Region-growing Method Research Articles

Colour image segmentation using region-growing and region-merging methods

Image segmentation is the process of dividing the given image into regions homogeneous with respect to certain features, and which hopefully correspond to real objects in the actual scene. Region-growing algorithms have been used mostly in the analysis of greyscale images; however, some significant work has been completed in the colour. The original image is divided into image blocks, which are not overlapped; then, the mean and variance of each image block were calculated, and the image blocks were divided into homogeneous colour blocks and texture blocks by the variance of it. The initial seed regions are automatically selected depending on calculating the homogeneous colour. The quadtree decomposition algorithm was applied to split and merge the texture blocks. This process is followed until no more pixels can be added. The experimental segmentation results hold favourable consistency in terms of human perception, and confirm effectiveness of the algorithm.

Optimizing the Use of Radiologist Seed Points for Improved Multiple Sclerosis Lesion Segmentation

Many current methods for multiple sclerosis (MS) lesion segmentation require radiologist seed points as input, but do not necessarily allow the expert to work in an intuitive or efficient way. Ironically, most methods also assume that the points are placed optimally. This paper examines how seed points can be processed with intuitive heuristics, which provide improved segmentation accuracy while facilitating quick and natural point placement. Using a large set of MRIs from an MS clinical trial, two radiologists are asked to seed the lesions while unaware that the points would be fed into a classifier, based on Parzen windows, that automatically delineates each marked lesion. To evaluate the impact of the new heuristics, an interactive region-growing method is used to provide ground truth and the Dice coefficient (DC) and Spearman’s rank correlation are used as the primary measures of agreement. A stratified analysis is performed to determine the effect on scans with low-, medium-, and high lesion loads. Compared to the unenhanced classifier, the heuristics dramatically improve the DC (+32.91 pt.) and correlation (+0.50) for the scans with low lesion loads, and also improve the DC (+14.55 pt.) and correlation (+0.15) for the scans with medium lesion loads, while having aminimal effect for the scans with high lesion loads, which are already segmented accurately by Parzen windows.With the heuristics, the DC is close to 80% and the correlation is above 0.9 for all three load categories.

Segmentation of artery areas on non-enhanced fresh blood imaging based on a dot enhancement filter and the 3-D region-growing method

Peripheral arterial disease (PAD) is characterized by arterial obstruction in the lower extremities due to atherosclerosis, and manifests itself as intermittent claudicating, limb ischemia, or gangrene. Diagnosis of PAD using magnetic resonance imaging (MRI) equipment without contrast medium is available as a useful visual screening in clinical practice. In this article, we propose a new method for the segmentation of arterial images which are obtained from non-contrast-enhanced magnetic resonance angiography (MRA) based on a dot enhancement filter and 3-D region-growing methods, and satisfactory experimental results are obtained.

歯顎顔面部ＣＴ画像からの歯の自動抽出

CT is an effective tool for image diagnosis because it enables noninvasive observation of internal organs. In the course of CT, 3D-CT, such as a helical scanning CT and a multi-detector row CT, has been developed. With the use of 3D-CT, organs can be observed from several viewing directions. Even now, however, 3D-CT images are generated by manual procedures to extract objective organs using the threshold method. These procedures waste time and effort. Therefore, development of highly automated and effective extracting techniques has been desired. The region growing (RG) method is one of the effective techniques of extracting internal organs. The conventional RG method, however, has some defects. Extracted regions are strongly affected by the threshold value for segmentation. A break point on a region contour yields a leakage of region. To overcome these defects, we formulated a modified region growing method with edge detection (MRGWED) which combines a three-dimensional region growing technique and an edge detection technique. Using the MRGWED, we tried to extract teeth from dentomaxillofacial 3D-CT image data.

