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

100 / 146

B.322
A NEW METHOD FOR THE ANATOMICAL DISTRICT RECOGNITION (ADR)
IN CT IMAGES
A. Vassena G. Pozzi A. Ostinell a A.O. Sant’Anna, Como, Italy; b DEIB
Politecnico di Milano, Milano, Italy
Introduction: Use of bio-medical images in clinical practice is becoming more
and more frequent and relevant. Despite standard image formats including
alpha-numerical tags, the anatomical district description may be incomplete,
missing, or even erroneous. We propose a novel approach based on CT scout
images (ADR), to identify the anatomical districts in CT examinations.
Materials and Methods: ADR exploits the Histogram of Oriented Gradi-
ents (HOG) method implemented by Matlab 2010b. The method, based on
feature detection, normalizes the image, and then splits it into cells. For
every cell, the method computes the gradient [0–179 deg.] of its pixels and
builds up the HOG for the image, obtaining a Feature Vector (FV). For every
image, we then compare that FV with the FVs of references images we pre-
viously manually classified. By looking for the minimal difference between
FVs (identified by Least Squares) we classify any image.
Results: For head, chest and abdomen districts the sensitivity results are 99.5%,
91.5%, 95.0% and 100%, 80.5%, 81.0 % for front and side scout, respectively. In
the same way, the accuracy results were 98%, 97%, 96% and 98%, 100%, 99% re-
spectively. Methods from the literature resulted in a poor sensitivity (<80% in
most cases). Computing time was also considered: for our workload of 73 CT
images/day, a 2 h 1min overall processing time is needed.
Conclusions: ADR defines a novel approach in anatomical district classi-
fication featuring an improved accuracy and sensitivity with respect to
methods from the literature. Future implementations are related to: op-
timized repository, integrating dose assessment system and automatically
running the classifier on the images stored by the PACS system.
B.323
PATIENT RADIATION DOSE IN INTERVENTIONAL CARDIOLOGY
PROCEDURES PERFORMED IN MODENA LOCAL HEALTH UNIT
G. Venturi *, F. Goldoni, D. Acchiappati. S.C Fisica Medica, AUSL Modena, Italy
Introduction: The aim of this work, which was carried on in collabora-
tion with the Medical Imaging Department and Cardiology Department of
the Nuovo Ospedale Sant’Agostino in Modena, and became part of a larger
Survey, is the dosimetric characterization of interventional cardiology pro-
cedures performed by the hemodynamic laboratory. We examined diagnostic
coronary angiography and percutaneous transluminal coronary angioplasty
procedures with the standard dosimetry indicators (DAP and fluoroscopy
time), and then evaluated the dose to the organs by using Monte Carlo simu-
lations and anthropomorphic phantomwith thermoluminescent dosimeters.
Materials and Methods: These dosimetric evaluations were part of a project
that included an educational component addressed to the x-ray techni-
cians, and were conducted by examining the activity of the hemodynamic
laboratory over the course of a year and on a sample of about 1800 pa-
tients. Information were collected about the type of procedure performed,
the composition of the team executing the procedure, the type of patient
(height and weight) and the dosimetric indicators (DAP and fluoroscopy
time). The coronary angiography procedure was evaluated by studying the
dosimetric contribution of every projection.
Results and Conclusions: The final results confirm the good level of op-
timization of the various operative protocols of the hospital and the full
respect of the Diagnostic Reference Levels, and provide a good estimation
of the dose that the patient’s organs received, even in relation to the type
or projections utilized during the procedures.
B.324
DIAGNOSTIC RADIATION DOSE IN ADULT TRAUMA PATIENTS: A SURVEY
PERFORMED AT MODENA LOCAL HEALTH UNIT
G. Venturi C. Malagol D. Acchiappati G. Bandiera L. Vallisner a S.C
Fisica Medica, AUSL Modena, Italy; b Dipartimento Di Emergenza, AUSL Modena,
Italy
Introduction: Diagnostic Imaging is essential to evaluate the conditions
of those patients who are transported to the Emergency Department fol-
lowing multiple traumas. Often combined with conventional radiology, CT
scan is certainly one of the most utilized procedures for the diagnosis. De-
pending on the type of trauma and the patient’s developing conditions,
radiological examinations must often be repeated, resulting in high radi-
ation dose for the ER patients.
The objective of this study is to identify diagnostic paths for the various
types of polytrauma and estimate for each of them the dose to the organs
for the patient.
Materials and Methods: The survey was performed on a sample of 355
patients. All of them suffered major trauma (Injury Severity Score > 15) and
where brought to the Emergency Department of Sant’Agostino Estense Hos-
pital in 2012.
A Regional database with data pertaining to the activation of the trauma
team provided the patients’ anagraphical information and gave some useful
information about the type of injury suffered by the patient.
Many different typologies of trauma were identified (the principal ones are
cranial, thoracic, abdominal, arms and legs) alongside with related diag-
nostic path. It was identified for each type of trauma object of study a patient
who has been subjected to a median number of RX and TC examinations.
The dosimetric evaluations were performed using the Monte Carlo soft-
ware PCXMC and ImPACT.
Results and Conclusions: Notwithstanding the possibly high exposure to
ionizing radiations, diagnostic imaging is certainly justifiable given the life
threatening nature of polytraumas. However, an accurate dosimetric eval-
uation for each diagnostic path is an important piece of information for
the doctors of Emergency Department, as well as for the communication
of the dosimetric data to patients and them caregivers.
B.325
EFFECT OF DIFFERENT FRAME RATES ON LOW CONTRAST DETECTABILITY
IN DIGITAL ANGIOGRAPHY
R. Villa C. Spadavecchi A. Radic C. Pasquali N. Paruccini
A. Crespi a Scuola di Specializzazione in Fisica Medica, UNIMI, Milano, Italy;
b A.O. San Gerardo, S.C. Fisica Sanitaria, Monza, Italy
Introduction: digital angiography systems produce different kinds of images:
fluoroscopic images obtained with low mA values, coronary angiography
images acquired with very high mA values besides very short times and
DSA images obtained through image subtraction. All these images are
dynamic ones and have to be analysed in motion. The analysis of low con-
trast detail characteristics of the above-mentioned images achieved with
standard phantom is difficult and time-consuming.
Materials and Methods: LCD was estimated applying Chao’s statistical
method and using a dedicated phantom. LCD values were calculated frame
by frame. Therefore we arranged a 4AFC experiment adding low contrast
details to a uniform region of fluoroscopy and fluorography runs acquired.
Results: details detectability depended on the viewing frame rate: higher
rates showed better details detectability.
Contrast detail curves derived from perception experiment were com-
pared with LCD curve calculated on a single frame and on images obtained
as a mean of different numbers of frames. Our results agreed with the as-
sumption that the time of integration of noise in human retina is 200 ms.
For example contrast detail values obtained with a viewing frame rate of
15 fps were in agreement with LCD calculated on an image which was a
mean of three different frames.
Conclusions: the statistical approach proved to be less time-consuming,
independent on inter- and intra-observers variability and repeatable, thus
being a valid method for image quality (or LCD) assessment and is cur-
rently adopted in image quality assurance protocols.
e95 Abstracts/Physica Medica 32 (2016) e71–e96