A Voxel-Based Approach to Explore Local Dose Differences Associated With Radiation-Induced Lung Damage
Giuseppe Palma1, Serena Monti2, Vittoria D'Avino1
1Institute of Biostructure and Bioimaging, National Research Council, Naples, Italy.
International Journal of Radiation Oncology, Biology, Physics
|August 12, 2016
Summary
This study used a voxel-based approach to analyze radiation dose differences in Hodgkin lymphoma survivors, finding higher doses in specific lung regions correlated with radiation-induced lung damage (RILD). This helps pinpoint dose-response relationships for late RILD.
Area of Science:
- Radiation Oncology
- Medical Physics
- Radiotherapy
Background:
- Late radiation-induced lung damage (RILD) is a significant concern for cancer survivors treated with radiation therapy.
- Understanding the relationship between radiation dose distribution and RILD is crucial for optimizing treatment plans and minimizing toxicity.
- Previous studies often analyzed dose-response relationships using whole-lung metrics, potentially overlooking localized effects.
Purpose of the Study:
- To apply a voxel-based (VB) approach to identify local radiation dose differences associated with late RILD in Hodgkin lymphoma (HL) survivors.
- To explore the spatial distribution of radiation dose in relation to the occurrence of late RILD.
- To develop and validate a framework for dose-response modeling in radiotherapy.
Main Methods:
- A cohort of 98 HL survivors treated with supradiaphragmatic radiation therapy was analyzed.
- Patient CT scans were normalized to a common coordinate system using a log-diffeomorphic approach for accurate dose mapping.
- Nonparametric multiple permutation inference and ROC analysis were used to identify significant dose differences in localized lung subregions.
Main Results:
- The voxel-based registration and dose mapping were found to be robust and accurate.
- Patients who experienced RILD consistently received a significantly higher radiation dose in specific voxel clusters in the peripheral medial-basal lung regions.
- The area under the ROC curve (AUC) was higher when analyzing mean doses from identified significant subregions compared to the total lung mean dose.
Conclusions:
- A robust framework combining registration and a VB approach was successfully implemented to model dose-response relationships for RILD.
- Significantly higher doses were delivered to parenchymal regions associated with RILD, even in areas receiving relatively low doses (around 6 Gy).
- The study suggests that localized dose variations, particularly in lower dose regions, play a critical role in the development of late RILD, while high-dose differences may lack a clear spatial signature.
