Structure-Enhanced Translation from PET to CT Modality with Paired GANs

X-ray computed tomography (CT) imaging plays an essential role in disease diagnosis due to its noninvasive, painless mode and superior penetration depth. However, the resolution of the soft tissue ...

Positron emission tomography (PET) and computed tomography (CT) play a vital role in tumor-related medical diagnosis, assessment, and treatment planning. However, full-dose PET and CT pose the risk of excessive radiation exposure to patients, whereas low-dose images compromise image quality, impacting subsequent tumor recognition and diseasediagnosis. To solve such problems, we propose a Noise-Assisted Hybrid Attention Network (NAHANet) to reconstruct full-dose PET and CT images from low-dose PET (LDPET) and CT (LDCT) images to reduce patient radiation risks while ensuring the performance of subsequent tumor recognition. NAHANet contains two branches: the noise feature prediction branch (NFPB) and the cascaded reconstruction branch. Among them, NFPB providing noise features for the cascade reconstruction branch. The cascaded reconstruction branch comprises a shallow feature extraction module and a reconstruction module which contains a series of cascaded noise feature fusion blocks (NFFBs). Among these, the NFFB fuses the features extracted from low-dose images with the noise features obtained by NFPB to improve the feature extraction capability. To validate the effectiveness of the NAHANet method, we performed experiments using two public available datasets: the Ultra-low Dose PET Imaging Challenge dataset and Low Dose CT Grand Challengedataset. As a result, the proposed NAHANet achieved higher performance on common indicators. For example, on the CT dataset, the PSNR and SSIM indicators were improved by 4.1dB and 0.06 respectively, and the rMSE indicator was reduced by 5.46 compared with the LDCT; on the PET dataset, the PSNR and SSIM was improved by 3.37dB and 0.02, and the rMSE was reduced by 9.04 compared with theLDPET. This paper proposes a transformer-based denoising algorithm, which utilizes hybrid attention to extract high-level features of low dose images and fuses noise features to optimize the denoising performance of the network, achieving good performance improvements on low-dose CT and PETdatasets.

Sinogram domain metal artifact correction of CT via deep learning

To mitigate the public health risks associated with the increasing utilization of computed tomography (CT), it is desirable to implement a low-dose scanning protocol. However, low-dose CT produces poor image quality due to the increased quantum noise, requiring an effective image denoising method for its clinical use. Recently, convolutional neural network (CNN)-based methods showed state-of-the-art performance in CT image denoising. However, a large amount of paired training data is essential for their high performance, which is generally not available in medical imaging fields. To alleviate this problem, we propose a new framework to train a CNN-based denoiser without real paired CT images or access to the proprietary information such as raw scan data and a reconstruction kernel. Specifically, we estimate a reconstruction kernel from normal-dose CT images and synthesize paired CT images using an image domain noise insertion technique. To validate the proposed method, we used extended cardiac-torso phantoms and the 2016 Low-Dose CT Grand Challenge datasets. The denoising performance of the CNN-based denoiser trained with the proposed method was compared with the conventional approach that generated paired training data by adding noise to raw scan data and reconstructing images with the known reconstruction kernel. To quantitatively evaluate the image quality, we used the peak signal-to-noise ratio, structural similarity index measure, modulation transfer function, and noise power spectrum. The extensive qualitative and quantitative results suggest that the proposed method enables CNN-based denoisers to be trained without access to raw scan data and a reconstruction kernel while maintaining their denoising performance.

