Search

Pansharpening is a crucial and challenging task that aims to obtain a high spatial resolution image by merging a multispectral (MS) image and a panchromatic (PAN) image. Current methods use CNNs with standard convolution, but we've observed strong correlation among channel dimensions in the kernel, leading to computational burden and redundancy. To address this, we propose Learnable Gaussian Perturbation Convolution (LGPConv), surpassing standard convolution. LGPConv leverages two properties of standard convolution kernels: 1) correlations within channels, learning a premier kernel as a base to reduce parameters and training difficulties caused by redundancy; 2) introducing Gaussian noise perturbations to simulate randomness and enhance nonlinear representation within channels. We incorporate LGPConv into a well-designed pansharpening network and demonstrate its superiority through extensive experiments, achieving state-of-the-art performance with minimal parameters (27K). Code is available on the GitHub page of the authors.

PurposeMagnetization transfer (MT) MRI can be effective for the diagnosis of a broad range of fibrotic diseases, including liver fibrosis. However, respiratory motion, a major source of artifacts in thoracic and abdominal MR imaging, can obscure important anatomic structures, making diagnosis difficult. In this study, we explored the potential to combine free-breathing (FB) respiratory self-gating (RSG) methods with MT saturation for FB MT ratio (MTR) measurements of abdominal organs.MethodsA respiratory self-gated multiple-gradient recalled echo sequence with MT presaturation (RSG-MT GRE) was developed and applied in a series of seven normal volunteers. We compared the MTR values of liver, pancreas, kidney, spleen, and posterior paraspinal muscle measured using our RSG-MT GRE sequence and a conventional MT GRE sequence.ResultsRSG consistently reduced motion artifacts within MT-weighted images acquired during FB, improved the accuracy of FB MTR measurements, and produced comparable MTRs to breath-holding MTR measurements.ConclusionRSG approaches may offer to improve the utility of MT-weighted imaging methods for the assessment of fibrotic diseases and tumor desmoplasia in abdominal organs.

BackgroundWithout the need of contrast media, diffusion-weighted imaging (DWI) has shown great promise for accurate detection of lipid-rich necrotic core (LRNC), a well-known feature of vulnerable plaques. However, limited resolution and poor image quality in vivo with conventional single-shot diffusion-weighted echo planar imaging (SS-DWEPI) has hindered its clinical application. The aim of this work is to develop a diffusion-prepared turbo-spin-echo (DP-TSE) technique for carotid plaque characterization with 3D high resolution and improved image quality.MethodsUnlike SS-DWEPI where the diffusion encoding is integrated in the EPI framework, DP-TSE uses a diffusion encoding module separated from the TSE framework, allowing for segmented acquisition without the sensitivity to phase errors. The interleaved, motion-compensated sequence was designed to enable 3D black-blood DWI of carotid arteries with sub-millimeter resolution. The sequence was tested on 12 healthy subjects and compared with SS-DWEPI for image quality, vessel wall visibility, and vessel wall thickness measurements. A pilot study was performed on 6 patients with carotid plaques using this sequence and compared with conventional contrast-enhanced multi-contrast 2D TSE as the reference.ResultsDP-TSE demonstrated advantages over SS-DWEPI for resolution and image quality. In the healthy subjects, vessel wall visibility was significantly higher with diffusion-prepared TSE (p < 0.001). Vessel wall thicknesses measured from diffusion-prepared TSE were on average 35% thinner than those from the EPI images due to less distortion and partial volume effect (p < 0.001). ADC measurements of healthy carotid vessel wall are 1.53 ± 0.23 × 10-3 mm2/s. In patients the mean ADC measurements in the LRNC area were significantly lower (0.60 ± 0.16 × 10-3 mm2/s) than those of the fibrous plaque tissue (1.27 ± 0.29 × 10-3 mm2/s, p < 0.01).ConclusionsDiffusion-prepared CMR allows, for the first time, 3D DWI of the caro