画像特徴量を用いた大規模点群データからの円柱面と矩形面の検出(機械要素,潤滑,設計,生産加工,生産システムなど)

The state-of-the-art mid/long-range laser scanners can capture dense point-clouds from engineering facilities. However, it is very tedious work to manually segment large-scale point-clouds and fit surfaces to the subsets of point-clouds for constructing 3D models. In this paper, we propose a new method to automatically extract cylindrical faces and rectangle faces, which are commonly used for standard parts in engineering facilities. Since it requires a lot of time and memory to search surfaces in large-scale point-clouds, we generate distance images and intensity images from dense point-clouds and estimate canditates of cylindrical and rectangle regions using image processing techniques. Then, we calculate equations of surfaces by applying the region-growing method to the candidate regions, and detect rectangle and cylindrical faces

Individual flexor tendon identification within the carpal tunnel: A semi-automated analysis method for serial cross-section magnetic resonance images

Carpal tunnel syndrome is commonly viewed as resulting from chronic mechanical insult of the median nerve by adjacent anatomical structures. Both the median nerve and its surrounding soft tissue structures are well visualized on magnetic resonance (MR) images of the wrist and hand. Addressing nerve damage from impingement of flexor digitorum tendons co-occupying the tunnel is attractive, but to date has been restricted by lack of means for making individual identifications of the respective tendons. In this image analysis work, we have developed a region-growing method to positively identify each individual digital flexor tendon within the carpal tunnel by tracking it from a more distal MR section where the respective tendon identities are unambiguous. Illustratively, the new method was applied to MRI scans from four different subjects in a variety of hand poses. Conventional shape measures yielded less discriminatory information than did evaluations of individual tendon location and arrangement. This new method of rapid identification of individual tendons will facilitate analysis of tendon/nerve interactions within the tunnel, thereby providing better information about mechanical insult of the median nerve.

Comparison Study of Different Bowel Preparation Regimens and Different Fecal-Tagging Agents on Tagging Efficacy, Patients' Compliance, and Diagnostic Performance of Computed Tomographic Colonography

To compare and find the optimal combination of bowel preparation regimens and different concentrations of barium-based fecal-tagging agents on computed tomographic colonography in terms of tagging efficacy, patients' compliance, and diagnostic performance. Sixty-nine consecutive patients were prospectively enrolled and randomly divided into 4 groups according to a combination of bowel preparation regimens (sodium phosphate [Colclean] vs magnesium citrate [LoSoPrep]) and different concentrations of barium-based fecal-tagging agents (Tagitol V [thick, 40% wt/vol] 60 mL vs EasyCT [thin, 4.6% wt/vol] 600 mL). Patients who received Colclean and EasyCT were designated as group 1; Colclean and Tagitol V as group 2; LoSoPrep and EasyCT, group 3; and LoSoPrep and Tagitol V, group 4. For objective analysis, the volume of tagged feces and fluid was semiautomatically calculated using a threshold of 150 Hounsfield units. The volume of untagged feces and fluid was also calculated using the seeded region-growing method. The tagging efficacy was then calculated and compared using the analysis of variance test. For subjective analysis, 2 radiologists visually determined the overall tagging efficacy on a segmental basis. The numbers of homogeneously tagged, heterogeneously tagged, heterogeneously untagged, and homogeneously untagged feces greater than or equal to 6 mm and fluid were counted in each segment. The rates of homogeneously tagged stool and fluid were compared between the groups using the chi test. Patients' compliance for taking the regimens was recorded and compared using the Kruskal-Wallis test. Per-polyp and per-patient sensitivity for polyps greater than or equal to 6 mm were analyzed and compared using the chi or Fisher exact test. Objective analysis showed that tagging efficacy was significantly lower in group 1 than in the other 3 groups. In a subjective analysis, overall tagging efficacy was significantly higher in group 4 than in the other groups. The rate of homogeneously tagged fluid was significantly higher in group 4 than in the other groups. The rate of homogeneously tagged feces was significantly higher in the thick barium group than in the thin barium group. Patients' compliance was significantly worse in group 1 than in the other 3 groups. Per-polyp and per-patient sensitivities for polyps greater than or equal to 6 mm were not significantly different among all groups. However, per-patient specificity for lesions greater than or equal to 10 mm was significantly lower in group 2 than in the other groups. With better tagging efficacy and compliance while maintaining comparable diagnostic performance, it is our belief that the best combination of bowel preparation and fecal-tagging regimen is a combination of magnesium citrate and a high concentration of a barium tagging agent.