ObjectiveWe developed a new computed tomography (CT)-based spatial normalization method and CT template to demonstrate its usefulness in spatial normalization of positron emission tomography (PET) images with [18F] fluorodeoxyglucose (FDG) PET studies in healthy controls.Materials and MethodsSeventy healthy controls underwent brain CT scan (120 KeV, 180 mAs, and 3 mm of thickness) and [18F] FDG PET scans using a PET/CT scanner. T1-weighted magnetic resonance (MR) images were acquired for all subjects. By averaging skull-stripped and spatially-normalized MR and CT images, we created skull-stripped MR and CT templates for spatial normalization. The skull-stripped MR and CT images were spatially normalized to each structural template. PET images were spatially normalized by applying spatial transformation parameters to normalize skull-stripped MR and CT images. A conventional perfusion PET template was used for PET-based spatial normalization. Regional standardized uptake values (SUV) measured by overlaying the template volume of interest (VOI) were compared to those measured with FreeSurfer-generated VOI (FSVOI).ResultsAll three spatial normalization methods underestimated regional SUV values by 0.3-20% compared to those measured with FSVOI. The CT-based method showed slightly greater underestimation bias. Regional SUV values derived from all three spatial normalization methods were correlated significantly (p < 0.0001) with those measured with FSVOI.ConclusionCT-based spatial normalization may be an alternative method for structure-based spatial normalization of [18F] FDG PET when MR imaging is unavailable. Therefore, it is useful for PET/CT studies with various radiotracers whose uptake is expected to be limited to specific brain regions or highly variable within study population.

Small-animal positron emission tomography as a tool for neuropharmacology

An automated system is developed for lung and mediastinum segmentation in lung CT (Computed Tomography) images for the purpose of using these segmentations not only in CT images but also in PET (Positron Emission Tomography) images to exploit the useful integration of the CT and PET images performed by the highly valuable oncological equipment PET/CT. Segmentation is the most crucial step in a CAD (Computer-Aided Diagnosis) system as lung borders delimit the region inside which pathological areas are searched for, while mediastinum borders identify the region containing the lymph nodes used for staging and restaging phases. Our method consists of an appropriate combination of image processing techniques. It was tested on CT images with different attributes such as resolution and slice thickness, containing 47 cancerous areas. We achieved almost 90% and 100% correct segmentation for lung and mediastinum, respectively.

Respiratory motion blurs positron emission tomography (PET) images in thorax and abdominal regions, and can cause attenuation-related errors in quantitative parameters. In addition, this motion can cause localization error and disappearance of tumor(s) near the diaphragm. Respiratory gated PET imaging, or 4D PET imaging, is an attractive solution. For 4D PET, a surrogate measurement of the patient's breathing is made during the scan, and based on this measurement the list-mode PET data are sorted and reconstructed into multiple images. In several studies, it has been reported that 4D PET imaging recovers tumor volume more accurately than 3D PET imaging, and yields more accurate standard uptake values. In spite of its promise, 4D PET images suffer from statistical noise because only a fraction of the acquired data can be used in each image. Moreover, they also suffer from inconsistent attenuation correction due to mismatches between PET and CT images. In this paper, we propose a novel framework for phased attenuation correction and respiratory motion compensation for a 3D PET image by using a 3D CT image and multiple respiratory-phase 3D MR images. For phased attenuation correction, we generate and utilize 4D CT image obtained via non-rigid registrations among a 3D CT image and multiple respiratory-phase 3D MR images. A 3D PET image at a selected respiratory phase is then obtained by adding attenuation corrected multiple 3D PET images through non-rigid transformation. Experimental result for a clinical dataset shows that the proposed algorithm can provide much clearer organ boundary with improved quality than the conventional method in a 3D PET image.

In this study, we propose a CNN-GAN-based real-time processing technique for filtering images of underwater cables used in power systems. This addresses the excessive interference impurities that are frequently observed in images captured by remotely operated vehicles (ROVs). The process begins with the input of the original image into the convolutional neural network (CNN). Subsequently, the training outcomes, which serve as input parameters for the generative adversarial network (GAN), facilitate the filtering process. The system also calculates both the structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR), performing model updates via backward propagation. This technique utilizes deep learning technologies to achieve rapid, real-time filtering of underwater cable images. The experimental results reveal that the loss function of the CNN reaches 0.16 with an accuracy of 97.5%, while the loss function of the adversarial GAN network approaches 0.05. Compared with traditional methods such as DDN, JORDOR, RESCAN, and PRENet, the proposed CNN-GAN algorithm exhibits superior performance, as evidenced by the higher PSNR and SSIM values. Specifically, for clear water images, the PSNR reaches 29.86 dB and the SSIM is 0.9045. For severely polluted images, the PSNR is 28.67 dB and the SSIM is 0.8965, while for unevenly illuminated images, the PSNR and SSIM values are 24.37 dB and 0.88, respectively. These enhancements significantly benefit the monitoring and maintenance of power systems.

