SEARCH · PubMed Health

Results for “Automated image segmentation”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 433 records · Page 24Linked to original sources

PubMed2004

Optic nerve head segmentation.

Reliable and efficient optic disk localization and segmentation are important tasks in automated retinal screening. General-purpose edge detection algorithms often fail to segment the optic disk due to fuzzy boundaries, inconsistent image contrast or missing edge features. This paper presents an algorithm for the localization and segmentation of the optic nerve head boundary in low-resolution images (about 20 microns/pixel). Optic disk localization is achieved using specialized template matching, and segmentation by a deformable contour model. The latter uses a global elliptical model and a local deformable model with variable edge-strength dependent stiffness. The algorithm is evaluated against a randomly selected database of 100 images from a diabetic screening programme. Ten images were classified as unusable; the others were of variable quality. The localization algorithm succeeded on all bar one usable image; the contour estimation algorithm was qualitatively assessed by an ophthalmologist as having Excellent-Fair performance in 83% of cases, and performs well even on blurred images.

Algorithms↗

PubMed2002

Automatic segmentation of echocardiographic sequences by active appearance motion models.

A novel extension of active appearance models (AAMs) for automated border detection in echocardiographic image sequences is reported. The active appearance motion model (AAMM) technique allows fully automated robust and time-continuous delineation of left ventricular (LV) endocardial contours over the full heart cycle with good results. Nonlinear intensity normalization was developed and employed to accommodate ultrasound-specific intensity distributions. The method was trained and tested on 16-frame phase-normalized transthoracic four-chamber sequences of 129 unselected infarct patients, split randomly into a training set (n = 65) and a test set (n = 64). Borders were compared to expert drawn endocardial contours. On the test set, fully automated AAMM performed well in 97% of the cases (average distance between manual and automatic landmark points was 3.3 mm, comparable to human interobserver variabilities). The ultrasound-specific intensity normalization proved to be of great value for good results in echocardiograms. The AAMM was significantly more accurate than an equivalent set of two-dimensional AAMs.

Algorithms↗

PubMed2002

Automated detection of diabetic retinopathy on digital fundus images.

AIMS: The aim was to develop an automated screening system to analyse digital colour retinal images for important features of non-proliferative diabetic retinopathy (NPDR). METHODS: High performance pre-processing of the colour images was performed. Previously described automated image analysis systems were used to detect major landmarks of the retinal image (optic disc, blood vessels and fovea). Recursive region growing segmentation algorithms combined with the use of a new technique, termed a 'Moat Operator', were used to automatically detect features of NPDR. These features included haemorrhages and microaneurysms (HMA), which were treated as one group, and hard exudates as another group. Sensitivity and specificity data were calculated by comparison with an experienced fundoscopist. RESULTS: The algorithm for exudate recognition was applied to 30 retinal images of which 21 contained exudates and nine were without pathology. The sensitivity and specificity for exudate detection were 88.5% and 99.7%, respectively, when compared with the ophthalmologist. HMA were present in 14 retinal images. The algorithm achieved a sensitivity of 77.5% and specificity of 88.7% for detection of HMA. CONCLUSIONS: Fully automated computer algorithms were able to detect hard exudates and HMA. This paper presents encouraging results in automatic identification of important features of NPDR.

Algorithms↗

PubMed2004

Automatic segmentation of digital micrographs: a survey.

Digital micrographs and play a crucial role in today's bio-medical research. Due to progresses in experiment standardization and automation large sets of digital microscopy images, so called micrographs are recorded and stored to databases. The subsequent analysis of the large number of digital images needs the image information to be transformed into quantitative data, which can be processed by statistical methods and datamining. This article summarizes applications of optical microscopy in biomedical research and describes the individual characteristics in micrograph segmentation and classification. An overview on past works based on image processing and artificial neural networks is given and the problem of segmentation evaluation. It concludes with recommendations for future works.

