Mask2Lesion: Mask-Constrained Adversarial Skin Lesion Image Synthesis

Abhishek, Kumar; Hamarneh, Ghassan

doi:10.1007/978-3-030-32778-1_8

Cited by 26 publications

(24 citation statements)

References 23 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The boundary between lesion and surrounding tissues on dermoscopic images is discernible when compared with the soft tissue boundaries on grayscale CT. Thus, it can be seen from Figure 5 that the 3D tumors generated by the method in [29] are blurry when compared with our proposed FRGAN. The tumors generated by FRGAN are natural with high visual texture similarities with surroundings.…”

Section: Qualitative Analysismentioning

confidence: 91%

“…Red: Best response 2, when using extra synthetic images generated by synthesis models, the overall segmentation performance was improved in terms of Dice, Jaccard, VOE, RVD, and HD over all the three datasets. Specifically, the AU-Net/U2-Net trained with tumors generated by our method achieved the best performance over LiTS and KiTS datasets in terms of spatial overlap measured by Dice, Jaccard, VOE, and HD when compared to those trained with tumors generated by [4] and [29]. For LiTS, the advanced U2-Net obtained the best performance with our synthetic samples, which outperformed the second best by 1.1%/1.1% in terms of Dice/Jaccard.…”

Section: Comparison With Other Methodsmentioning

confidence: 97%

“…The proposed model is compared with two state-of-the-art models. The model in [4] is a cGAN model to generate 3D lung nodules of different sizes and the model in [29] is based on the pix2pix model [3] with the original purpose to generate masked skin lesions. Red: Best response 2, when using extra synthetic images generated by synthesis models, the overall segmentation performance was improved in terms of Dice, Jaccard, VOE, RVD, and HD over all the three datasets.…”

Section: Comparison With Other Methodsmentioning

confidence: 99%

“…Even though GAN based approaches and its variants have been used to increase data volume and data variety in different image analysis tasks, the methods mentioned above cannot support user-customized free-form tumor synthesis. Besides, high-quality synthetic results cannot always be guaranteed because the detailed texture and boundary characteristics may not be preserved during the synthesis process [29,10].…”

Section: Related Workmentioning

confidence: 99%

“…Even though the model in [29] and the model in [4] can also support user-customized tumor synthesis, our model improved the visual appearance of the synthesized tumors using the newly introduced RicehrGonv-Branch and hybrid loss functions as shown in Figure 5. The original model in [29] is designed for 2D skin lesions from dermoscopic images. The boundary between lesion and surrounding tissues on dermoscopic images is discernible when compared with the soft tissue boundaries on grayscale CT.…”

Section: Qualitative Analysismentioning

confidence: 99%

See 4 more Smart Citations

Free-form tumor synthesis in computed tomography images via richer generative adversarial network

Cui

Sun

Meng

et al. 2021

Knowledge-Based Systems

View full text Add to dashboard Cite

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.

show abstract

Section: Qualitative Analysismentioning

confidence: 91%

Section: Comparison With Other Methodsmentioning

confidence: 97%

Section: Comparison With Other Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Qualitative Analysismentioning

confidence: 99%

See 3 more Smart Citations

Free-form tumor synthesis in computed tomography images via richer generative adversarial network

Cui

Sun

Meng

et al. 2021

Knowledge-Based Systems

View full text Add to dashboard Cite

show abstract

A partial convolution generative adversarial network for lesion synthesis and enhanced liver tumor segmentation

Liu

Yang

2023

J Applied Clin Med Phys

View full text Add to dashboard Cite

Lesion segmentation is critical for clinicians to accurately stage the disease and determine treatment strategy. Deep learning based automatic segmentation can improve both the segmentation efficiency and accuracy. However, training a robust deep learning segmentation model requires sufficient training examples with sufficient diversity in lesion location and lesion size. This study is to develop a deep learning framework for generation of synthetic lesions with various locations and sizes that can be included in the training dataset to enhance the lesion segmentation performance. The lesion synthesis network is a modified generative adversarial network (GAN). Specifically, we innovated a partial convolution strategy to construct a U‐Net‐like generator. The discriminator is designed using Wasserstein GAN with gradient penalty and spectral normalization. A mask generation method based on principal component analysis (PCA) was developed to model various lesion shapes. The generated masks are then converted into liver lesions through a lesion synthesis network. The lesion synthesis framework was evaluated for lesion textures, and the synthetic lesions were used to train a lesion segmentation network to further validate the effectiveness of the lesion synthesis framework. All the networks are trained and tested on the LITS public dataset. Our experiments demonstrate that the synthetic lesions generated by our approach have very similar distributions for the two parameters, GLCM‐energy and GLCM‐correlation. Including the synthetic lesions in the segmentation network improved the segmentation dice performance from 67.3% to 71.4%. Meanwhile, the precision and sensitivity for lesion segmentation were improved from 74.6% to 76.0% and 66.1% to 70.9%, respectively. The proposed lesion synthesis approach outperforms the other two existing approaches. Including the synthetic lesion data into the training dataset significantly improves the segmentation performance.

show abstract