Whole-body positron emission tomography (PET) with F-fluorodeoxyglucose (FDG) has proven to be a very effective staging imaging modality in manymalignant tumours, particularly in lung cancer, malignant melanoma, lymphoma, colorectal cancer and head and neck tumours [1–7]. Since PET images have a fairly high resolution (,6 mm), even small lesions with an increased FDG uptake can be detected. This represents a critical advantage of PET over the conventional cross-sectional imaging modalities CT and MRI. PET can be used to identify tumour tissue on the basis of FDG uptake, rather than on the basis of morphological appearance or size, and thus also in lesions with a size less than 1 cm. Unfortunately, FDG–PET provides little anatomical information, and therefore it is sometimes difficult to determine where the lesions detected are located and what they represent. It is important for the surgeon to know precisely where lesions are located, especially for small tumours. Currently, comparison and juxtaposition of PET images and CT images serves to define the localization of the lesions. Owing to the different image slice thickness, the different positioning of the patient during the two examinations and the different states of nutrition, i.e. fasting for 4 h prior to FDG–PET, it is often very difficult and time consuming to match the images for exact anatomical localization. Image co-registration has been used to overcome these problems [8]. It has been recognized that proper image co-registration is rather difficult for whole-body examinations with separate scanners. For identical patient positioning during PET and CT scanning, special ‘‘cast forming’’ vacuum mattresses or other immobilization methods should be used. However, patients frequently find these methods of immobilization uncomfortable. Logistic problems also limit the use of image co-registration of PET and CT with separate scanners. Scheduling problems arise when performing the two examinations one after the other. These limitations restrict image co-registration with separate scanners to individual cases. In our experience, the exact localization of lesions is often not possible when juxtaposing PET and CT images. Examinations had to be repeated for exact image co-registration. As a result, image co-registration with separate scanners has never become routine in our practice. To overcome the problems related to software fusion, various prototype systems combining molecular and morphologic imaging have been developed. Hasegawa and colleagues proposed an integrated SPECT/CT scanner [9], and in 1999 Townsend and his colleagues at the University of Pittsburgh presented a prototype integrated PET/CT scanner by combining a CT scanner and a partial ring, rotating PET scanner in a single gantry [10, 11]. Such systems are attractive not only because of the ‘‘hardware’’ image coregistration capabilities but also because the CT data can be used to correct PET scans for patient self-absorption of the PET emission rays. This decreases the acquisition time for the correcting attenuation scans by an order of magnitude compared with the conventional techniques used in PET for this purpose, and decreases the overall imaging time by up to 30% depending on the CT system used. In March 2001, the first clinical integrated PET/ CT device (Discovery LS, GE Medical Systems, Milwaukee, USA) was installed at the University Hospital of Zurich in Switzerland. This novel inline system consists of a dedicated high resolution/ high sensitivity full-ring PET scanner (Advance Nxi) and a very fast high-end multi-detector spiral CT scanner (Light Speed Plus). The axes of both systems are mechanically aligned so that simple translation of the patient table by 60 cm moves the patient from the CT into the PET gantry. The resulting PET and CT images are ‘‘hardware’’ coregistered to an accuracy in the range of 1 mm. Data acquisition in the combined system is as follows. At 45–55 min post-injection of FDG and after bladder emptying a multidetector computed tomography scan is performed from the pelvic floor to the head (scan length, 86.7 cm) and normal end-expiration. Immediately after CT scanning, a PET emission scan is obtained covering the identical axial field of view using six PET fields of view. The acquisition time is 4 min at each table position. With this protocol it is The British Journal of Radiology, 75 (2002), S36–S38 E 2002 The British Institute of Radiology