Image Processing, Computer-Assisted↗

PubMed2005

Quantitative diagnosis of stress-induced myocardial ischemia using analysis of contrast echocardiographic parametric perfusion images.

AIMS: Parametric imaging of myocardial perfusion provides useful visual information for the diagnosis of coronary artery disease (CAD). We developed a technique for automated detection of perfusion defects based on quantitative analysis of parametric perfusion images and validated it against coronary angiography. METHODS AND RESULTS: Contrast-enhanced, apical 2-, 3- and 4-chamber images were obtained at rest and with dipyridamole in 34 patients with suspected CAD. Images were analyzed to generate parametric perfusion images of the standard contrast-replenishment model parameters A, beta and A.beta. Each parametric image was divided into six segments, and mean parameter value (MPV) was calculated for each segment. Segmental MPV ratio between stress and rest was defined as a flow reserve index (FRI). Receiver operating characteristics (ROC) analysis was used in a Study group (N=17) to optimize FRI threshold and the minimal number of abnormal segments per vascular territory (LAD and non-LAD), required for automated detection of stress-induced perfusion defects. The optimized detection algorithm was then tested prospectively in the remaining 17 patients (Test group). LAD and non-LAD stenosis >70% was found in 19 and 17 patients, respectively. In the Study group, FRI threshold was: LAD=0.95 and non-LAD=0.68, minimal number of abnormal segments was four and two, correspondingly. Sensitivity, specificity and accuracy in the Test group were: 75%, 67% and 71% in the LAD, and 75%, 75% and 75% in the non-LAD territories. CONCLUSION: Automated quantitative analysis of contrast echocardiographic parametric perfusion images is feasible and may aid in the objective detection of CAD.

Aged↗

PubMed2001

Automated hippocampal segmentation by regional fluid registration of serial MRI: validation and application in Alzheimer's disease.

The application of voxel-level three-dimensional registration to serial magnetic resonance imaging (MRI) is described. This fluid registration determines deformation fields modeling brain change, which are consistent with a model describing a viscous fluid. The objective was to validate the measurement of hippocampal volumetric change by fluid registration in Alzheimer's disease (AD) against current methodologies. The hippocampus was chosen for this study because it is difficult to measure reproducibly by manual segmentation and is widely studied; however, the technique is applicable to any structure which can be delineated on a scan. First, suitable values for the viscosity-body-force-ratio, alpha (0.01), and the number of iterations (300), were established and the convergence, repeatability, linearity, and accuracy investigated and compared with expert manual segmentation. A simple model of hippocampal atrophy was used to compare simulated volumetric change against that obtained by fluid registration. Finally the serial segmentation was compared with the current gold standard technique-expert human labeling with a volume repeatability of approximately 4%-in 27 subjects (15 normal controls, 12 clinically diagnosed with Alzheimer's disease). The scan-rescan volumetric consistency of serial segmentation by fluid-registration was shown to be superior to human serial segmentors ( approximately 2%). The mean absolute volume difference between fluid and manual segmentation was 0.7%. Fluid registration has potential importance for tracking longitudinal structural changes in brain particularly in the context of the clinical trial where large numbers of subjects may have multiple MR scans.

Aged↗

PubMed1986

Morphometric assessment of drug effects in experimental spinal cord injury.

The effect of large doses of methylprednisolone sodium succinate (MPSS) and two protease inhibitors, leupeptin and bestatin, on experimental acute spinal cord injury was evaluated by morphometric analysis of degenerating axons with the aid of an automated image analyzer. Spinal cord injury was produced by epidural compression with a surgical clip on the T-11 segment in rats. The extent of axonal damage was assessed in Rexed's lamina VIII in the L-6 segment by measuring the amount of silver grains, representing degenerating axons and their terminals, using the Fink-Heimer method. The severity of axonal damage was expressed as the degeneration index: that is, the amount of silver grains in experimental animals/the amount of silver grains in cord-transected animals. When examined on the 7th postoperative day, axonal degeneration in MPSS-treated rats was significantly decreased, with an average degeneration index difference of 6 (p less than 0.05). Increased preservation of axons was seen in the leupeptin-treated rats sacrificed 7, 10, and 14 days after trauma. The difference in the degeneration index between the leupeptin-treated and untreated groups was 16 on Day 7 (p less than 0.001), 12 on Day 10 (p less than 0.001), and 13 on Day 14 (p less than 0.01). Bestatin had no beneficial effect. The implications for the use of calcium-activated neutral protease inhibitors in acute spinal cord injury are discussed.

