Real-time automatic extraction of lumen region and boundary from endoscopic images

A new approach to the automatic extraction of the lumen region and its boundary for gastrointestinal (GI) endoscopic images is presented. First, a quasi region of interest, the darker regions of the image, is segmented using a region splitting scheme termed progressive thresholding. The centre of mass of this segmented region acts as a seed for further processing. Then the lumen region is obtained using a region growing technique called the integrated neighbourhood search (INS). A new quad structure based technique is introduced to enhance the INS speed significantly. A back projection algorithm is suggested to optimise the search for pixels belonging to the lumen region and boundary. A boundary-thinning algorithm is also proposed to remove the redundant pixels from the lumen boundary and to generate a connected single pixel width boundary. The proposed approach does not need a priori knowledge about the image characteristics. The experimental results indicate that the proposed technique enhances the speed of conventional INS by 45.5% to 28.6% based on the lumen size varying from 22 709 pixels to 4947 pixels. The main advantage of the proposed technique is its high-speed response that facilitates real-time analysis of endoscopic images.     

Real-time automatic extraction of lumen region and boundary from endoscopic images

A new approach to the automatic extraction of the lumen region and its boundary for gastrointestinal (GI) endoscopic images is presented. First, a quasi region of interest, the darker regions of the image, is segmented using a region splitting scheme termed progressive thresholding. The centre of mass of this segmented region acts as a seed for further processing. Then the lumen region is obtained using a region growing technique called the integrated neighbourhood search (INS). A new quad structure based technique is introduced to enhance the INS speed significantly. A back projection algorithm is suggested to optimise the search for pixels belonging to the lumen region and boundary. A boundary-thinning algorithm is also proposed to remove the redundant pixels from the lumen boundary and to generate a connected single pixel width boundary. The proposed approach does not need a priori knowledge about the image characteristics. The experimental results indicate that the proposed technique enhances the speed of conventional INS by 45.5% to 28.6% based on the lumen size varying from 22,709 pixels to 4947 pixels. The main advantage of the proposed technique is its high-speed response that facilitates real-time analysis of endoscopic images.     

胸部CT图像中感兴趣区域的提取与量化分析

目的：根据临床应用需求，研究了胸部高分辨率CT图像中感兴趣区域（region of interest，ROI）的提取与量化诊断问题。方法：首先由人工勾勒感兴趣区域边界，再应用Bresenham扫描线算法生成连续的区域边界，然后，应用基于四邻域的背景标记扫描线方法，对区域外像素作出标记，从而得到选定区域。最后，计算区域的量化参数，并根据肺气肿量化诊断标准，对感兴趣区域进行分析与辅助诊断。结果：计算得到肺气肿占整个肺部容积的百分比为39．2％，该患者属于3级重度肺气肿。结论：实验证明，该方法能快速、准确地提取任意形状的区域，并对给定区域进行统计分析，非常有利于医生的准确诊断。     

Broadband ultrasound attenuation imaging: algorithm development and clinical assessment of a region growing technique

This paper presents a computerized method for the selection of an irregular region of interest (ROI) in broadband ultrasound attenuation (BUA) images. A region growing algorithm searches an initial region in the posterior part of the calcaneus until the pixel with the lowest attenuation value is found; this is the starting seed. Then, the algorithm evaluates the values of the eight pixels neighbouring the starting seed. Pixels that have the closest value to the starting seed are accepted. This procedure is the first processing level. The procedure is repeated for the group of pixels neighbouring those accepted from the previous processing level. The algorithm ceases when the number of accepted pixels reaches a user-specified number. The clinical part of this study compares measurements of BUA at an automatic ROI implemented on a quantitative ultrasound imaging device, defined as the circular region of lowest attenuation in the posterior part of the calcaneus, and at irregular ROIs of various sizes generated by the algorithm developed in this study. The algorithm was applied to BUA images obtained from 24 post-menopausal women with hip fractures and 26 age-matched healthy female subjects. The use of the irregular ROI with a size of 2400 pixels is proposed because that region yielded better clinical results compared to irregular ROIs with different size and the circular automatic ROI.     