Denoising computed tomography (CT) medical images is crucial in preserving information and restoring images contaminated with noise. Standard filters have extensively been used for noise removal and fine details’ preservation. During the transmission of medical images, noise degrades the visibility of anatomical structures and subtle abnormalities, making it difficult for radiologists to accurately diagnose and interpret medical conditions. In recent studies, an optimum denoising filter using the wavelet threshold and deep-CNN was used to eliminate Gaussian noise in CT images using the image quality index (IQI) and peak signal-to-noise ratio (PSNR). Although the results were better than those with traditional techniques, the performance resulted in a loss of clarity and fine details’ preservation that rendered the CT images unsuitable. To address these challenges, this paper focuses on eliminating noise in CT scan images corrupted with additive Gaussian blur noise (AGBN) using an ensemble approach that integrates anisotropic Gaussian filter (AGF) and wavelet transform with a deep learning denoising convolutional neural network (DnCNN). First, the noisy image is denoised by AGF and Haar wavelet transform as preprocessing operations to eliminate AGBN. The DnCNN is then combined with AGF and wavelet for post-processing operation to eliminate the rest of the noises. Specifically, we used AGF due to its adaptability to edge orientation and directional information, which prevents blurring along edges for non-uniform noise distribution. Denoised images are evaluated using PSNR, mean squared error (MSE), and the structural similarity index measure (SSIM). Results revealed that the average PSNR value of the proposed ensemble approach is 28.28, and the average computational time is 0.01666 s. The implication is that the MSE between the original and reconstructed images is very low, implying that the image is restored correctly. Since the SSIM values are between 0 and 1.0, 1.0 perfectly matches the reconstructed image with the original image. In addition, the SSIM values at 1.0 or near 1.0 implicitly reveal a remarkable structural similarity between the denoised CT image and the original image. Compared to other techniques, the proposed ensemble approach has demonstrated exceptional performance in maintaining the quality of the image and fine details’ preservation.

Lung cancer is a leading cause of death in human globally. Owing to the low survival rates among lung cancer patients, it is essential to detect and treat cancer at an early stage. In some countries, positron emission tomography (PET)/X-ray computed tomography (CT) examination is also used for the cancer screening in addition to diagnosis and follow-up of treatment. PET/CT images provide both anatomical and functional information of the lung cancer. However, radiologists must examine a large number of these images and therefore, support tools for the localization of lung nodule are desired. This chapter highlights our recent contributions to a hybrid detection scheme of lung nodules in PET/CT images. In the CT image, a massive region is first detected using a cylindrical nodule enhancement filter (CNEF), which is a cylindrical kernel shaped by contrast enhancement filter. Subsequently, high-uptake regions detected by the PET images are merged with the region detected by the CT image. False positives (FPs) among the leading candidates are eliminated by a rule-based classifier and three support vector machines based on the characteristic features obtained from CT and PET images. Experimentally, the detection capability was evaluated using 100 cases of PET/CT images. As a result, the sensitivity in detecting candidates was 83%, with 5 FPs/case. These results indicate that the proposed hybrid method may be useful for the computer-aided detection of lung cancer in clinical practice.

The use of MRI-guided radiation therapy (MRIgRT) has shown considerable advantages. However, the acquisition of electron density information still relies on Computed tomography (CT) images. In the context of medical image synthesis, achieving high global and local accuracy, as well as efficient synthetic CT (sCT) generation, remains a key challenge. This study aims to improve the accuracy of sCT image generation from MRI by introducing a novel deep learning model that leverages global context and local detail, with a minimal increase in model complexity. In this study, we propose a novel generative adversarial network based on Mamba block and Residual Constraint (RC) strategy, namely RC-MambaGAN, which achieves a significant improvement in accuracy with only a minimal increase in computation time required for sCT generation. Specifically, we integrate Mamba block into the generator to enhance the model's ability to capture long-range contextual information, while preserving the precision of local features and maintaining the global structure of the image. The RC strategy improves the alignment between sCT and real CT by minimizing residuals. The model was validated on multi-center datasets that included both tumor and pelvic-region data. Furthermore, the dosimetric accuracy of the generated sCT images was evaluated to assess their clinical applicability. RC-MambaGAN achieved high performance with mean absolute error (MAE) of 37.094±8.761 HU, peak signal-to-noise ratio (PSNR) of 28.424±1.145dB, structural similarity index measure (SSIM) of 0.947±0.004, and Mutual Information (MI) of 1.426±0.012. External validation demonstrated that RC-MambaGAN outperformed existing state-of-the-art methods in both image quality and quantitative accuracy, even for structures visible in MRI but not in CT images. The proposed RC-MambaGAN markedly improves the image quality and anatomical accuracy of sCT images derived from MRI, thereby advancing the feasibility of MRI-only radiotherapy workflows.

