Copenhagen, Denmark
Onsite/Online

ESTRO 2022

Session Item

CNS
6002
Poster (digital)
Clinical
MRI texture analysis of radionecrosis and isodoses levels after radiosurgery in brain metastases
Sandra Fernandez Alonso, Spain
PO-1162

Abstract

MRI texture analysis of radionecrosis and isodoses levels after radiosurgery in brain metastases
Authors:

Sandra Fernandez Alonso1, Isabel Casado Ruiz2, Sandra Guardado Gonzales1, Gustavo Pozo3, Rafael D´Ambrosi1, Pedro Adaimi3, Eduardo Cabello3, M. Angeles Perez Escutia1, Jose Fermin Perez-Regadera1

1Hospital Universitario 12 de Octubre, Radiation Oncology, Madrid, Spain; 2Universidad Politécnica de Madrid, Photonic Technology and Bioengineering, Madrid, Spain; 3Hospital Universitario 12 de Octubre, Radiophysics, Madrid, Spain

Show Affiliations
Purpose or Objective

To quantitatively analyze the relationship between the appearance of radionecrosis (RN) and isodoses levels, comparing textural features between these regions.

Material and Methods
From a total of 28 cases of RN in 286 patients treated in our center with radiosurgery (doses delivered interval 15-18 Gy) or with fractionated stereotactic radiotherapy (30-35 Gy in 6 or 7 fractions) between 2011 and 2020, the MRI images of T1 and T2 with contrast of 9 selected patients were analyzed

The regions of interest (ROIs) analyzed were: GTV, RN and isodoses levels of 110%, 107%, 100% and 50% of total dose.

RT treatment planning was performed using ECLIPSE planner (Varian®) and Radiomics, an extension of 3D Slicer was used to extract radiomic features of ROIs (3D Slicer image computing platform®). 
Statistical analysis was performed using SPSS (IBM®). 
The selected Bin Width value was 15.

We extracted features from five feature classes: Shape, First Order, GLCM, GLSZM y GLRLM.

Of the RN structure, 14 characteristics of Shape, 17 of First Order, 24 of GLCM, 16 of GLSZM and 16 of GLRLM were analyzed and compared with the same characteristics of the isodoses levels described.

 

To compare textural differences, two analyzes were carried out: the sign test and the Pearson correlation coefficient
In order to establish a relationship between the structure of RN and isodoses levels, two conditions had to be met: that there were no significant differences in the characteristics of both regions (p<0.05) and that the correlation was close to 1 or -1.

Results
The 110% and 107% isodoses levels had greater number of characteristics without significant differences with RN, being the ones that were most related to it (Table 1A).
Regarding the level of correlation, there were a greater number of correlated characteristics between the 110% isodose level and RN (Table1B).

According to the concordance of characteristics without significant differences and adequate level of correlation (Table 2), the isodoses levels with a greater relationship with RN were 110%, which is the one that receives the highest radiation dose, and 50%. These are followed by the 100% and finally the 107% isodoses levels.




Conclusion
RN is closely related to the 110% isodose level.
It is also related to the 50% isodose level, so these regions should be kept away from the organs at risk.
Besides, it is related to the initial lesion, which implies that certain properties of RN can be predicted thanks to the textural analysis of the metastatic lesion.
Finally, repeatability of radiomics features could provide good differentiation between RN and tumor progression.