Shaode Yu

Nonrigid Registration of Lung CT Images Based on Tissue Features

Nonrigid image registration is a prerequisite for various medical image process and analysis applications. Much effort has been devoted to thoracic image registration due to breathing motion. Recently, scale-invariant feature transform (SIFT) has been used in medical image registration and obtained promising results. However, SIFT is apt to detect blob features. Blobs key points are generally detected in smooth areas which may contain few diagnostic points. In general, diagnostic points used in medical image are often vessel crossing points, vascular endpoints, and tissue boundary points, which provide abundant information about vessels and can reflect the motion of lungs accurately. These points generally have high gradients as opposed to blob key points and can be detected by Harris. In this work, we proposed a hybrid feature detection method which can detect tissue features of lungs effectively based on Harris and SIFT. In addition, a novel method which can remove mismatched landmarks is also proposed. A series of thoracic CT images are tested by using the proposed algorithm, and the quantitative and qualitative evaluations show that our method is statistically significantly better than conventional SIFT method especially in the case of large deformation of lungs during respiration.

An edge-directed interpolation method for fetal spine MR images

BackgroundFetal spinal magnetic resonance imaging (MRI) is a prenatal routine for proper assessment of fetus development, especially when suspected spinal malformations occur while ultrasound fails to provide details. Limited by hardware, fetal spine MR images suffer from its low resolution.High-resolution MR images can directly enhance readability and improve diagnosis accuracy. Image interpolation for higher resolution is required in clinical situations, while many methods fail to preserve edge structures. Edge carries heavy structural messages of objects in visual scenes for doctors to detect suspicions, classify malformations and make correct diagnosis. Effective interpolation with well-preserved edge structures is still challenging.MethodIn this paper, we propose an edge-directed interpolation (EDI) method and apply it on a group of fetal spine MR images to evaluate its feasibility and performance. This method takes edge messages from Canny edge detector to guide further pixel modification. First, low-resolution (LR) images of fetal spine are interpolated into high-resolution (HR) images with targeted factor by bi-linear method. Then edge information from LR and HR images is put into a twofold strategy to sharpen or soften edge structures. Finally a HR image with well-preserved edge structures is generated. The HR images obtained from proposed method are validated and compared with that from other four EDI methods. Performances are evaluated from six metrics, and subjective analysis of visual quality is based on regions of interest (ROI).ResultsAll these five EDI methods are able to generate HR images with enriched details. From quantitative analysis of six metrics, the proposed method outperforms the other four from signal-to-noise ratio (SNR), peak signal-to-noise ratio (PSNR), structure similarity index (SSIM), feature similarity index (FSIM) and mutual information (MI) with seconds-level time consumptions (TC). Visual analysis of ROI shows that the proposed method maintains better consistency in edge structures with the original images.ConclusionsThe proposed method classifies edge orientations into four categories and well preserves structures. It generates convincing HR images with fine details and is suitable in real-time situations. Iterative curvature-based interpolation (ICBI) method may result in crisper edges, while the other three methods are sensitive to noise and artifacts.

Feature and contrast enhancement of mammographic image based on multiscale analysis and morphology.

A new algorithm for feature and contrast enhancement of mammographic images is proposed in this paper. The approach is based on multiscale transform and mathematical morphology. First of all, the Laplacian Gaussian pyramid transform is applied to decompose the mammography into different multiscale subband sub-images. In addition, the detail or high frequency sub-images are equalized by the contrast limited adaptive histogram equalization (CLAHE) and low frequency sub-images are processed by mathematical morphology. Finally, the enhanced image of feature and contrast is reconstructed from the Laplacian Gaussian pyramid coefficients modified at one or more levels by CLAHE and mathematical morphology. The enhanced image is processed by global non-linear operator in order to obtain natural result. The experimental results show that the presented algorithm is effective for feature and contrast enhancement of mammogram. The performance evaluation of the proposed algorithm is contrast evaluation criterion for image, signal-noise-ratio (SNR) and contrast improvement index (CII).

Performance evaluation of edge-directed interpolation methods for noise-free images

Many interpolation methods have been developed for high visual quality, but fail for preserving image structures. Edges carry heavy structural messages for visual tasks. Importance of edge preservation imposes edge-directed interpolation (EDI) methods a center of focus. How to measure edge-preserving ability has not been mentioned. In this paper, two metrics are proposed to measure the ability by edge-preserving ratio from accuracy and robustness. Performance of four edge-directed interpolation with two traditional methods are evaluated on two groups of standard images with other six commonly-used metrics. Experimental results show that EDI methods are better than traditional methods with highly improved edge-preserving ratio.

A shallow convolutional neural network for blind image sharpness assessment

Blind image quality assessment can be modeled as feature extraction followed by score prediction. It necessitates considerable expertise and efforts to handcraft features for optimal representation of perceptual image quality. This paper addresses blind image sharpness assessment by using a shallow convolutional neural network (CNN). The network takes single feature layer to unearth intrinsic features for image sharpness representation and utilizes multilayer perceptron (MLP) to rate image quality. Different from traditional methods, CNN integrates feature extraction and score prediction into an optimization procedure and retrieves features automatically from raw images. Moreover, its prediction performance can be enhanced by replacing MLP with general regression neural network (GRNN) and support vector regression (SVR). Experiments on Gaussian blur images from LIVE-II, CSIQ, TID2008 and TID2013 demonstrate that CNN features with SVR achieves the best overall performance, indicating high correlation with human subjective judgment.