Algebraic Decomposition of Fat and Water in MRI

The decomposition of magnetic resonance imaging (MRI) data to generate water and fat images has several applications in medical imaging, including fat suppression and quantification of visceral fat. We introduce a novel algorithm to overcome some of the problems associated with current analytical and iterative decomposition schemes. In contrast to traditional analytical schemes, our approach is general enough to accommodate any uniform echo-shift pattern, any number of metabolites and signal samples. In contrast to region-growing method that use a smooth field-map initialization to resolve the ambiguities with the IDEAL algorithm, we propose to use an explicit smoothness constraint on the final field-map estimate. Towards this end, we estimate the number of feasible solutions at all the voxels, prior to the evaluation of the roots. This approach enables the algorithm to evaluate all the feasible roots, thus avoiding the convergence to the wrong solution. The estimation procedure is based on a modification of the harmonic retrieval (HR) framework to account for the chemical shift dependence in the frequencies. In contrast to the standard linear HR framework, we obtain the frequency shift as the common root of a set of quadratic equations. On most of the pixels with multiple feasible solutions, the correct solution can be identified by a simple sorting of the solutions. We use a region-merging algorithm to resolve the remaining ambiguity and phase-wrapping. Experimental results indicate that the proposed algebraic scheme eliminates most of the difficulties with the current schemes, without compromising the noise performance. Moreover, the proposed algorithm is also computationally more efficient.

Toward breast cancer diagnosis based on automated segmentation of masses in mammograms

This work explores the use of characterization features extracted based on breast-mass contours obtained by automated segmentation methods, for the classification of masses in mammograms according to their diagnosis (benign or malignant). Two sets of mass contours were obtained via two segmentation methods (a dynamic-programming-based method and a constrained region-growing method), and simplified versions of these contours (modeling the contours as ellipses) were employed to extract a set of six features designed for characterization of mass margins (contrast between foreground region and background region, coefficient of variation of edge strength, two measures of the fuzziness of mass margins, a measure of spiculation based on relative gradient orientation, and a measure of spiculation based on edge-signature information). Three popular classifiers (Bayesian classifier, Fisher's linear discriminant, and a support vector machine) were then used to predict the diagnosis of a set of 349 masses based on each of said features and some combinations of these. The systems (each system consists of a segmentation method, a featureset, and a classifier) were compared with each other in terms of their performance on the diagnosis of the set of breast masses. It was found that, although there was a percent difference of about 14% in the average segmentation quality between methods, this was translated into an average percent difference of only 4% in the classification performance. It was also observed that the spiculation feature based on edge-signature information was distinctly better than the rest of the features, although it is not very robust to changes in the quality of the segmentation. All systems were more efficient in predicting the diagnosis of benign masses than that of the malignant masses, resulting in low sensitivity and high specificity values (e.g. 0.6 and 0.8, respectively) since the positive class in the classification experiments is the set of malignant masses. It was concluded that features extracted from automated contours can contribute to the diagnosis of breast masses in screening programs by correctly identifying a majority of benign masses.

