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

136 / 146

interest, wherein summary statistics of all the parameters were derived.
Cross-correlations between all the variables were studied using the Spear-
man rank test.
Results: RSS maps demonstrated the goodness of the fit quality by the DKI
model for each patient. Increased values of K were found in all patients,
indicating a non-Gaussianity of diffusion processes in tumors. Dk and ADC
were strongly related, while there was a low correlation between Dk and
K, suggesting that K may provide additional information about the tumor.
An inverse relationship emerged between K and D, indicating that both pa-
rameters may evidence the possible interactions of water molecules inside
the lesion with underlying cell membranes.
Conclusions: Our preliminary data from HN and brain tumors show the
feasibility of DKI that may lead to additional information for a more com-
prehensive tumor characterization.
E.443
IMPROVING DIFFERENTIAL DIAGNOSIS OF BREAST LESIONS WITH ADC
AND DCE KINETIC DESCRIPTORS
L. Moro *, E. Sardi, G. Bertoli, I. Carne, P. Caprotti, M. Baldi. Fondazione
Salvatore Maugeri, Pavia, Italy
Introduction: The aim of this work is to evaluate the usefulness of diffusion-
weighted sequences in the differentiation of benign and malignant breast
diseases using magnetic resonance imaging (MRI) at 3 T.
Materials and Methods: 343 patients who underwent routine MRI ex-
aminations on a GE Discovery MR750, including dynamic contrast enhanced
(DCE) and diffusion-weighted imaging (DWI) scanning for breast cancer,
between January 2013 and April 2015, were considered. The study in-
cluded 65 female patients (average age, 43 years; age range, 32–82 years),
corresponding to 67 known breast lesions which had been diagnosed by
histological examination. Apparent diffusion coefficient (ADC) and kinetic
parameters (uptake phase and delayed enhancement characteristics) were
evaluated using a dedicated 8-channel breast coil.
Results: Mean ADC value for malignant lesions was 0.92 × 10−3 mm2/s (ADC
range, 1.80–0.41 × 10−3 mm2/s), for benign lesion was 1.11 × 10−3 mm2/s (ADC
range, 2.10–0.50 × 10−3 mm2/s): benign lesions displayed higher ADCs than
malignant ones (p < 0.05). Fixing a threshold ADC value at 1.0 × 10−3 mm2/s,
10 false negatives and 7 false positives were found. Kinetic parameters ex-
hibited a predominance of rapid uptake and washout curve in invasive ductal
carcinoma, medium/rapid uptake and plateau curve in ductal carcinoma
in situ, slow/rapid uptake and washout curve in invasive lobular carcino-
ma, and rapid uptake and washout/plateau curve in lobular carcinoma in
situ.
Conclusions: According to the available data, the diffusion coefficient has
correctly evaluated the benign or malignant nature of the lesion in 50 cases
out of 67 (74.6%), providing useful information to the diagnosis. Kinetic pa-
rameters were consistent to histological data only in 49.2% of cases,
indicating a lower reliability.
E.444
LOSS OF REPRODUCIBILITY IN A TASK BRAIN
M. Muti K. Sammartano G. Cestellini A. Direnz I. Aprile a S.Maria
Hospital, Terni, Italy; b School for Radiology’s Technicians, Rieti, Italy; c Bietti
Foundation, Rome, Italy
Introduction: We study the brain during a repetitive simple motor task
and, in a brain’s black box approach, we consider the reproducibility lost
in this process between input and output.
Materials and Methods: Each of eighteen right-handers healthy subjects
was examined on 3 Tesla tomograph (Siemens Magnetom Verio) through
acquisition of a standard functional magnetic resonance (fmri) motor task
protocol, repeated four times. An spm8 standard post processing analysis
was done and were obtained cortex activation maps. The visually stimu-
lated motor task, in a (30’ON-OFF)×4 fmri protocol, involves the alternate
compression of two buttons on a joystick with thumb and forefinger; the
frequency and number of compressions were electronically recorded and
subjects’ input-reproducibility evaluates by their coefficient of variations
(CV). The subjects’ output-reproducibility was evaluated in the L motor and
R + L visual cortex maps with the CV of mean distance between points of
maximum activation in the repetitive tasks. The ratio output/input mean
reproducibility expresses the loss of reproducibility.