Extraction and reconstruction of retinal vasculature

Information about retinal vasculature morphology is used in grading the severity and progression of diabetic retinopathy. An image analysis system can help ophthalmologists make accurate and efficient diagnoses. This paper presents the development of an image processing algorithm for detecting and reconstructing retinal vasculature. The detection of the vascular structure is achieved by image enhancement using contrast limited adaptive histogram equalization followed by the extraction of the vessels using bottom-hat morphological transformation. For reconstruction of the complete retinal vasculature, a region growing technique based on first-order Gaussian derivative is developed. The technique incorporates both gradient magnitude change and average intensity as the homogeneity criteria that enable the process to adapt to intensity changes and intensity spread over the vasculature region. The reconstruction technique reduces the required number of seeds to near optimal for the region growing process. It also overcomes poor performance of current seed-based methods, especially with low and inconsistent contrast images as normally seen in vasculature regions of fundus images. Simulations of the algorithm on 20 test images from the DRIVE database show that it outperforms many other published methods and achieved an accuracy range (ability to detect both vessel and non-vessel pixels) of 0.91 – 0.95, a sensitivity range (ability to detect vessel pixels) of 0.91 – 0.95 and a specificity range (ability to detect non-vessel pixels) of 0.88 – 0.94.     

Extraction and reconstruction of retinal vasculature

Information about retinal vasculature morphology is used in grading the severity and progression of diabetic retinopathy. An image analysis system can help ophthalmologists make accurate and efficient diagnoses. This paper presents the development of an image processing algorithm for detecting and reconstructing retinal vasculature. The detection of the vascular structure is achieved by image enhancement using contrast limited adaptive histogram equalization followed by the extraction of the vessels using bottom-hat morphological transformation. For reconstruction of the complete retinal vasculature, a region growing technique based on first-order Gaussian derivative is developed. The technique incorporates both gradient magnitude change and average intensity as the homogeneity criteria that enable the process to adapt to intensity changes and intensity spread over the vasculature region. The reconstruction technique reduces the required number of seeds to near optimal for the region growing process. It also overcomes poor performance of current seed-based methods, especially with low and inconsistent contrast images as normally seen in vasculature regions of fundus images. Simulations of the algorithm on 20 test images from the DRIVE database show that it outperforms many other published methods and achieved an accuracy range (ability to detect both vessel and non-vessel pixels) of 0.91 - 0.95, a sensitivity range (ability to detect vessel pixels) of 0.91 - 0.95 and a specificity range (ability to detect non-vessel pixels) of 0.88 - 0.94.     

Segmentation automatique d’échographies trans-abdominales de prostate pour le recalage d''images

This article presents a method for automatic segmentation of prostate from abdominal freehand ultrasound images. A statistical model of prostate is estimated from a manually delineated images. The segmentation starts by smoothing the image to enhance edges by applying a morphological and adaptive filter which detects individual speckles and remove them, while it preserves valuable details. Then the boundary is initialised starting from the model and the final form is estimated by a simulated annealing optimisation algorithm. The performances of the algorithm were compared with manual segmentation by an expert, the average distance was 3.7 pixels and an overlap surface of 93%.     

The enhancement of radiotherapy verification images by an automated edge detection technique.

Adaptive histogram equalization techniques are known to be effective for the enhancement of contrast in portal images acquired during radiotherapy treatments. A significant drawback is the loss of definition on the edges of the treatment field. Analysis of this problem shows that it can be remedied by separating the treatment field from the background prior to the enhancement, and using only the pixels within the field boundary in the enhancement procedure. An edge extraction algorithm has been developed for delineating the treatment field in portal images, and consists of four modules that are applied to the original portal image in sequence. In the first step, edges are enhanced with a derivative of Gaussian operator that assures high response to the field edges relative to anatomical or other edges in the image. Pixels for which the response of the edge operator was the strongest are subsequently connected by an edge following algorithm to produce a raw contour of the field. In the last two steps the contour is refined by converting it into straight line segments and appending to the contour any parts of the field edge that might have been missed out during the initial edge following. The final contour encloses exclusively those pixels that belong to the treatment field, and the adaptive histogram equalization is applied selectively to this region. The combination of edge detection and selective enhancement was shown to produce images of superior contrast on the patient's anatomical features as well as accurate definition of treatment field edges.     

A fast and accurate segmentation technique for the extraction of gastrointestinal lumen from endoscopic images

A novel region-growing algorithm for the segmentation of endoscopic images is proposed in this paper. The objective of the research work is fast and accurate segmentation of gastrointestinal lumen from the endoscopic images for real-time applications. The proposed technique consists of a dual-step methodology in which a quasi Region of Interest (qROI) is segmented first using a global thresholding technique and then the actual lumen is extracted using differential region growing. An adaptive progressive thresholding technique is used to obtain qROI for a given endoscopic image. The centre of mass of qROI acts as the seed for the region growing in the next step. A differential region growing technique is suggested which grows the region on the basis of a similarity criterion. A dynamic hill-clustering method is utilised to ensure the effectiveness of the terminating condition during the growth process. The proposed scheme is faster than the conventional gradient based region-growing technique. The accuracy and high speed response of the proposed technique is validated with several endoscopic images and the results are presented.     