Automated Segmentation and Morphometric Analysis of the Human Airway Tree from Multidetector CT Images

Remarkable advances in computed tomography (CT) technology geared our research toward investigating the integrative function of the lung and the development of a database of the airway tree incorporating anatomical and functional data with computational models. As part of this project, we are developing the algorithm to construct an anatomically realistic geometric model of airways from CT images. The basic concept of the algorithm is to segment as many airway trees as possible from CT images and later correct quantified parameters based on CT values. CT images are acquired with a 64-channel multidetector CT, and the airway is then extracted from them by the region-growing method while maintaining connectivity. Using this method, we extracted 428 airways up to the 14th branching generation. Although the airway diameters up to the 4th generation showed good agreement with those reported in an autopsy study, those in later generations were all greater than the reported values because of the limited resolution of the CT images. We corrected the errors in diameters by assessing the relationship between the diameter and median value of Hounsfield unit (HU) intensity of each airway; consequently, the diameters up to generation 8 agreed well with the reported values. Based on these results, we conclude that the use of HU-based correction algorithm combined with rough segmentation can be another way to improve data accuracy in the morphometric analysis of airways from CTs.

Morphological Study of Fibroblasts Treated with Cytochalasin D and Colchicine Using a Confocal Laser Scanning Microscopy

The role of actin filaments and microtubules in 3D cell morphology was investigated using confocal laser scanning microscopy and image analysis based on a region-growing method. Fibroblasts were treated with cytochalasin D or colchicine to disrupt the actin filaments or microtubules, respectively, and the structure and distribution of these cytoskeletal filaments were observed using a confocal laser scanning microscope. From the 3D reconstructed fluorescence images of the cytoskeleton, morphological parameters such as volume, adhesion area, height, and volume ratio of individual cells were determined. The volume ratio was defined as the ratio of the partial volume for every 10% of the height to the total cell volume. The cell volume decreased slightly after the disruption of actin filaments and microtubules, but the change was not significant. The cell adhesion area was significantly decreased after the disruption of actin filaments and microtubules, and was significantly smaller in actin filament-disrupted cells than in microtubule-disrupted cells. Cell height increased significantly after actin filament disruption, whereas it remained almost unchanged after microtubule disruption. Analysis of the volume ratio revealed that the cell shape changed from a cone to a hemisphere after disruption of actin filaments and slightly shifted toward a hemisphere-like shape after microtubule disruption. These results suggest that actin filaments are the major component responsible for the maintenance of global cell shape and that the contribution of microtubules to global cell morphology is much less than that of actin filaments.

A dual‐stage method for lesion segmentation on digital mammograms

Mass lesion segmentation on mammograms is a challenging task since mass lesions are usually embedded and hidden in varying densities of parenchymal tissue structures. In this article, we present a method for automatic delineation of lesion boundaries on digital mammograms. This method utilizes a geometric active contour model that minimizes an energy function based on the homogeneities inside and outside of the evolving contour. Prior to the application of the active contour model, a radial gradient index (RGI)-based segmentation method is applied to yield an initial contour closer to the lesion boundary location in a computationally efficient manner. Based on the initial segmentation, an automatic background estimation method is applied to identify the effective circumstance of the lesion, and a dynamic stopping criterion is implemented to terminate the contour evolution when it reaches the lesion boundary. By using a full-field digital mammography database with 739 images, we quantitatively compare the proposed algorithm with a conventional region-growing method and an RGI-based algorithm by use of the area overlap ratio between computer segmentation and manual segmentation by an expert radiologist. At an overlap threshold of 0.4, 85% of the images are correctly segmented with the proposed method, while only 69% and 73% of the images are correctly delineated by our previous developed region-growing and RGI methods, respectively. This resulting improvement in segmentation is statistically significant.

Automatic Segmentation of Pelvic Structures From Magnetic Resonance Images for Prostate Cancer Radiotherapy

