Super-Resolution Imaging of Mammograms Based on the Super-Resolution Convolutional Neural Network

Purpose: To apply and evaluate a super-resolution scheme based on the super-resolution convolutional neural network (SRCNN) for enhancing image resolution in digital mammograms. Materials and Methods: A total of 711 mediolateral oblique (MLO) images including breast lesions were sampled from the Curated Breast Imaging Subset of the Digital Database for Screening Mammography (CBIS-DDSM). We first trained the super-resolution convolutional neural network (SRCNN), which is a deep-learning based super-resolution method. Using this trained SRCNN, high-resolution images were reconstructed from low-resolution images. We compared the image quality of the super-resolution method and that obtained using the linear interpolation methods (nearest neighbor and bilinear interpolations). To investigate the relationship between the image quality of the SRCNN-processed images and the clinical features of the mammographic lesions, we compared the image quality yielded by implementing the SRCNN, in terms of the breast density, the Breast Imaging-Reporting and Data System (BI-RADS) assessment, and the verified pathology information. For quantitative evaluation, peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) were measured to assess the image restoration quality and the perceived image quality. Results: The super-resolution image quality yielded by the SRCNN was significantly higher than that obtained using linear interpolation methods (p p Conclusion: SRCNN can significantly outperform conventional interpolation methods for enhancing image resolution in digital mammography. SRCNN can significantly improve the image quality of magnified mammograms, especially in dense breasts, high-risk BI-RADS assessment groups, and pathology-verified malignant cases.

Perivascular Fat Attenuation Index on Non-Contrast-Enhanced Cardiac Computed Tomography: Comparison with Coronary Computed Tomography Angiography

Objective: Perivascular fat attenuation index (FAI) measurement on non-contrast-enhanced cardiac computed tomography (NCCT) has not been rigorously validated in previous studies. Herein, we compared perivascular FAI values between NCCT and coronary computed tomography angiography (CCTA). We also investigated the variability and reproducibility of perivascular FAI measurement. Methods: A total of 44 patients who underwent NCCT and CCTA were included in this study. For NCCT, perivascular FAI was measured using three threshold settings: from −30 to −190 Hounsfield Units (HU), −20 to −180 HU, and −10 to −170 HU. For CCTA, perivascular FAI was measured using one threshold setting: from −30 to −190 HU. Perivascular FAI measurements by NCCT were compared with those by CCTA using the paired t-test, and correlations were assessed using Pearson’s correlation coefficient. The intra- and inter-observer variabilities for the measurements with NCCT and CCTA were evaluated with the intraclass correlation coefficient. Results: Perivascular FAI measurements with the threshold setting of −30 to −190 HU were significantly lower on NCCT than on CCTA. There were no significant differences between the perivascular FAI measurements at the remaining thresholds on NCCT and those on CCTA. The perivascular FAI at all thresholds on NCCT correlated significantly with those on CCTA. The intra- and inter-observer agreements were excellent for the measurements on NCCT and CCTA. Conclusion: There were significant differences between the perivascular FAI measurements on NCCT and CCTA. However, the differences could be modified by threshold adjustment.

Performance Evaluation of Super-Resolution Methods Using Deep-Learning and Sparse-Coding for Improving the Image Quality of Magnified Images in Chest Radiographs

Purpose: To detect small diagnostic signals such as lung nodules in chest radiographs, radiologists magnify a region-of-interest using linear interpolation methods. However, such methods tend to generate over-smoothed images with artifacts that can make interpretation difficult. The purpose of this study was to investigate the effectiveness of super-resolution methods for improving the image quality of magnified chest radiographs. Materials and Methods: A total of 247 chest X-rays were sampled from the JSRT database, then divided into 93 training cases with non-nodules and 154 test cases with lung nodules. We first trained two types of super-resolution methods, sparse-coding super-resolution (ScSR) and super-resolution convolutional neural network (SRCNN). With the trained super-resolution methods, the high-resolution image was then reconstructed using the super-resolution methods from a low-resolution image that was down-sampled from the original test image. We compared the image quality of the super-resolution methods and the linear interpolations (nearest neighbor and bilinear interpolations). For quantitative evaluation, we measured two image quality metrics: peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). For comparative evaluation of the super-resolution methods, we measured the computation time per image. Results: The PSNRs and SSIMs for the ScSR and the SRCNN schemes were significantly higher than those of the linear interpolation methods (p p p Conclusion: Super-resolution methods provide significantly better image quality than linear interpolation methods for magnified chest radiograph images. Of the two tested schemes, the SRCNN scheme processed the images fastest;thus, SRCNN could be clinically superior for processing radiographs in terms of both image quality and processing speed.

Application of Sparse-Coding Super-Resolution to 16-Bit DICOM Images for Improving the Image Resolution in MRI

Purpose: To improve the image resolution of magnetic resonance imaging (MRI), conventional interpolation methods are commonly used to magnify images via various image processing approaches;however, these methods tend to produce artifacts. While super-resolution (SR) schemes have been introduced as an alternative approach to apply medical imaging, previous studies applied SR only to medical images in 8-bit image format. This study aimed to evaluate the effectiveness of sparse-coding super-resolution (ScSR) for improving the image quality of reconstructed high-resolution MR images in 16-bit digital imaging and communications in medicine (DICOM) image format. Materials and Methods: Fifty-nine T1-weighted images (T1), 84 T2-weighted images (T2), 85 fluid attenuated inversion recovery (FLAIR) images, and 30 diffusion-weighted images (DWI) were sampled from The Repository of Molecular Brain Neoplasia Data as testing datasets, and 1307 non-medical images were sampled from the McGill Calibrated Color Image Database as a training dataset. We first trained the ScSR to prepare dictionaries, in which the relationship between low- and high-resolution images was learned. Using these dictionaries, a high-resolution image was reconstructed from a 16-bit DICOM low-resolution image downscaled from the original test image. We compared the image quality of ScSR and 4 interpolation methods (nearest neighbor, bilinear, bicubic, and Lanczos interpolations). For quantitative evaluation, we measured the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). Results: The PSNRs and SSIMs for the ScSR were significantly higher than those of the interpolation methods for all 4 MRI sequences (PSNR: p p Conclusion: ScSR provides significantly higher image quality in terms of enhancing the resolution of MR images (T1, T2, FLAIR, and DWI) in 16-bit DICOM format compared to the interpolation methods.