Design, Fabrication, and Implementation of Voxel-Based 3D Printed Heterogeneous Lung Lesion Phantoms for Assessment of CT Imaging Conditions on Texture Quantification
Realistic virtual lesion models are valuable in medical imaging applications including phantom design and observer studies. Radiologic diagnostic information rarely include lesion texture due to the fact that texture quantification is sensitive to changing imaging conditions. These effects are not well understood, in part, due to a lack of ground-truth phantoms with realistic textures. Internal tumor heterogeneity in nodules can be predictive of lesion malignancy but is not well understood and virtual lesion models will facilitate research in this area. The purpose of this study was to design and fabricate realistic virtual lung nodules with internal heterogeneity characteristics, and assess the variability as well as determine which imaging conditions provides the most accurate texture features compared to voxel-based 3D printed textured lesions for witch the true texture features are known.
We propose a texture synthesis method that accounts for the effects of the imaging system to mimic the appearance of texture in real nodules. Modulation Transfer Function blurring effects and noise contamination was included in the texture generation based on a 3D-Clustered Lumpy Background (3D-CLB). The governing parameters of the 3D-CLB were optimized using a Generic Algorithm with an objective function of Mahalanobis distance between synthesized textures and real lesion textures features. The resultant texture was objectively and visually similar to real nodules of the same heterogeneity category.
The heterogeneous lesion phantoms were designed with three shapes (spherical, lobulated, spiculated), two textures (homogenous, heterogeneous), and two sizes (diameter < 1.5cm, 1.5cm<diameter<3cm), resulting in 24 lesions (replica of each). The lesions were inserted into an anthropomorphic thorax phantom (Multipurpose Chest Phantom N1, Kyoto Kagaku) and imaged using a commercial CT system (SOMATOM Definition Flash, Siemens Healthcare) at three CTDI levels (0.¬¬¬67, 1.42, 5.80mGy), three reconstruction algorithm (FBP, IR-2, IR-4), four reconstruction kernels (standard, soft, edge), and two slice thicknesses (0.6mm, 5mm). Texture features from these images were extracted and compared to the ground truth feature values. High bias and variance was seen for each feature. Significance level of effects of imaging and lesion conditions were explored. Variability related to CT imaging acquisition and reconstruction techniques is a clinically important source of bias and variance during lesion heterogeneity quantification. Lesion size and shape should also be taken into consideration. Different features are influenced by imaging factors and lesion conditions differently, and as such, feature quantities are highly susceptible to parameter choices and lesion attributes.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Masters Theses