Results: The mean (median/min–max) CV% of frequency is 11.6 (7/2–75),
n° of compressions 18.1 (14/2–52), distance of maximum activation on L
motor cortex 55.6 (59/17–100), on L visual cortex 53.9 (50/22–89), on R
visual cortex 54.4 (55/17–84); the correspondent loss of reproducibility is,
with respect to the frequency: 4.8, 4.7,4.7 times and respect to the n° of
compressions : 3.1,3,3 times.
Conclusions: In a linear static or dynamic system there is an input–
output reproducibility’s conservation. In the brain system there is a loss
of reproducibility and seems dependent by variable but not by the brain’s
area used for its measure. In the brain the evidence of a nonlinear input–
output process system, even with respect to a simple motor task, could be
due to the low frequency resonant rest networks noise font.
E.445
BRAIN SIZE BIAS CORRECTION METHOD IN A VOLUMETRIC GROUP
ANALYSIS: THEORY AND NORMALIZATION
M. Muti A. Direnzo S. Caproni M. Princip C. Piccolini a S.Maria
Hospital, Terni, Italy; b Bietti Foundation, Rome, Italy
Introduction: We propose a brain size bias correction method for mag-
netic resonance images volumetric studies based on a theoric relation and
proved in a group comparison analysis.
Materials and Methods: 19 dyslexics (assessed with dyslexia test Sartori
DDE -II after evaluating cognitive IQ > 85 with scale Wechsler WISC -III)
and 10 controls aged matched were examined on 3 Tesla tomograph
(Siemens Magnetom Verio) through acquisition of high resolution and high
contrast gray-white matter T1-dependent MR images. The images were then
processed with Freesurfer (Martinos Center for Biomedical Imaging) for the
automatic segmentation of cortical and subcortical areas with final assess-
ment of their volume. With an interpretative model of brain cortex based
on the geometry of a hollow sphere, we describe the volume of each cor-
tical area as linearly dependent by three variables: surface (by solid angle)
and relative thickness of the area and volume of whole brain subjects. In
the statistical analysis are selected sixteen pathology linked areas. For each
subject, the volume of each mean cortical area has been evaluated both
as such and as a ratio with its intracranial volume (normalization method).
The effect of normalization is evaluated considering the decreasing of the
p-values in a T-test group statistical analysis in a with vs without uses of
method comparison.
Results: The normalization produces: an annulment of brain size bias effect;
a decrease of p-values in the 81% of dyslexics vs Controls volume area T-test
in which the mean p-values variation is −55%; and highlights significant
difference in the amygdala volume (p = 0.03) and a marginal significant dif-
ference in Temporal Planum volume (p = 0.07) between two groups, was
not highlighted before (volume corrected VBM or Freesurfer standard group
comparison).
Conclusions: The proposed method can drive the volumetric analysis in
the comparison between groups and produces significant results better than
the standard methods.
E.446
COMPARISON OF PCA VS KPCA FOR PHYSIOLOGICAL NOISE REMOVAL IN
RESTING STATE FMRI
F. Pennarola M. Fanfoni V. Cannata’ B. Bernardi A. Napolitano
a Medical Physics Department, Enterprise Risk Management, Bambino Gesù
Children’s Hospital, Rome, Italy; b Physics Department, Universtiy of Rome ‘Tor
Vergata’, Rome, Italy; c Unit of Neuroradiology, Bambino Gesù Children’s
Hospital, Rome, Italy
Introduction: Functional magnetic resonance imaging (fMRI) is a versa-
tile technique to study brain activation by exploiting the BOLD contrast and
there is a relative recent use of this method to monitor resting state brain
activity. Because of the lower SNR in resting state signal compared to task
e131 Abstracts/Physica Medica 32 (2016) e124–e134