APSR2026-679

 
Regular
Submitted
Poster Presentation
 
AP14-012
P31
 
Format: PPTX Only
 
Format: PDF Only
08. Lung Cancer
8

No more than 15 authors can be listed.
Please ensure the authors are listed in the right order.
If there are multiple affiliations, please use the "+" button next to "Select Country/Region" to add additional affiliation fields.

Peixin YU * The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Yao PU - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Mayang ZHAO - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Qingpei LAI - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Yitao ZHUANG - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Zhenxi ZHANG - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Jing CAI - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
Ge REN - The Hong Kong Polytechnic University Department of Health Technology and Informatics Hong Kong Hong Kong
-
-
-
-
-
-
-
Presenting Author Details
Peixin
YU
pei-xin.yu@connect.polyu.hk
+85251001390
Under 30
Hong Kong
Original Research
CBCT-based Lung Function Imaging using Deep Learning for Functional Image-Guided Adaptative Radiation Therapy
Maximum 25 words
Abstract Content
<p>Recent studies indicate that lung function can change significantly during radiation therapy (RT). However, additional function imaging scans are not part of routine RT workflow, limiting their use 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).</p>
<p>This study prospectively included 60 lung cancer patients treated with intensity-modulated RT. Each patient received a planning CT, fractional CBCTs, and a baseline single-photon emission CT (SPECT) perfusion scan. We developed a 3D Gated-Attention U-Net (GAU-Net) architecture, incorporating gated attention modules and 3D deformable convolutions integrated into skip-connection pathways to mitigate CBCT noise and artifacts. The model was trained to synthesize SPECT-equivalent perfusion maps from CBCT inputs. Performance was evaluated via cross-validation, employing voxel-wise metrics (Spearman’s correlation coefficient R, structural similarity index SSIM, mean squared error MSE) and region-based functional agreement (Dice similarity coefficient DSC for high/low perfusion regions). Dosimetric evaluation compared plans optimized using synthetic function maps versus conventional anatomic plans.</p>
<p>The proposed GAU-Net successfully generated synthetic perfusion images that correlated well with ground-truth SPECT. Quantitative analysis showed strong agreement: Spearman’s R=0.68, SSIM=0.75, MSE=0.03. Functional region analysis yielded DSC values of 0.76 (high-function) and 0.85 (low-function). Dosimetrically, auto-planning based on CBCT-derived function maps significantly reduced the mean dose (6.97±4.22 Gy) and V20 (13.54±8.66%) to high-function lung regions, while simultaneously improving target dose homogeneity (homogeneity index: 5.26±0.65) compared to conventional planning.</p>
<p>This study demonstrates a feasible deep learning framework for synthesizing lung perfusion maps directly from routine CBCT. The results underscore its potential to support functional image-guided adaptive radiotherapy (FIGART), enabling perfusion-informed treatment planning that effectively spares functional lung tissue.</p>
 
 
 
 
 
No
No
 
JPG or JPEG or PNG format ONLY
Perfusion imaging, Deep learning, Functional image-guided adaptive radiotherapy.
maximum 3 keywords and no abbreviations, separated by commas
This research was partly supported by the Health and Medical Research Fund (11222456) of the Health Bureau, the Pneumoconiosis Compensation Fund Board in HKSAR, and Shenzhen Science and Technology Program (JCYJ20230807140403007).
300