Room: AAPM ePoster Library
Purpose: To investigate the potential association between spatially-encoded radiomic features of the lungs and pulmonary function.
Methods: We developed a technique to generate radiomic filtered images of the lungs based on a voxel-based extraction technique. The lungs were segmented on 25 CT images, from which radiomic features were extracted regionally using a sliding window technique. For each voxel in the lungs, 62 radiomic features were calculated in a rotationally-invariant 3D neighborhood to capture spatially-encoded information. In general, such an approach results in an image tensor object, i.e., each voxel in the original lung CT is represented by a 62-dimensional radiomic feature vector. To test the technique as a potential pulmonary biomarker, the radiomic filtered images were subsequently compared to corresponding Galligas PET images as ground truth for pulmonary function based on Spearman coefficients (r). A Canonical Correlation Analysis (CCA)-based feature fusion method was then implemented to enhance the association between the radiomic filtered images and Galligas ventilation. Finally, the Spearman distributions were compared with 37 individual CT ventilation image (CTVI) algorithms from the VAMPIRE Challenge to assess the algorithm’s overall performance relative to conventional CT-based techniques.
Results: Several radiomic filtered images were identified to be correlated with Galligas PET lung imaging. The most robust association was found to be the Run Length Encoding feature, Run-Length Non-uniformity (0.21
Conclusion: This preliminary study indicates that spatially-encoded lung texture and lung density are potentially associated with pulmonary function as measured via Galligas PET ventilation images. Collectively, low density, heterogeneous coarse lung texture was often associated with lower Galligas radiotracer amounts.