3D fully convolutional networks for co-segmentation of tumors on PET-CT images

Zhi, Zhao; Kim, Yusung; Zhou, Leixin; Plichta, Kristin A.; Allen, Bryan G.; Buatti, John M.; Wu, Xiaodong

doi:10.1109/isbi.2018.8363561

Cited by 78 publications

(63 citation statements)

References 13 publications

(34 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many studies [4][5][6] have indicated that the co-segmentation method by combining different segmentation methods is treated as an energy minimization problem to delineate the gross tumor contours. Besides, due to the superior contrast of PET images and high spatial resolution of CT images, more recent methods and techniques [7][8][9][10] in the field of clinic and lesion segmentation prefer to integrate PET and CT images.…”

Section: Co-segmentation Methodsmentioning

confidence: 99%

Fast and Accurate Lung Tumor Spotting and Segmentation for Boundary Delineation on CT Slices in a Coarse-to-Fine Framework

Pang

et al. 2019

Communications in Computer and Information Science

View full text Add to dashboard Cite

Label noise and class imbalance are two of the critical challenges when training image-based deep neural networks, especially in the biomedical image processing domain. Our work focuses on how to effectively and accurately address the two challenges in the task of lesion segmentation from biomedical/medical images. To address the pixel-level label noise problem, an advanced transfer training and learning approach with a detailed DICOM pre-processing work is proposed in this paper. To address the tumor/non-tumor class imbalance problem, we exploit a self-adaptive fully convolutional neural network with an automated weight distribution mechanism to accurately spot the Radiomics lung tumor regions. Furthermore, an improved conditional random field method is utilized to obtain a sophisticated lung tumor contour delineation and segmentation. Finally, our approach has been evaluated using several well-known evaluation metrics on the Radiomics Lung Tumor segmentation dataset used in the 2018 IEEE VIP-CUP Challenge, which totally contains 422 patient records with 51513 Radiomics Lung slices. Experimental results show that our weakly supervised learning algorithm outperforms other deep models and state-of-the-art approaches.

show abstract

Section: Co-segmentation Methodsmentioning

confidence: 99%

Fast and Accurate Lung Tumor Spotting and Segmentation for Boundary Delineation on CT Slices in a Coarse-to-Fine Framework

Pang

et al. 2019

Communications in Computer and Information Science

View full text Add to dashboard Cite

show abstract

“…Xu et al [11] cascaded two V-Nets [12] to detect bone lesions, using CT alone as the input to the first V-Net and a pre-fused PET-CT image for the second. Similarly, Zhong et al [13] trained one U-Net [44] for PET and one for CT, combining the results using a graph cut algorithm. None of this prior work, however, considered how the visual characteristics, specific to each image at different locations, could be integrated in a spatially varying manner.…”

Section: Related Workmentioning

confidence: 99%

“…• A two-branch (TB) CNN, implementing a fusion strategy where each modality was processed separately and the outputs from each modality were combined [9], [13], [41]. The CNN was similar to the architecture in Fig.…”

Section: G Experimental Designmentioning

confidence: 99%

Co-Learning Feature Fusion Maps From PET-CT Images of Lung Cancer

Kumar

Fulham

Feng

et al. 2020

IEEE Trans. Med. Imaging

166

View full text Add to dashboard Cite

The analysis of multi-modality positron emission tomography and computed tomography (PET-CT) images requires combining the sensitivity of PET to detect abnormal regions with anatomical localization from CT. However, current methods for PET-CT image analysis either process the modalities separately or fuse information from each modality based on knowledge about the image analysis task. These methods generally do not consider the spatially varying visual characteristics that encode different information across the different modalities, which have different priorities at different locations. For example, a high abnormal PET uptake in the lungs is more meaningful for tumor detection than physiological PET uptake in the heart. Our aim is to improve fusion of the complementary information in multi-modality PET-CT with a new supervised convolutional neural network (CNN) that learns to fuse complementary information for multi-modality medical image analysis. Our CNN first encodes modality-specific features and then uses them to derive a spatially varying fusion map that quantifies the relative importance of each modality's features across different spatial locations. These fusion maps are then multiplied with the modality-specific feature maps to obtain a representation of the complementary multi-modality information at different locations, which can then be used for image analysis, e.g. region detection. We evaluated our CNN on a region detection problem using a dataset of PET-CT images of lung cancer. We compared our method to baseline techniques for multi-modality image analysis (pre-fused inputs, multi-branch techniques, multichannel techniques) and demonstrated that our approach had a significantly higher accuracy (p < 0.05) than the baselines.

show abstract

“…The performance of these methods relies heavily on a large set of training data, which include manually annotated images for learning features in the medical scans. Neural Networks (Goceri and Goceri, 2015;Ramrez et al, 2018;Zhong et al, 2018), Support Vectors Machines (Lee et al, 2008;Amiri et al, 2016) and Random Forests (Shah et al, 2017;Bauer et al, 2012;Conze et al, 2016) are amount the popular methods used for medical images segmentation. Unlike these supervised methods, unsupervised segmentation techniques are very flexible as they can be applied on a small set of data from which they can learn classification rules based some similarity criterion.…”

Section: Related Workmentioning

confidence: 99%

Nephroblastoma Analysis in Mri Images

et al. 2019

View full text Add to dashboard Cite

The annotation of the tumour from medical scans is a crucial step in nephroblastoma treatment. Therefore, an accurate and reliable segmentation method is needed to facilitate the evaluation and the treatments of the tumour. The proposed method serves this purpose by performing the segmentation of nephroblastoma in MRI scans. The segmentation is performed by adapting and a 2D free hand drawing tool to select a region of interest in the scan slices. Results from 24 patients show a mean root-mean-square error of 0.0481 ± 0.0309, an average Dice coefficient of 0.9060 ± 0.0549 and an average accuracy of 99.59% ± 0.0039. Thus the proposed method demonstrated an effective agreement with manual annotations.

show abstract

3D fully convolutional networks for co-segmentation of tumors on PET-CT images

Cited by 78 publications

References 13 publications

Fast and Accurate Lung Tumor Spotting and Segmentation for Boundary Delineation on CT Slices in a Coarse-to-Fine Framework

Fast and Accurate Lung Tumor Spotting and Segmentation for Boundary Delineation on CT Slices in a Coarse-to-Fine Framework

Co-Learning Feature Fusion Maps From PET-CT Images of Lung Cancer

Nephroblastoma Analysis in Mri Images

Contact Info

Product

Resources

About