Target-volume and organ-at-risk delineation is a time-consuming task in radiotherapy planning. The development of automated segmentation tools remains problematic, because of pelvic organ shape variability. We evaluate a three-dimensional (3D), deformable-model approach and a seeded region-growing algorithm for automatic delineation of the prostate and organs-at-risk on magnetic resonance images. Manual and automatic delineation were compared in 24 patients using a sagittal T2-weighted (T2-w) turbo spin echo (TSE) sequence and an axial T1-weighted (T1-w) 3D fast-field echo (FFE) or TSE sequence. For automatic prostate delineation, an organ model-based method was used. Prostates without seminal vesicles were delineated as the clinical target volume (CTV). For automatic bladder and rectum delineation, a seeded region-growing method was used. Manual contouring was considered the reference method. The following parameters were measured: volume ratio (Vr) (automatic/manual), volume overlap (Vo) (ratio of the volume of intersection to the volume of union; optimal value = 1), and correctly delineated volume (Vc) (percent ratio of the volume of intersection to the manually defined volume; optimal value = 100). For the CTV, the Vr, Vo, and Vc were 1.13 (+/-0.1 SD), 0.78 (+/-0.05 SD), and 94.75 (+/-3.3 SD), respectively. For the rectum, the Vr, Vo, and Vc were 0.97 (+/-0.1 SD), 0.78 (+/-0.06 SD), and 86.52 (+/-5 SD), respectively. For the bladder, the Vr, Vo, and Vc were 0.95 (+/-0.03 SD), 0.88 (+/-0.03 SD), and 91.29 (+/-3.1 SD), respectively. Our results show that the organ-model method is robust, and results in reproducible prostate segmentation with minor interactive corrections. For automatic bladder and rectum delineation, magnetic resonance imaging soft-tissue contrast enables the use of region-growing methods.

A new region-primitive method for classification of colour meat image texture based on size, orientation, and contrast

A new structural method was developed to characterise texture of colour meat images. Structural primitives or region-primitives were constructed using a region-growing method with a global and automatically calculated threshold. A total of eight textural features under the consideration of size, orientation, and contrast, which are the primary factors of human perception of texture pattern, were obtained from the region-primitives. An experiment was set up to classify the texture pattern of a set of 60 cooked meat colour images consisting of 20 beef images, 20 lamb images, and 20 pork images using texture features obtained from the proposed region-primitive method and run-length matrix method, respectively. The proposed region-primitive method (error rate: 18.3%) was found to perform better than the run-length matrix method (error rate: 35.0%). Results also indicated that the texture features obtained by the proposed method carried adequate texture information for classification and that the proposed region-primitive method could be potentially used for the characterisation of irregular textural pattern in cooked meat images, which was inapproachable by conventional structural texture analysis techniques.

An improved algorithm for the detection of plumes caused by natural or technological hazards using AVHRR imagery

Fires caused by natural or technological disasters emit large amounts of smoke which, once formed into plumes, may affect the human health and the environment. Satellite remote sensing data provide an effective tool to achieve detection and monitoring of these plumes over large areas on a routine basis. Discrimination of plumes on satellite images is a prerequisite to study and retrieve physical, chemical and optical properties of emitted smoke. An improved algorithm for the detection of plumes caused by natural or technological hazards using AVHRR imagery is presented in this study. The method is based on a multi-temporal and multi-spectral change detection algorithm. It is performed in two main steps: a) appropriate spectral and spatial filters are applied on the images acquired before and after a fire event in visible and near-infrared ranges in order to extract the core of the plume; b) a criterion on spectral information is defined as an homogeneity measure that enables, through a modified version of the region-growing method, the spatial expansion of the detected core to include the complete area covered by the plume. Through this approach, a pixel is identified as a plume pixel if it is “close” to the core plume pixels in both spatial and spectral spaces. The algorithm was developed and calibrated using AVHRR images acquired over Spain before and during a major forest fire event on July 16, 2005. It was applied using past events of natural and technological hazards in several locations to ensure its global applicability and robustness. The algorithm produced accurate results in all cases of plumes, either in natural or in technological fire events. Three application cases are presented in this study: A major fire in an industrial installation in London (December 11, 2005), a major fire in Baghdad during the recent war in Iraq (April 1, 2003) and a forest fire in California (September 29, 2005).

