduced image noise.
Keywords: Computed tomogra-phy
(CT), Deep learning reconstruction
(DLR), Iterative reconstruction (IR)
INTRODUCTION
Computed tomography (CT) is a com-monly
used imaging modality and its
usage has been rapidly increasing due to
its strong diagnostic performance (den
Harder et al. 2015). The growing popu-larity
of CT as a diagnostic tool has raised
concern about the risk of radiation‐
related carcinogenesis, prompting re-
Pang Zhuo Wei
Student, BSc (Hons)
(Diagnostic Radiography)
Singapore Institute of Technology
1800066@sit.singaporetech.edu.sg
Lim Wei Ming Shawn
Student, BEng (Hons)
(Software Engineering)
Singapore Institute of Technology
1802643@sit.singaporetech.edu.sg
Joonas Kuru
Student, BHsc
(Radiography and Radiotherapy)
Metropolia University of Applied Sciences
joonas.kuru@metropolia.fi
Tiina Pakarinen
Student, BHsc
(Radiography and Radiotherapy)
Metropolia University of Applied Sciences
tiina.pakarinen@metropolia.fi
Eija Metsälä
Docent, Principal Lecturer
Radiographer
Metropolia University of Applied Sciences
eija.metsala@metropolia.fi
Corresponding author:
Sanna Törnroos
MHsc, Senior Lecturer
Radiographer Metropolia University
of Applied Sciences
sanna.tornroos@metropolia.fi
search in CT to focus on minimising radi-ation
doses (Power et al. 2016). Through-out
the years after the commercialised
introduction of CT for clinical use, sev-eral
approaches were made to reduce CT
radiation dose (Power et al. 2016). One
of such key methods was to determine
the appropriate usage of CT scans by de-veloping
evidence-based guidelines and
recommendations on when the benefits
of a CT examination outweighs the risks
(Furterer 2014). Other existing options
involve concentrating on the optimisa-tion
of CT technical parameters (e.g tube
current, tube voltage, pitch and recon-structed
section thickness), utilisation
of dual-energy CT and applying dose re-duction
technology (e.g. automatic tube
current modulation) (Kerl et al. 2011;
Li et al. 2011; Willemink et al. 2013).
In recent years, CT manufacturers
have started to focus on developing a
newer CT dose reduction strategy that
revolves around optimisation of CT im-age
reconstruction (Willemink et al.
link
link
link
link
link
link