Iterative image-domain ring artifact removal in cone-beam CT

Ring artifacts in cone beam computed tomography (CBCT) images are caused by pixel gain variations using flat-panel detectors, and may lead to structured non-uniformities and deterioration of image quality. The purpose of this study is to propose a method of general ring artifact removal in CBCT images. This method is based on the polar coordinate system, where the ring artifacts manifest as stripe artifacts. Using relative total variation, the CBCT images are first smoothed to generate template images with fewer image details and ring artifacts. By subtracting the template images from the CBCT images, residual images with image details and ring artifacts are generated. As the ring artifact manifests as a stripe artifact in a polar coordinate system, the artifact image can be extracted by mean value from the residual image; the image details are generated by subtracting the artifact image from the residual image. Finally, the image details are compensated to the template image to generate the corrected images. The proposed framework is iterated until the differences in the extracted ring artifacts are minimized. We use a 3D Shepp-Logan phantom, Catphan©504 phantom, uniform acrylic cylinder, and images from a head patient to evaluate the proposed method. In the experiments using simulated data, the spatial uniformity is increased by 1.68 times and the structural similarity index is increased from 87.12% to 95.50% using the proposed method. In the experiment using clinical data, our method shows high efficiency in ring artifact removal while preserving the image structure and detail. The iterative approach we propose for ring artifact removal in cone-beam CT is practical and attractive for CBCT guided radiation therapy.

Applications of edge preservation ratio in image processing

Edge preservation ratio (EPR) is a full-reference metric for objective image quality assessment (IQA). It is under the assumption that key messages to human visual systems are mainly from image structures, and these structures can be extracted by edge detection. EPR measure is twofold: accuracy and robustness, and a color map is synthesized to reveal structure changes before and after image processing. The feasibility and superiority of EPR have been validated via image magnification and noise reduction. Experimental results suggest that: (1) it is challenging to fully recover lost messages by image magnification; (2) high image contrast may be derived from concise and distinct image structures.

Efficient Segmentation of a Breast in B-Mode Ultrasound Tomography Using Three-Dimensional GrabCut (GC3D)

As an emerging modality for whole breast imaging, ultrasound tomography (UST), has been adopted for diagnostic purposes. Efficient segmentation of an entire breast in UST images plays an important role in quantitative tissue analysis and cancer diagnosis, while major existing methods suffer from considerable time consumption and intensive user interaction. This paper explores three-dimensional GrabCut (GC3D) for breast isolation in thirty reflection (B-mode) UST volumetric images. The algorithm can be conveniently initialized by localizing points to form a polygon, which covers the potential breast region. Moreover, two other variations of GrabCut and an active contour method were compared. Algorithm performance was evaluated from volume overlap ratios (TO, target overlap; MO, mean overlap; FP, false positive; FN, false negative) and time consumption. Experimental results indicate that GC3D considerably reduced the work load and achieved good performance (TO = 0.84; MO = 0.91; FP = 0.006; FN = 0.16) within an average of 1.2 min per volume. Furthermore, GC3D is not only user friendly, but also robust to various inputs, suggesting its great potential to facilitate clinical applications during whole-breast UST imaging. In the near future, the implemented GC3D can be easily automated to tackle B-mode UST volumetric images acquired from the updated imaging system.

Edge preservation ratio for image sharpness assessment

Image sharpness is one of the most determining factors for image readability and scene understanding. How to accurately quantify it is a hot topic. This paper systematically validates a previously proposed index for full-reference image sharpness assessment (edge preservation ratio, EPR). Based on Gaussian blurring images in LIVE, CSIQ, TID2008 and TID2013 databases, we firstly evaluated EPR accuracy on five edge detectors on LIVE and selected an optimal one for further analysis. Then nine state-of-the-art image quality assessment metrics are compared, including full-reference, no-reference and dedicated image sharpness assessment categories. Experimental results demonstrate (1) Canny is an optimal edge detector for EPR implementation; (2) EPR is a top-ranking image sharpness assessment metric that outperforms PSNR and SSIM and rivals FSIM; and (3) EPR accords more closely with human subjective judgment than involved image sharpness assessment metrics. This study also indicates that image sharpness assessment is still full of challenges and utilizing deep learning architectures to learning the direct mapping from images to quality will be a trend in the near future.

CNN-GRNN for Image Sharpness Assessment

Image sharpness is key to readability and scene understanding. Because of the inaccessible reference information, blind image sharpness assessment (BISA) is useful and challenging. In this paper, a shallow convolutional neural network (CNN) is proposed for intrinsic representation of image sharpness and general regression neural network (GRNN) is utilized for precise score prediction. The hybrid CNN-GRNN model tends to build functional relationship between retrieved features and subjective human scores by supervised learning. Superior to traditional algorithms based on handcrafted features and machine learning, CNN-GRNN fuses feature extraction and score prediction into an optimization procedure. Experiments on Gaussian blurring images in LIVE, CSIQ, TID2008 and TID2013 show that CNN-GRNN outperforms the state-of-the-art algorithms and gets closer to human subjective judgment.