基于Graph Cuts的脑部MRI图像脑组织提取方法

脑部MRI图像的脑组织提取是神经影像学分析的一项重要预处理过程,为提高提取精度,提出一种基于graph cuts的脑组织自动提取方法,主要适用于T1加权MRI图像。首先采用Smith等提出的脑组织提取工具（BET）得到感兴趣区域 (ROI),仅在该区域内用graph cuts方法进行演化;并在graph cuts中加入一个速度限制因子,解决脑组织提取过程中的局部收敛和边界泄漏问题;此外,还采用一种逐层处理2D图像切片的3D数据初始化方法。利用IBSR(Internet Brain Segmentation Repository)网站提供的18组数据,将所提出方法与现有的部分脑组织提取方法（脑组织提取工具(BET)、 脑组织表面提取算法(BSE)、 分水岭算法(WAT)、 混合分水岭算法(HWAT)、 图割算法(GCUT) 和鲁棒脑组织提取算法(ROBEX)）,进行对比试验。结果显示,本方法最接近于标准分割,平均Dice系数达到095,并且在多个评价参数（假阳性率32%和Hausdorff距离96）上都取得最好结果。实验表明,所提出方法具有较好的精确性和稳定性。     

Accuracy and reproducibility of a subpixel extended phase correlation method to determine micron level displacements in the heart

Future treatment of heart disease may involve local perturbations of mechanical function, such as intramyocardial injections of angiogenic growth factors or progenitor cells. This necessitates an accurate measurement technique to determine regional heart function. We have previously developed a method to determine regional heart function using a phase correlation algorithm. However, in determining regional function over a single heartbeat it is necessary to sum displacements between many images. We have therefore incorporated a subpixel algorithm that models the result of phase correlation as a sinc function in order to increase the accuracy of our technique. This method, which we have named high density mapping (HDM), determines the subpixel displacement of 64 x 64 pixel regions from images of the heart. To determine the accuracy and precision of the technique, a high contrast image of a heart was digitally shifted 1, 2 or 3 pixels. The original and shifted images were then downsampled four times resulting in 0.25, 0.50 or 0.75 pixel shifts between the original and shifted images. The average accuracy of HDM in the digitally shifted images was 0.06 pixels, with a precision of 0.08 pixels. Effectiveness of HDM in characterization of deformation was also assessed in digitally stretched images. Error in quantification of strain was found to be less than 3.5% of the calculated strain. In an additional set of experiments, in which accuracy was determined using physical motion instead of digital shifting and downsampling, a speckle pattern was displaced by known distances using a micromanipulator, such that the displacement between the captured images was 0.5 pixels. These data demonstrated an accuracy of 0.09 pixels and a precision of 0.02 pixels. Finally, as HDM is used to determine the regional stroke work index (RSW) in beating hearts, the repeatability of using this method to compute RSW was assessed. RSW, the integral of intraventricular pressure with respect to unitless regional area, where end diastolic area was normalized to unity, was assessed in consecutive beats from four different hearts. The average standard deviation of RSW was 0.098 mmHg. Uncertainty analysis determined the maximum error of RSW to be +/-0.41 mmHg, approximately two-thirds of the measured biologic variability. These data demonstrate the ability of HDM to accurately and reproducibly measure displacement and regional function in the beating heart.     

A novel PET tumor delineation method based on adaptive region-growing and dual-front active contours

