Author: Bas W. Raaymakers, Mario Ries, Paris Tzitzimpasis, Cornel Zachiu 👨🔬
Affiliation: Department of Radiotherapy, University Medical Center Utrecht, University Medical Center Utrecht, UMC Utrecht 🌍
Purpose: Radiation pneumonitis affects approximately 10-30% of lung cancer patients treated with radiation therapy (RT), posing a significant dose-limiting factor. Recently developed CT-ventilation methods enable the generation of temporally dense longitudinal ventilation data. The assessment of regional ventilation changes from this data can provide essential insight into treatment response.
Methods: We propose a Bayesian inference framework to estimate genuine functional changes from longitudinal ventilation scans. Our method prioritizes changes that follow a monotonic trend and assigns lower confidence to regions with large fluctuations, often caused by noise or imaging artifacts. The framework outputs the estimated volumes of significant functional increases and declines by using a Bayesian sequential update component which uses posterior data from a time-point as prior information for subsequent estimates.
To validate our method, we created a synthetic dataset by simulating localized functional changes to a PET ventilation scan and adding Gaussian noise. This dataset was used for calibrating the internal parameters of our algorithm. We then applied the framework to a dataset of 11 lung cancer patients, each with multiple 4DCT scans acquired during treatment. Ventilation maps generated using a CT-ventilation method developed by our group. Those maps were then used as input for the Bayesian model.
Results: In the synthetic dataset, the proposed method identified true regions of functional change with an optimal accuracy exceeding 90%. When applied to the patient dataset, significant regions of functional decline were detected in four patients. These regions were primarily located proximal or contralateral to the gross tumor volume (GTV).
Conclusion: Our findings demonstrate that significant functional changes can occur during RT treatment. More generally, our developed framework provides a robust method for assessing longitudinal ventilation changes and has the potential to guide adaptive treatment strategies, helping to reduce post-treatment toxicities such as radiation pneumonitis.