ESTRO 36 Abstract Book

S308

ESTRO 36 2017

_______________________________________________________________________________________________

Since the MR-Linac aims for online adaptive radiotherapy,

online plan QA needs to be developed additional to the

offline QA procedures and devices. At the MRI-linac, an

IMRT plan will be created online based on the daily

anatomy while the patient is on the treatment couch.

Therefore, individual plan QA via measurements can’t be

performed. As a solution, the use of an independent dose

calculation based on 3D collapsed cone dose calculation

algorithm was investigated. Although the effect of the

magnetic field is not directly incorporated in this

algorithm it seems to be sufficient to be used for online

QA of the online generated IMRT plan.

Symposium: Novel approaches for combining imaging

and non-imaging data for radiotherapy response

predicition

SP-0595 Modeling the interplay among volume,

vascularization and radio-sensitivity in cervical cancer

exploiting 3D-Doppler data

A. Belfatto

, D. Ciardo

, A.M. Vidal Urbinati

, F. Cattani

R. Lazzari

, B.A. Jereczek-Fossa

, D. Franchi

, R.

Orecchia

, G. Baroni

, P. Cerveri

Politecnico di Milano University, DEIB, Milano, Italy

European Institute of Oncology, Unit of Medical Physics,

Milano, Italy

European Institute of Oncology, Division of Preventive

Gynecology, Milano, Italy

European Institute of Oncology, Division of

Radiotherapy, Milano, Italy

European Institute of Oncology, Scientific Director,

Milano, Italy

The role of oxygenation in tumor radio-sensitivity

The cell damage due to radiation exposure is conveyed by

means of both direct effects, and indirect ionization of

DNA due to free-radicals (e.g. O-). As a consequence,

tumor oxygenation affects the response to radiotherapy by

enhancing the radiosensitivity, conversely, hypoxic tumors

are likely to be more radio-resistant. Moreover, literature

findings suggest a systemic repercussion of changes in

oxygen levels in the lesion area, for example hypoxia may

reduce the tumor metabolism inducing a quiescent status

and preventing the mass development. Mathematical

models of tumor growth and response to irradiation are

able to mimic the interplay among oxygenation, volume

and radiosensitivity addressing the problem as a whole.

Tumor data (e.g. volume, PO

, necrotic fraction) are

mandatory for an accurate and patient-specific model

setting.

Current methods to assess tumor oxygenation

This lecture does not aim to provide an exhaustive

overview of the methods currently available for the

evaluation of tumor oxygenation, however a brief

description of the most used techniques is in order. The

Eppendorf probes are still considered the gold standard in

the assessment of oxygenation and consist in a direct

measure of PO

. Despite their accuracy, the electrodes are

not always applicable in clinics due to their intrinsic

invasiveness and the local nature of the measurements.

Alternative methods require the use of radioactive tracers

and PET scans or functional MRI (e.g. interleaved BOLD-

TOLD protocols). Another option is the 3D Doppler

ultrasonography, which is inexpensive and non-invasive,

although some restrictions reduces the extent of its

application.

Focus on the US Doppler advantages and limitations

The 3D Doppler technique allows the evaluation of the

hemodynamics by exploiting the Doppler effect. The

regions containing moving particles will be shown in false

colors according to their velocity. Three main indices are

provided, namely the vascularization index (VI), the flow

index (FI) and the vascularization flow index (VFI). The VI

is the percentage of color pixels in the region of interest

and represents the amount of vasculature, while FI is the

average (%) color-value of the pixels which were

recognized as vessels, indicating the mean flow. The

rescaled product of the above equals to the

vascularization

flow

index

(0<VFI<100).

The

vascularization/flow indices are not able to completely

define the tissue oxygenation, especially for tumors. This

is due to the chaotic development of vasculature in case

of rapid angiogenesis promoted by the cancer

uncontrolled growth. The vessels are likely to be leaking

and immature supplying a reduced amount of oxygen to

the cells they drain. The gap between vascularization and

oxygenation is also due the lack of knowledge about vessel

permeability.

Example of volume and oxygen-related data integration

in tumor evolution models.

We implemented a macroscale model of tumor

development encompassing the tumor evolution where

viable and necrotic portion dynamics evolve separately,

along with oxygenation changes. The model was set

according to the clinical data (volume) of 7 cervical cancer

patients undergoing external beam radiotherapy (EBRT)

subdivided into 28 fraction (1.8Gy/fraction). For each

ptient, two expert clinicians performed and manually

contoured five Doppler ultrasound tests before, during and

after the treatment. The software 4D View® (General

Electric Company - Fairfield, Connecticut, United States)

was used to compute the tumor volume and the above

mentioned vascularization/flow indices (VI, FI, VFI) in the

identified tumor region (

Fig.1

Our algorithm was trained to provide the lowest fitting

error on the tumor size evolution, while the oxygenation

was allowed to adjust itself feely to cope with the

predicted volume. The model-based curves of tumor

evolution resulted in over 90% accuracy with respect to

the measured volumes. Using the Pearson coefficient (r),

the correlation between VI, FI, VFI and the model

predicted oxygenation was analyzed. In three cases r 0.9

(strong correlation), while two patients showed an

average r value (0.5-0.7). For the remaining two patients,

who presented r<0.5, low indices values (expecially VFI)

and the smallest tumor shrinkage at the end of EBRT

(according to data), the model predicted poor average

oxygenation.

In conclusion, the relation included in the model to mimic

the interplay between tumor volume and oxygenation

dynamics was sufficient to provide oxygenation trends

correlating to Doppler indices (r>0.5) in five cases out of

seven. Although non-responsive tumors were more critical

to mimic, the model was able to identify their hypoxic

status.