Cone Beam Computed Tomography-based Lung Function Imaging using Deep Learning for Functional Image-Guided Adaptative...

Recent studies indicate that lung function can change significantly during radiation therapy (RT) course. However, additional function imaging scans are not part of routine RT workflow, limiting the use of functional avoidance for adaptive therapy. To bridge this gap, we aimed to develop a deep learning model to synthesize functional maps directly from fractional cone-beam computed tomography (CBCT) images, enabling potential functional image-guided adaptive radiotherapy (FIGART).

Data were prospectively collected from 60 lung cancer patients who underwent intensity-modulated radiation therapy. In addition to standard planning CT and fractional cone-beam computed tomography (CBCT) scans, all patients received a baseline single-photon emission computed tomography (SPECT) perfusion scan before radiotherapy. A subset of 16 patients also underwent a follow-up SPECT scan after completing RT. A three-dimensional Gated-Attention U-Net (GAU-Net) was developed to synthesize perfusion maps directly from CBCT images. To address the challenges of inherent CBCT noise and artifacts, the network architecture was augmented by integrating gated attention modules and 3D deformable convolutions within the skip-connection pathways. This design enhances multi-scale feature fusion for more robust image synthesis. The synthesized perfusion images were quantitatively compared to the reference SPECT scans using cross validation. Voxel-wise agreement was assessed using the Spearman correlation coefficient (R), structural similarity index (SSIM) and mean squared error (MSE), while functional region agreement was evaluated using the Dice similarity coefficient (DSC). The potential clinical benefit was assessed through dosimetric evaluation.

Quantitative analysis demonstrated CBCT-based functional images strong agreement with ground-truth SPECT, with a R of 0.68±0.09, SSIM of 0.75±0.09, MAE of 0.15±0.03, MSE of 0.04±0.02, and DSC values of 0.76±0.07 and 0.85±0.05 for high- and low-functional regions, respectively. Regarding dose sparing of the lung high-function region, CBCT-based functional imaged guided plan significantly reduced the mean dose by 6.97±4.22 Gy and V20 by 13.54±8.66% compared to anatomical plan, while achieving a better target dose homogeneity index of 5.26±0.65.

We developed a CBCT-based lung function imaging method using a deep learning model. The evaluation demonstrated its feasibility for functional guidance planning in FIGART. A larger cohort study is warranted in the future.