To more accurately and precisely delineate a tumor in a 3D PET image, we proposed a novel, semi-automatic, two-stage method by utilizing an adaptive region-growing algorithm and a dual-front active contour model. First, a rough region of interest (ROI) is manually drawn by a radiation oncologist that encloses a tumor. The voxel having the highest intensity in the ROI is chosen as a seed point. An adaptive region growing algorithm successively appends to the seed point all neighboring voxels whose intensities > = T of the mean of the current region. When T varies from 100% to 0%, a sharp volume increase, indicating the transition from the tumor to the background, always occurs at a certain T value. A preliminary tumor boundary is determined just before the sharp volume increase, which is found to be slightly outside of the known tumor in all tested phantoms. A novel dual-front active contour model utilizing region-based information is then applied to refine the preliminary boundary automatically. We tested the two-stage method on six spheres (0.5-20 ml) in a cylindrical container under different source to background ratios. Comparisons between the two-stage method and an iterative threshold method demonstrate its higher detection accuracy for small tumors (less than 6 ml). One patient study was tested and evaluated by two experienced radiation oncologists. The study illustrated that this two-stage method has several advantages. First, it does not require any threshold-volume curves, which are different and must be calibrated for each scanner and image reconstruction method. Second, it does not use any iso-threshold lines as contours. Third, the final result is reproducible and is independent of the manual rough ROIs. Fourth, this method is an adaptive algorithm that can process different images automatically.     

Tracking moving objects with megavoltage portal imaging: a feasibility study

Four different algorithms were investigated with the aim to determine their suitability to track an object in conventional megavoltage portal images. The algorithms considered were the mean of the sum of squared differences (MSSD), mutual information (MI), the correlation ratio (CR), and the correlation coefficient (CC). Simulation studies were carried out with various image series containing a rigid object of interest that was moved along a predefined trajectory. For each of the series the signal-to-noise ratio (SNR) was varied to compare the performance of the algorithms under noisy conditions. For a poor SNR of -6 dB the mean tracking error was 2.4, 6.5, 39.0, and 17.2 pixels for MSSD, CC, CR and MI, respectively, with a standard deviation of 1.9, 12.9, 19.5, and 7.5 pixels, respectively. The size of a pixel was 0.5 mm. These results improved to 1.1, 1.3, 1.3, and 2.0 pixels, respectively, with a standard deviation of 0.6, 0.8, 0.8, and 2.1 pixels, respectively, when a mean filter was applied to the images prior to tracking. The implementation of MSSD into existing in-house software demonstrated that, depending on the search range, it was possible to process between 2 and 15 images/s, making this approach capable of real-time applications. In conclusion, the best geometric tracking accuracy overall was obtained with MSSD, followed by CC, CR, and MI. The simplest and best algorithm, both in terms of geometric accuracy as well as computational cost, was the MSSD algorithm and was therefore the method of choice.     

Speckle reduction in ultrasound medical images using adaptive filter based on second order statistics

This article discusses an adaptive filtering technique for reducing speckle using second order statistics of the speckle pattern in ultrasound medical images. Several region-based adaptive filter techniques have been developed for speckle noise suppression, but there are no specific criteria for selecting the region growing size in the post processing of the filter. The size appropriate for one local region may not be appropriate for other regions. Selection of the correct region size involves a trade-off between speckle reduction and edge preservation. Generally, a large region size is used to smooth speckle and a small size to preserve the edges into an image. In this paper, a smoothing procedure combines the first order statistics of speckle for the homogeneity test and second order statistics for selection of filters and desired region growth. Grey level co-occurrence matrix (GLCM) is calculated for every region during the region contraction and region growing for second order statistics. Further, these GLCM features determine the appropriate filter for the region smoothing. The performance of this approach is compared with the aggressive region-growing filter (ARGF) using edge preservation and speckle reduction tests. The processed image results show that the proposed method effectively reduces speckle noise and preserves edge details.     

Speckle reduction in ultrasound medical images using adaptive filter based on second order statistics

This article discusses an adaptive filtering technique for reducing speckle using second order statistics of the speckle pattern in ultrasound medical images. Several region-based adaptive filter techniques have been developed for speckle noise suppression, but there are no specific criteria for selecting the region growing size in the post processing of the filter. The size appropriate for one local region may not be appropriate for other regions. Selection of the correct region size involves a trade-off between speckle reduction and edge preservation. Generally, a large region size is used to smooth speckle and a small size to preserve the edges into an image. In this paper, a smoothing procedure combines the first order statistics of speckle for the homogeneity test and second order statistics for selection of filters and desired region growth. Grey level co-occurrence matrix (GLCM) is calculated for every region during the region contraction and region growing for second order statistics. Further, these GLCM features determine the appropriate filter for the region smoothing. The performance of this approach is compared with the aggressive region-growing filter (ARGF) using edge preservation and speckle reduction tests. The processed image results show that the proposed method effectively reduces speckle noise and preserves edge details.     