BackgroundMulti-contrast weighted imaging is a commonly used cardiovascular magnetic resonance (CMR) protocol for characterization of carotid plaque composition. However, this approach is limited in several aspects including low slice resolution, long scan time, image mis-registration, and complex image interpretation. In this work, a 3D CMR technique, named Multi-contrast Atherosclerosis Characterization (MATCH), was developed to mitigate the above limitations.MethodsMATCH employs a 3D spoiled segmented fast low angle shot readout to acquire data with three different contrast weightings in an interleaved fashion. The inherently co-registered image sets, hyper T1-weighting, gray blood, and T2-weighting, are used to detect intra-plaque hemorrhage (IPH), calcification (CA), lipid-rich necrotic core (LRNC), and loose-matrix (LM). The MATCH sequence was optimized by computer simulations and testing on four healthy volunteers and then evaluated in a pilot study of six patients with carotid plaque, using the conventional multi-contrast protocol as a reference.ResultsOn MATCH images, the major plaque components were easy to identify. Spatial co-registration between the three image sets with MATCH was particularly helpful for the reviewer to discern co-existent components in an image and appreciate their spatial relation. Based on Cohen's kappa tests, moderate to excellent agreement in the image-based or artery-based component detection between the two protocols was obtained for LRNC, IPH, CA, and LM, respectively. Compared with the conventional multi-contrast protocol, the MATCH protocol yield significantly higher signal contrast ratio for IPH (3.1±1.3 vs. 0.4±0.3, p<0.001) and CA (1.6±1.5 vs. 0.7±0.6, p=0.012) with respect to the vessel wall.ConclusionsTo the best of our knowledge, the proposed MATCH sequence is the first 3D CMR technique that acquires spatially co-registered multi-contrast image sets in a single scan for characterization of carotid plaque composition.

BackgroundWithout the need of contrast media, diffusion-weighted imaging (DWI) has shown great promise for accurate detection of lipid-rich necrotic core (LRNC), a well-known feature of vulnerable plaques. However, limited resolution and poor image quality in vivo with conventional single-shot diffusion-weighted echo planar imaging (SS-DWEPI) has hindered its clinical application. The aim of this work is to develop a diffusion-prepared turbo-spin-echo (DP-TSE) technique for carotid plaque characterization with 3D high resolution and improved image quality.MethodsUnlike SS-DWEPI where the diffusion encoding is integrated in the EPI framework, DP-TSE uses a diffusion encoding module separated from the TSE framework, allowing for segmented acquisition without the sensitivity to phase errors. The interleaved, motion-compensated sequence was designed to enable 3D black-blood DWI of carotid arteries with sub-millimeter resolution. The sequence was tested on 12 healthy subjects and compared with SS-DWEPI for image quality, vessel wall visibility, and vessel wall thickness measurements. A pilot study was performed on 6 patients with carotid plaques using this sequence and compared with conventional contrast-enhanced multi-contrast 2D TSE as the reference.ResultsDP-TSE demonstrated advantages over SS-DWEPI for resolution and image quality. In the healthy subjects, vessel wall visibility was significantly higher with diffusion-prepared TSE (p < 0.001). Vessel wall thicknesses measured from diffusion-prepared TSE were on average 35% thinner than those from the EPI images due to less distortion and partial volume effect (p < 0.001). ADC measurements of healthy carotid vessel wall are 1.53 ± 0.23 × 10-3 mm2/s. In patients the mean ADC measurements in the LRNC area were significantly lower (0.60 ± 0.16 × 10-3 mm2/s) than those of the fibrous plaque tissue (1.27 ± 0.29 × 10-3 mm2/s, p < 0.01).ConclusionsDiffusion-prepared CMR allows, for the first time, 3D

BackgroundMulti-contrast weighted imaging is a commonly used cardiovascular magnetic resonance (CMR) protocol for characterization of carotid plaque composition. However, this approach is limited in several aspects including low slice resolution, long scan time, image mis-registration, and complex image interpretation. In this work, a 3D CMR technique, named Multi-contrast Atherosclerosis Characterization (MATCH), was developed to mitigate the above limitations.MethodsMATCH employs a 3D spoiled segmented fast low angle shot readout to acquire data with three different contrast weightings in an interleaved fashion. The inherently co-registered image sets, hyper T1-weighting, gray blood, and T2-weighting, are used to detect intra-plaque hemorrhage (IPH), calcification (CA), lipid-rich necrotic core (LRNC), and loose-matrix (LM). The MATCH sequence was optimized by computer simulations and testing on four healthy volunteers and then evaluated in a pilot study of six patients with carotid plaque, using the conventional multi-contrast protocol as a reference.ResultsOn MATCH images, the major plaque components were easy to identify. Spatial co-registration between the three image sets with MATCH was particularly helpful for the reviewer to discern co-existent components in an image and appreciate their spatial relation. Based on Cohen's kappa tests, moderate to excellent agreement in the image-based or artery-based component detection between the two protocols was obtained for LRNC, IPH, CA, and LM, respectively. Compared with the conventional multi-contrast protocol, the MATCH protocol yield significantly higher signal contrast ratio for IPH (3.1±1.3 vs. 0.4±0.3, p<0.001) and CA (1.6±1.5 vs. 0.7±0.6, p=0.012) with respect to the vessel wall.ConclusionsTo the best of our knowledge, the proposed MATCH sequence is the first 3D CMR technique that acquires spatially co-registered multi-contrast image sets in a single scan for characterization of carotid plaque composition.