Physica Medica: European Journal of Medical Physics - Volume 32 - Supplement 1 - February 2016

90 / 146

Careggi, Firenze, Italy; c Head of Medical Imaging Department, AOU Careggi,
Firenze, Italy
Introduction: The dramatically increased use of CT has driven the tech-
nological progress to develop low dose solutions while preserving the
diagnostic-quality. In order to validate dose reduction claims, a standard-
izable method of assessing the quality of quantum-limited CT images would
be needed. In this work we present a new method for an automatic and
objective evaluation of CT image quality, which can be easily applied for
optimization of clinical protocol.
Materials and Methods: The method comprises a CHO model observer ad-
justed to predict as close as possible the performances of the human
observer, a phantom, and a procedure for simple generation of the ensem-
ble of images statistically required for quality evaluation. The image
processing and data analysis are managed by a homemade software written
in MatLab®. The task consists of the classification of images that may or
may not contain a low contrast detail in an unknown location. The model
observer includes a searching process to locate the signal and its overall
performances are measured by means of the area under the curve (AUC)
of localization ROC (LROC) analysis.
Results: The implemented CHO observer resulted to be in good agree-
ment with the human observer. The performances on iterative reconstructed
(IR) images appeared slightly better than on backprojection (BP) images
relatively to the low contrast detection task here proposed. The CHO was
applied to the optimization of a protocol consisting of repeated acquisi-
tions with different slice thicknesses. The CHO recognized the possibility
of reducing the exposure for the images with larger slice thickness, while
preserving the quality, but the amount of the reduction is not linear with
the thickness as expected if only noise is considered.
Conclusions: The method revealed to be very practical and easy to use.
The obtained results agree with the perception of the clinicians and thus
seem valuable for protocol optimization.
B.287
SYNTHETIC 2D RECONSTRUCTIONS OF DIGITAL BREAST TOMOSYNTHESIS:
AN ATTEMPT OF OBJECTIVE IMAGE QUALITY ASSESSMENT
N. Oberhofer *. Azienda Sanitaria dell’Alto Adige, Bolzano, Italy
Purpose: In recent years, several Digital Breast Tomosynthesis (DBT) systems
are proposing synthetic 2D mammograms, reconstructed from DBT pro-
jections. This should permit to skip the traditional FFDM acquisitions, in
order to keep patient dose low. This study investigates the possibility of
using objective image quality criteria known from FFDM for image quality
assessment of synthetic 2D images.
Methods and Materials: Two objective metrics have been used: (a) signal
difference-to-noise-ratio (SDNR) for a 0.1 mm Al detail embedded in 2 cm–
7 cm homogenous PMMA slabs and (b) Image Quality Figure inverse (IQFinv)
from automated evaluation of eight scans of the CDMAM phantom (phantom
ver. 3.4, software Cdmam Analyser, both Artinis, The Netherlands), sand-
wiched between 2 PMMA slabs of 2 cm each. A system of one vendor (Fuji
Innovality) has been studied in detail. Two different acquisition geom-
etries for the DBT modality (small angle 15°, large angle 40°), with two
different detector readout handling possibilities each (high resolution, normal
resolution), have been investigated. SDNR was evaluated from single DBT
scans with automatic exposure control active; for CDMAM evaluation, ex-
posure was varied to cover 0.7 mGy–3.8 mGy average glandular dose.
Results: With respect to dose variations, SDNR of synthetic 2D images re-
flects the behavior known from FFDM, increasing with dose, although values
are lower and may be not comparable. Synthetic images scored signifi-
cantly lower in IQFinv compared to FFDM images of the same dose level.
Nevertheless the metrics proved to be quite sensitive to dose variations, too.
Conclusion: SDNR and IQFinv measured on synthetic 2D images have both
proven to be quantities sensitive to variations in exposure settings. The re-
sulting values are lower than in 2D. Their suitability as a metrics for exposure
optimization in DBT has to be further investigated.
