MRI to CT Synthesis of the Lumbar Spine from a Pseudo-3D Cycle GAN

Oulbacha, Reda; Kadoury, Samuel

doi:10.1109/isbi45749.2020.9098421

Cited by 28 publications

(15 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This synthesis can be used in lesion detection solutions decreasing false–positive reduction rates, hence providing treatment for CT environment drugs that decrease extra expensive modality PET/CT scan requirement. Oulbacha and Kadoury (2020) used an unsupervised technique to generate lumbar spine CT image from MRI scan. The synthesized image was used for the surgical procedure before surgery.…”

Section: Literature Reviewmentioning

confidence: 99%

Deep convolutional generative adversarial network for Alzheimer's disease classification using positron emission tomography (PET) and synthetic data augmentation

Sajjad

Ramzan

Khan

et al. 2021

Microscopy Res & Technique

View full text Add to dashboard Cite

With the evolution of deep learning technologies, computer vision-related tasks achieved tremendous success in the biomedical domain. For supervised deep learning training, we need a large number of labeled datasets. The task of achieving a large number of label dataset is a challenging. The availability of data makes it difficult to achieve and enhance an automated disease diagnosis model's performance. To synthesize data and improve the disease diagnosis model's accuracy, we proposed a novel approach for the generation of images for three different stages of Alzheimer's disease using deep convolutional generative adversarial networks. The proposed model out-perform in synthesis of brain positron emission tomography images for all three stages of Alzheimer disease. The three-stage of Alzheimer's disease is normal control, mild cognitive impairment, and Alzheimer's disease. The model performance is measured using a classification model that achieved an accuracy of 72% against synthetic images. We also experimented with quantitative measures, that is, peak signal-to-noise (PSNR) and structural similarity index measure (SSIM). We achieved average PSNR score values of 82 for AD, 72 for CN, and 73 for MCI and SSIM average score values of 25.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Deep convolutional generative adversarial network for Alzheimer's disease classification using positron emission tomography (PET) and synthetic data augmentation

Sajjad

Ramzan

Khan

et al. 2021

Microscopy Res & Technique

View full text Add to dashboard Cite

show abstract

“…Such an approach has the advantage of factoring in neighboring 3D information, while still maintaining a reasonable memory footprint, contrary to 3D convolutions 14 . To form the output volume, we use the Pseudo‐3D slice recombination strategy as we presented in, 17 selecting the third slice of each nonlimit quadruplet which will be added to the generated final output volume. The missing boundary slices are completed from the first and last quadruplets.…”

Section: Methodsmentioning

confidence: 99%

“…Few applications were dedicated for the spine anatomy, which exposes several pathologies and disk degenerations hindering the labeling process. In recent work, 17 we demonstrated the feasibility of MRI to CT synthesis with a CycleGAN model, showing the advantages of a pseudo-3D approach. However, it lacked voxel distribution cycle consistency between MRI and CT.…”

Section: Introductionmentioning

confidence: 95%

See 1 more Smart Citation

MRI to C‐arm spine registration through Pseudo‐3D CycleGANs with differentiable histograms

Oulbacha

Kadoury

2020

Medical Physics

Self Cite

View full text Add to dashboard Cite

Purpose Image‐guided spine surgery increasingly relies on diagnostic MRI for device navigation, as it allows to visualize the nerves and soft tissues during screw insertion in the pedicle region, which is not possible with preoperative CT or cone beam CT. However, registration of MRI to C‐arm images remains difficult due to differences in visible tissue. Methods In this paper, we introduce a three‐dimensional/two‐dimensional (3D/2D) registration method of preoperative T2‐weighted MRI of the lumbar spine to C‐arm X‐ray using synthetic CT images. The registration work is based on a pseudo‐3D CycleGAN integrating a new cyclic loss function to ensure consistency in MRI and CT synthesis using differentiable histograms to match the multimodal distributions. The unified framework allows to improve bony tissue inference as opposed to regular 2D CycleGAN for image synthesis. A multiplanar digitally reconstructed radiograph (DRR) registration approach aligns the 3D and 2D images. Results Experiments performed on a public dataset of 18 pathological spines yielded a mean dice coefficient of 0.84 ± 0.015 on synthetic CTs. The DRR registration experiments, on the other hand, presented a target localization error of 2.1 ± 0.2mm. Conclusion Intensity distributions and voxel‐wise errors in Hounsfield units show encouraging results, illustrating the network's flexibility of producing qualitatively and quantitatively reasonable synthetic CT scans that can be used in a surgical 3D/2D registration framework. These promising results demonstrate the potential of the synthesis tool prior to integration in an image‐guidance system.

show abstract

“…In recent years, the more frequently used GAN variants in medical imaging are pix2pix [17] and CycleGAN [18]. GAN is widely used in image synthesis [19][20][21][22][23], noise reduction [24][25][26][27], cross-modality [28][29][30][31][32][33], and many other aspects, providing significant help for CAD.…”

Section: Application Of Gan In Medical Imagementioning

confidence: 99%

GAN-based Medical Image Small Region Forgery Detection via a Two-Stage Cascade Framework

Zhang¹,

Huang²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Using generative adversarial network (GAN) [1] for data enhancement of medical images is significantly helpful for many computer-aided diagnosis (CAD) tasks. A new attack called CT-GAN [2] has emerged. It can inject or remove lung cancer lesions to CT scans. Because the tampering region may even account for less than 1% of the original image, even state-of-the-art methods are challenging to detect the traces of such tampering. This paper proposes a cascade framework to detect GAN-based medical image small region forgery like CT-GAN. In the local detection stage, we train the detector network with small sub-images so that interference information in authentic regions will not affect the detector. We use depthwise separable convolution and residual to prevent the detector from over-fitting and enhance the ability to find forged regions through the attention mechanism. The detection results of all sub-images in the same image will be combined into a heatmap. In the global classification stage, using gray level co-occurrence matrix (GLCM) can better extract features of the heatmap. Because the shape and size of the tampered area are uncertain, we train PCA and SVM methods for classification. Our method can classify whether a CT image has been tampered and locate the tampered position. Sufficient experiments show that our method can achieve excellent performance.

show abstract

MRI to CT Synthesis of the Lumbar Spine from a Pseudo-3D Cycle GAN

Cited by 28 publications

References 7 publications

Deep convolutional generative adversarial network for Alzheimer's disease classification using positron emission tomography (PET) and synthetic data augmentation

Deep convolutional generative adversarial network for Alzheimer's disease classification using positron emission tomography (PET) and synthetic data augmentation

MRI to C‐arm spine registration through Pseudo‐3D CycleGANs with differentiable histograms

GAN-based Medical Image Small Region Forgery Detection via a Two-Stage Cascade Framework

Contact Info

Product

Resources

About