超声医学图像滤波算法研究进展

主要讨论超声医学图像滤波算法的研究现状,几种主要的滤波方法（多方位滤波方法、自适应权值调节滤波方法、自适应窗口选取滤波方法、两步法等）面临的问题及发展的方向。作者通过实践,将有关算法应用于超声医学图像的处理,给出了处理结果,进行了几种算法的比较分析。     

An Algorithm for Tracking Microcatheters in Fluoroscopy

Currently, a large number of endovascular interventions are performed for treatment of intracranial aneurysms. For these treatments, correct positioning of microcatheter tips, microguide wire tips, or coils is essential. Techniques to detect such devices may facilitate endovascular interventions. In this paper, we describe an algorithm for tracking of microcatheter tips during fluoroscopically guided neuroendovascular interventions. A sequence of fluoroscopic images (1,024 × 1,024 × 12 bits) was acquired using a C-arm angiography system as a microcatheter was passed through a carotid phantom which was on top of a head phantom. The carotid phantom was a silicone cylinder containing a simulated vessel with the shape and curvatures of the internal carotid artery. The head phantom consisted of a human skull and tissue-equivalent material. To detect the microcatheter in a given fluoroscopic frame, a background image consisting of an average of the four previous frames is subtracted from the current frame, the resulting image is filtered using a matched filter, and the position of maximum intensity in the filtered image is taken as the catheter tip position in the current frame. The distance between the tracked position and the correct position (error distance) was measured in each of the fluoroscopic images. The mean and standard deviation of the error distance values were 0.277 mm (1.59 pixels) and 0.26 mm (1.5 pixels), respectively. The error distance was less than 3 pixels in the 93.0% frames. Although the algorithm intermittently failed to correctly detect the catheter, the algorithm recovered the catheter in subsequent frames.     

Cell contour extraction on multi-threshold images.

A procedure for the extraction of cell and nuclear contours from digital images is presented. The procedure is simple and suitable for efficient implementation in an interactive environment where the user selects a region of interest containing the cell to be segmented. Segmentation is facilitated by multi-thresholding where the threshold values are determined by an iterative algorithm using the mean square error criteria to cluster the image pixel values. Edges are detected on the threshold images by employing a binary morphological filter and the desired contours are selected by considering closed contours in the edge image. The cell region is first extracted by performing segmentation, edge detection and contour extraction on the region of interest. The nuclear region is subsequently extracted by reapplication of the algorithms within the extracted cell region. Segmentation examples are presented.     

Authentication and Data Hiding Using a Hybrid ROI-Based Watermarking Scheme for DICOM Images

Authenticating medical images using watermarking techniques has become a very popular area of research, and some works in this area have been reported worldwide recently. Besides authentication, many data-hiding techniques have been proposed to conceal patient’s data into medical images aiming to reduce the cost needed to store data and the time needed to transmit data when required. In this paper, we present a new hybrid watermarking scheme for DICOM images. In our scheme, two well-known techniques are combined to gain the advantages of both and fulfill the requirements of authentication and data hiding. The scheme divides the images into two parts, the region of interest (ROI) and the region of non-interest (RONI). Patient’s data are embedded into ROI using a reversible technique based on difference expansion, while tamper detection and recovery data are embedded into RONI using a robust technique based on discrete wavelet transform. The experimental results show the ability of hiding patient’s data with a very good visual quality, while ROI, the most important area for diagnosis, is retrieved exactly at the receiver side. The scheme also shows some robustness against certain levels of salt and pepper and cropping noise.     

基于深度学习的颈动脉超声图像内中膜厚度测量

目的 颈动脉血管内中膜厚度(IMT)是衡量动脉粥样硬化程度的重要标准.一般采用人工标定进行测量,该过程耗时且繁琐,由此提出一种总体性能较好的全自动分割(AS)算法.方法 该算法首先利用卷积神经网络(CNN)识别出颈动脉血管远端,进而提取包含颈动脉内膜、中膜部分的感兴趣区域(ROI).采用基于堆栈式自编码器(SAE)构造的模式分类器将ROI中的像素进行分类.最后利用分类区域的面积信息和位置信息对分类结果进行甄别,运用曲线拟合提取边界完成测量任务.结果 针对本研究所用图像库中的84幅颈动脉超声图像进行实验,金标准(GT)由两名专家4次测量的平均值产生,其与AS之间的绝对误差和标准差为(13.3±20.5) μm,协方差系数为0.990 7.结论 实验结果表明,此算法总体性能较好,能够实现超声颈动脉血管内中膜全自动、快速、准确分割,从而满足临床需要.