Note: A system from a different vendor (Hologic, Dimensions) is under
testing.
B.288
A NEW METHOD FOR THE DIRECT EVALUATION OF DOSIMETRIC
PARAMETERS FOR PEDIATRIC PATIENTS IN HIGH PITCH CT SCANS
OVERCOMING AAPM RTP111
C. Orlandi E. Genovese S. Donatiell P. Toma’ V. Cannata’ a Enterprise
Risk Management/Medical Physics, Bambino Gesù Children’s Hospital, IRCCS,
Rome, Italy; b Department of Imaging, Bambino Gesù Children’s Hospital, IRCCS,
Rome, Italy
Introduction: The increasing complexity of new technologies concerning
CT scanners imposes the redefinition of the standard dosimetric param-
eters in order to obtain an accurate characterization in terms of radiation
dose. The use of high pitch, or the presence of overbeaming and overscanning
in helical scans, highlights the need to introduce a novel CTDI metric. An
innovative method for dosimetric parameters determination in latest gen-
eration CT scanners is proposed, with the aim to achieve a comprehensive
dose evaluation.
Methods and Materials: Dosimetric parameters (CTDIvol and DLP) deter-
mination for high pitch helical scans is performed on a dual source multislice
CT (Siemens), with a standard pencil ion chamber (Radcal10X6-3CT) in-
serted in PMMA Body and Head phantoms. The use of an innovative
experimental setup permits to change reference system and thus measure
dosimetric parameters without the dedicated phantoms and ion chamber
described in AAPM Report n.111. The method is validated for different
helical scans: dosimetric parameters are obtained analyzing the dose
rate as a function of time and extrapolating the contribution of a single
rotation for CTDIvol, while the whole dose integral is considered for
DLP.
Results: CTDIvol for high pitch helical scans evaluated with the proposed
method is comparable with the one obtained from the corresponding se-
quential scan (differences below 5%). In addition, DLP represents a precise
estimation of radiation dose, including the entire contribution of
overscanning and overbeaming dose.
Conclusion: The proposed method is powerful to evaluate the total radi-
ation dose for latest generation CTs, without needing dedicated or bulky
instrumentation. The accurate determination of radiation dose is partic-
ularly important for pediatric patients, to take into account the real
exposition and avoid dose underestimations that can reach 30%
References
[1]
[2]
B.289
RADIATION DOSE TRACKING: EXPERIENCE IN CLINICAL MAMMOGRAPHY
O. Ortenzia C. D’Aloia G. Piccinini C. Ghetti a Servizio di Fisica
Sanitaria, AOU Parma, Parma, Italy; b Centro Senologico, AOU Parma, Parma,
Italy
Introduction: The aim of this study was to outline the status of radiation
exposure in mammographic clinical imaging. We evaluated mammo-
graphic dosimetry indexes and improved clinical procedures using a radiation
dose tracking system in order to face the requests of the new Euratom Di-
rective 2013/59.
Materials and Methods: Since the beginning of 2015 at AOU of Parma we
started using Radimetrics (Bayer) dose tracking software connected to GE
Senographe Essential and GE Senographe DS digital mammography systems.
We have evaluated the tube current, peak kilovoltage, breast thickness and
compression force and also the Average Glandular Dose (AGD) to the breast
during clinical mammography.
Results: We have collected the data from February to August 2015 and pro-
cessed 15,875 radiograms equal to 3627 mammographic examinations (2956
bilateral and 671 monolateral).
The 70% of exams were performed on GE Senographe DS, the remaining
on GE Senographe Essential. The mean age of the patients was 57 years
old and mean compression force was 184 N. The mean AGD for each ra-
diogram calculated for both mammography systems was 1.15 mGy and the
75th percentile was 1.33 mGy.
The third-quartile of total AGD was 6.2 mGy and third-quartile of acqui-
sition counts was 5 images for GE Senographe Essential. The third-quartile
e85 Abstracts/Physica Medica 32 (2016) e71–e96