Three-Dimensional Computed Tomography for Planning Donor Hepatectomy

Background An accurate estimation of preoperative volumetric measurements of the donor liver is essential in living donor liver transplantation. Methods Three-dimensional (3-D) computed tomography (CT) was applied to 56 living donors. 3-D images of the liver were constructed using the region-growing method and the volume of each sector was measured. Results The median volume ratios of the left liver, caudate lobe, right paramedian, and lateral sectors were 34%, 4%, 38%, and 25% of the total liver volume, respectively. The shape of the congestive area in the right paramedian sector was properly demonstrated by 3-D CT. The volume of the region corresponded to 32% of the right liver. The actual volume of the graft correlated well with the estimated graft volume ( n = .86). Conclusions The region-growing method was useful for graft selection and for determining the indication of middle hepatic vein reconstruction in right liver grafts.

Method for the automatic detection and segmentation of the spinal canal in computed tomographic images

A method for the automatic 3-D segmentation of the spi- nal canal in computed tomographic (CT) images is presented. The method uses a priori radiological and anatomical information, math- ematical morphology, and region-growing methods. The skin and peripheral fat structures are determined by delineating the air and other materials external to the body. Using the fat layer as a refer- ence, the bone structure is segmented. The Hough transform for the detection of circles is applied to a cropped bone edge map that includes the thoracic vertebral structure. The centers of the detected circles are used to derive the information required for the fuzzy con- nectivity algorithm that is employed to segment the spinal canal. In a preliminary study, the method successfully segmented the spinal canal in eight CT volumes of four patients, with Hausdorff distances with reference to contours drawn independently by a radiologist in the range 1.7±0.8 mm. © 2006 SPIE and IS&T.

Gray matter volume reduction in obsessive–compulsive disorder with schizotypal personality trait

Objectives Obsessive–compulsive disorder (OCD) with schizotypal personality trait (SPT) and OCD without SPT demonstrated differences in cognitive dysfunction and treatment response. This study aimed to investigate whether brain volume differs between OCD with SPT and OCD without SPT. Methods OCD with SPT ( n = 20), OCD without SPT ( n = 47), healthy comparison subjects ( n = 83) and schizophrenic patients ( n = 59) participated in this study. We assessed brain volume such as gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) using magnetic resonance imaging. The imaging data sets were filtered using anisotropic diffusion methods to improve the signal to noise ratio. The semi-automated region-growing method was used to remove images of tissues exterior to the brain. The fuzzy C-means algorithm was used to segment the extracted brain images into gray matter, white matter and cerebrospinal fluid. Results OCD with SPT ( p = 0.048) and Schizophrenic patients ( p < 0.001) demonstrated a significant GM volume reduction compared to healthy controls. OCD without SPT revealed no significant GM volume reduction compared to healthy controls ( p = 0.504). Conclusion This study suggests that OCD with SPT could be distinguished as a distinct subtype of OCD, based on observations of gray matter volume.

Generate triangulated surfaces from massive unorganized points

A region-growing method for reconstructing triangulated surfaces from massive unorganized points is presented. To save memory space, a ring data-structure is adopted to build connections between points and triangulated surfaces. The data-structure allows the efficient retrieval of all neighboring vertices and triangles of a given vertice. To narrow the search range of adjacent points and avoid mangle intersection, an influence area is defined for each active-edge. In the region-growing process of triangulated surfaces, a minimum-edge-angle-product algorithm is put forward to select an appropriate point to form a new triangle for an active edge. Results indicate that the presented method has high efficiency and needs less memory space, optimized triangulated surfaces with reliable topological quality can be obtained after triangulation.

Detection of Flaws in Weldment Images Using Region Growing Segmentation Technique

This paper presents an efficient detection method to highlight the flaws present in weld images. The obtained results further help in finding the size and location o f the flaw. The method is applied on X-ray images having flaws generated during the welding process. Using region-growing method we merge the adjacent detected pixels on homogeneity criteria, to obtain the flaws. Region growing starts with seed(s). The seed value is determined with the help of histogram analysis. The peaks and valleys o f histogram help in determining the seed value. The developed algorithm is explained and applied on the welding images and some preliminary results are shown which are encouraging.