AUTHOR=Puttanawarut Chanon , Sirirutbunkajorn Nat , Tawong Narisara , Khachonkham Suphalak , Pattaranutaporn Poompis , Wongsawat Yodchanan 

TITLE=Impact of Interfractional Error on Dosiomic Features

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.726896

DOI=10.3389/fonc.2022.726896

ISSN=2234-943X

ABSTRACT=Objectives: The purpose of this study was to investigate the stability of dosiomics features under random interfractional error.
Material and Methods: The isocenters of the treatment plans of 15 lung cancer patients were translated by a maximum of ±3 mm in each axis with a mean of (0, 0, 0) and a standard deviation of (1.2, 1.2, 1.2) mm in the x, y and z directions for each fraction. A total of 81 dose distributions for each patient were then calculated considering 4 fraction number groups (2, 10, 20, and 30). A total of 93 dosiomic features were extracted from each dose distribution in 4 different regions of interest (ROIs): gross tumor volume (GTV), planning target volume (PTV), heart and both lungs. The stability of dosiomic features was analyzed for each fraction number group by the coefficient of variation (CV) and intraclass correlation coefficient (ICC). The agreements in the means of dosiomic features among the 4 fraction number groups were tested by ICC. The percent differences (PD) between the dosiomic features extracted from the original dose distribution and the dosiomic features extracted from the dose distribution with interfractional error were calculated.
Results: Eleven out of 93 dosiomic features demonstrated a large CV (CV ≥ 20%). Overall CV values were highest in GTV ROIs and lowest in lung ROIs. The stability of dosiomic features decreased as the total number of fractions decreased. The ICC results showed that 5 out of 93 dosiomic features had an ICC lower than 0.75, which indicates intermediate or poor stability under interfractional error. The mean dosiomic feature values were shown to be consistent with different numbers of fractions (ICC ≥ 0.9). Some of the dosiomic features had PD greater than 50% and showed different PD values with different numbers of fractions.
Conclusion: Some dosiomic features have low stability under interfractional error. The stability and values of the dosiomic features were affected by the total number of fractions. The effect of interfractional error on dosiomic features should be considered in further studies regarding dosiomics for reproducible results.