Computer-assisted delineation of lung tumor regions in treatment planning CT images with PET/CT image sets based on an optimum contour selection method

Quantitative imaging: Correlating image features with the segmentation accuracy of PET based tumor contours in the lung

The purpose of this study was to investigate the correlation between image features extracted from PET images and the accuracy of manually drawn lesion contours in the lung. Such correlations are interesting in that they could potentially be used in predictive models to help guide physician contouring. In this work, 26 synthetic PET datasets were created using an anthropomorphic phantom and Monte Carlo simulation. Manual contours of simulated lesions were provided by 10 physicians. Contour accuracy was quantified using five commonly used similarity metrics which were then correlated with several features extracted from the images. Features were sub-divided into three groups using intensity, geometry, and texture as categorical descriptors. When averaged among the participants, the results showed relatively strong correlations with complexity and contrast_I (r≥0.65, p<0.001), and moderate correlations with several other image features (r≥0.5, p<0.01). The predictive nature of these correlations was improved through stepwise regression and the creation of multi-feature models. Imaging features were also correlated with the standard deviation of contouring error in order to investigate inter-observer variability. Several features were consistently identified as influential including integral of mean curvature and complexity. These relationships further the understanding as to what causes variation in the contouring of PET positive lesions.

Computer-assisted framework for machine-learning-based delineation of GTV regions on datasets of planning CT and PET/CT images

We have proposed a computer-assisted framework for machine-learning-based delineation of gross tumor volumes (GTVs) following an optimum contour selection (OCS) method. The key idea of the proposed framework was to feed image features around GTV contours (determined based on the knowledge of radiation oncologists) into a machine-learning classifier during the training step, after which the classifier produces the 'degree of GTV' for each voxel in the testing step. Initial GTV regions were extracted using a support vector machine (SVM) that learned the image features inside and outside each tumor region (determined by radiation oncologists). The leave-one-out-by-patient test was employed for training and testing the steps of the proposed framework. The final GTV regions were determined using the OCS method that can be used to select a global optimum object contour based on multiple active delineations with a LSM around the GTV. The efficacy of the proposed framework was evaluated in 14 lung cancer cases [solid: 6, ground-glass opacity (GGO): 4, mixed GGO: 4] using the 3D Dice similarity coefficient (DSC), which denotes the degree of region similarity between the GTVs contoured by radiation oncologists and those determined using the proposed framework. The proposed framework achieved an average DSC of 0.777 for 14 cases, whereas the OCS-based framework produced an average DSC of 0.507. The average DSCs for GGO and mixed GGO were 0.763 and 0.701, respectively, obtained by the proposed framework. The proposed framework can be employed as a tool to assist radiation oncologists in delineating various GTV regions.