516 / 1096
S501
ESTRO 36
_______________________________________________________________________________________________
Conclusion
Automatic propagated contours for target and OARs need
manual adjustments for clinical acceptance. Mesorectum
and OARs adjustments can be made by an experienced
technologist and are not clinically relevant different from
the manual contours of an experienced radiation
oncologist. This semi-automatic contouring strategy can
be used in an online workflow for rectal boost treatment,
however further speed-optimisation is desirable.
PO-0906 Textural analysis of MR images to improve the
characterisation of recurrent prostate cancer
J. Stirling
1
, R. Alonzi
2
, P.J. Hoskin
2
, N.J. Taylor
1
, W.L.
Wong
1
, A.R. Padhani
1
, B. Sanghera
1
1
Paul Strickland Scanner Centre, Research, Northwood,
United Kingdom
2
Mount Vernon Hospital, Academic Oncology Unit,
Northwood, United Kingdom
Purpose or Objective
MR has the ability to assess numerous physiological and
biochemical tumour characteristics. Fractal analysis may
provide a better insight into the biology and behaviour of
prostate tumour than simplistic comparisons of
multiparametric data. In this pilot study, we aim to
determine whether fractal and lacunarity analysis can
characterize the properties of radio-recurrent prostate
cancer, using Apparent Diffusion Coefficient (ADC) MR
Images.
Material and Methods
Retrospective analysis of eight patients with recurrent
prostate cancer after previous radical radiotherapy (mean
age: 71.25 years), underwent MRI examination for re-
staging prior to consideration of salvage therapy. ADC
images of the prostate were manually segmented from
surrounding tissue and a region of interest (ROI) drawn to
distinguish between restrictive diffusion and non-
restrictive tissue (figure 1b). Low, medium and high ADC
value maps were generated by intensity thresholding the
respective restrictive and non-restrictive ROIs. These
were processed and converted to 8-bit black and white
images (figure 1c, low intensity in restricted diffusion) for
application to in-house textural analysis software (image
1d) to estimate (a) fractal dimension (b) fractal
abundance and (c) lacunarity
Curve1
Figure 1:
(a) an ADC image of the prostate gland (b) an
ADC image showing areas of restricted diffusion (red) and
non-restricted diffusion (blue) (c) shows a binary image
used for fractal and lacunarity analysis (d) lacunarity
curves from restricted areas (red) and non-restricted
areas
(blue)
Results
The average fractal characteristics are summarised in
table 1 with the fractal dimension between areas of
restricted diffusion and non-restrictive diffusion of the
low and medium intensity images being of significant
difference (p=0.0014 and 0.0023 respectively). The
fractal abundance of the medium intensity image between
the restricted diffusion and non-restrictive diffusion was
also significant (p=0.0012).