PurposeThe purpose of this study was to evaluate the accuracy and perceptual quality of red-green-blue (RGB) fundus images generated from red-green (RG) images using Pix2Pix-based generative adversarial network (GAN).MethodsRG images were extracted from original RGB images recorded by a newer model of the Optos system by turning off the blue laser source. A conditional GAN based on the Pix2Pix architecture was trained to convert this extracted RG image to an RGB image. The performance was evaluated quantitatively by using a five-fold cross-validation and assessments with structural similarity index measure (SSIM), peak signal-to-noise ratio (PSNR), learned perceptual image patch similarity (LPIPS), mean absolute error (MAE), and root mean square error (RMSE). To assess the perceptual accuracy, 39 ophthalmologists examined 60 randomly selected images, 30 true and 30 generated RGB images, using Google Form-based interface. They were instructed to classify each image as a true or an artificial intelligence (AI)-generated image.ResultsThe generated images had high similarity to the original RGB images (SSIM = 0.97 ± 0.00; PSNR = 35.74 ± 0.17 decibel [dB]; LPIPS = 0.09 ± 0.00; MAE = 1.76 ± 0.04; and RMSE = 4.25 ± 0.10). The overall correct classification rate was 57.1%. Certain lesion types, such as epiretinal membrane and optic disc cupping, had higher discrimination rates. The receiver operating characteristic (ROC) analysis across all evaluations yielded an area under the curve (AUC) of 0.497 (P = 0.314).ConclusionsPix2Pix-based GANs can generate perceptually and structurally consistent RGB images from RG images without lesion-specific attention mechanisms.Translational RelevanceThis method enables a reinterpretation of legacy RG fundus images and may support future AI diagnostic applications.

Medical image fusion (MIF) significantly improves diagnosis accuracy by combining complementary data from several imaging modalities. However, the conventional fusion methods often fail to maintain the edge details, uniformity of texture and most relevant characteristics. To overcome these challenges, a novel multi-modality medical image fusion using Triple generator network (MIMO-TGAN) is proposed for accurate brain abnormality detection. The proposed MIMO-TGAN uses the multi-modality images such as Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Positron Emission Tomography (PET) images for MIF process. Initially, the gathered MRI and CT images are denoised with scalable range-based adaptive trilateral (SCRAT) filter and PET images are processed by multi-scale retinex (MSR) technique. The proposed MIMO-TGAN model contains three generators with a discriminator and an abnormality detection phase. The Generator-1 consists of grouped convolutional with different attention blocks for retrieving the relevant features from MRI and CT images to generate a synthesized image. Simultaneously, Generator-2 retrieves the most important voxel features from PET images and MRI images to generate another synthesized image. Finally, Generator-3 is designed to combine the both synthesized images from two generators by using four fusion rules to accurately detect brain abnormalities. The experimental evaluations demonstrate that the proposed MIMO-TGAN achieves an accuracy of 99.43% and 98.24% for dataset-1and dataset-2 respectively. Moreover, the proposed MIMO-TGAN achieves the entropy values of 7.05 and 6.84 for dataset-1 and dataset-2, and overall SSIM of 0.94 that indicates superior fusion quality results compared to existing methods.