No abstract

Accurate lung tumor segmentation is problematic when the tumor boundary or edge, which reflects the advancing edge of the tumor, is difficult to discern on chest CT or PET. We propose a 'topo-poly' graph model to improve identification of the tumor extent. Our model incorporates an intensity graph and a topology graph. The intensity graph provides the joint PET-CT foreground similarity to differentiate the tumor from surrounding tissues. The topology graph is defined on the basis of contour tree to reflect the inclusion and exclusion relationship of regions. By taking into account different topology relations, the edges in our model exhibit topological polymorphism. These polymorphic edges in turn affect the energy cost when crossing different topology regions under a random walk framework, and hence contribute to appropriate tumor delineation. We validated our method on 40 patients with non-small cell lung cancer where the tumors were manually delineated by a clinical expert. The studies were separated into an 'isolated' group (n = 20) where the lung tumor was located in the lung parenchyma and away from associated structures / tissues in the thorax and a 'complex' group (n = 20) where the tumor abutted / involved a variety of adjacent structures and had heterogeneous FDG uptake. The methods were validated using Dice's similarity coefficient (DSC) to measure the spatial volume overlap and Hausdorff distance (HD) to compare shape similarity calculated as the maximum surface distance between the segmentation results and the manual delineations. Our method achieved an average DSC of 0.881 ± 0.046 and HD of 5.311 ± 3.022 mm for the isolated cases and DSC of 0.870 ± 0.038 and HD of 9.370 ± 3.169 mm for the complex cases. Student's t-test showed that our model outperformed the other methods (p-values <0.05).

The capability of generalization to unseen domains is crucial for deep learning models when considering real-world scenarios. However, current available medical image datasets, such as those for COVID-19 CT images, have large variations of infections and domain shift problems. To address this issue, we propose a prior knowledge driven domain adaptation and a dual-domain enhanced self-correction learning scheme. Based on the novel learning schemes, a domain adaptation based self-correction model (DASC-Net) is proposed for COVID-19 infection segmentation on CT images. DASC-Net consists of a novel attention and feature domain enhanced domain adaptation model (AFD-DA) to solve the domain shifts and a self-correction learning process to refine segmentation results. The innovations in AFD-DA include an image-level activation feature extractor with attention to lung abnormalities and a multi-level discrimination module for hierarchical feature domain alignment. The proposed self-correction learning process adaptively aggregates the learned model and corresponding pseudo labels for the propagation of aligned source and target domain information to alleviate the overfitting to noises caused by pseudo labels. Extensive experiments over three publicly available COVID-19 CT datasets demonstrate that DASC-Net consistently outperforms state-of-the-art segmentation, domain shift, and coronavirus infection segmentation methods. Ablation analysis further shows the effectiveness of the major components in our model. The DASC-Net enriches the theory of domain adaptation and self-correction learning in medical imaging and can be generalized to multi-site COVID-19 infection segmentation on CT images for clinical deployment.

The insufficiency of annotated medical imaging scans for cancer makes it challenging to train and validate data-hungry deep learning models in precision oncology. We propose a new richer generative adversarial network for free-form 3D tumor/lesion synthesis in computed tomography (CT) images. The network is composed of a new richer convolutional feature enhanced dilated-gated generator (RicherDG) and a hybrid loss function. The RicherDG has dilated-gated convolution layers to enable tumor-painting and to enlarge perceptive fields; and it has a novel richer convolutional feature association branch to recover multi-scale convolutional features especially from uncertain boundaries between tumor and surrounding healthy tissues. The hybrid loss function, which consists of a diverse range of losses, is designed to aggregate complementary information to improve optimization. We perform a comprehensive evaluation of the synthesis results on a wide range of public CT image datasets covering the liver, kidney tumors, and lung nodules. The qualitative and quantitative evaluations and ablation study demonstrated improved synthesizing results over advanced tumor synthesis methods.

No abstract