Animals↗

PubMed2005

Identifying lesions on structural brain images--validation of the method and application to neuropsychological patients.

The study of neuropsychological disorders has been greatly facilitated by the localization of brain lesions on MRI scans. Current popular approaches for the assessment of MRI brain scans mostly depend on the successful segmentation of the brain into grey and white matter. These methods cannot be used effectively with large lesions because lesions usually impair segmentation. We propose a novel, fully automated approach for the delineation of brain lesions on MR scans. This method involves comparing a skull stripped, smoothed, unsegmented T1 images to a control group using the general linear model. We tested this method by using images with simulated lesions of different sizes and images containing real lesions from patients with language deficits. We also tested how varying the size of the Gaussian smoothing kernel affects detection. The simulation was informed by findings of a lesion morphological study also presented here. The proposed method detected simulated lesions effectively in the range of 30--90%<normal signal. Smoothing kernels in the range of 8--12 mm resulted in the most accurate lesion detection. Both artificial and real lesions were optimally detected when the results were uncorrected for multiple comparisons at p<.001. This proposed method produced highly satisfactory results and can be used to generate reproducible detection of lesions.

Adult↗

PubMed2002

Image processing techniques for noise removal, enhancement and segmentation of cartilage OCT images.

Osteoarthritis, whose hallmark is the progressive loss of joint cartilage, is a major cause of morbidity worldwide. Recently, optical coherence tomography (OCT) has demonstrated considerable promise for the assessment of articular cartilage. Among the most important parameters to be assessed is cartilage width. However, detection of the bone cartilage interface is critical for the assessment of cartilage width. At present, the quantitative evaluations of cartilage thickness are being done using manual tracing of cartilage-bone borders. Since data is being obtained near video rate with OCT, automated identification of the bone-cartilage interface is critical. In order to automate the process of boundary detection on OCT images, there is a need for developing new image processing techniques. In this paper we describe the image processing techniques for speckle removal, image enhancement and segmentation of cartilage OCT images. In particular, this paper focuses on rabbit cartilage since this is an important animal model for testing both chondroprotective agents and cartilage repair techniques. In this study, a variety of techniques were examined. Ultimately, by combining an adaptive filtering technique with edge detection (vertical gradient, Sobel edge detection), cartilage edges can be detected. The procedure requires several steps and can be automated. Once the cartilage edges are outlined, the cartilage thickness can be measured.

Animals↗

PubMed2007

Diffusion tensor imaging segmentation of white matter structures using a Reproducible Objective Quantification Scheme (ROQS).

Reproducible Objective Quantification Scheme (ROQS) is a novel method for regional white matter measurements of diffusion tensor imaging (DTI) parameters that overcomes the limitations of previous approaches for analyzing large cohorts of subjects reliably. ROQS is a semi-automated technique that exploits the fiber orientation information from the diffusion tensor in conjunction with a binary masking and chain-linking algorithm to segment anatomically distinct white matter tracts for subsequent quantitative analysis of DTI parameters such as fractional anisotropy and apparent diffusion coefficient. When applied to 3-T whole-brain DTI of normal adult volunteers, ROQS is shown to segment the corpus callosum much faster than manual region of interest (ROI) delineation, and with better reproducibility and accuracy.

Adult↗

PubMed2003

Validation of gated blood-pool SPECT cardiac measurements tested using a biventricular dynamic physical phantom.

UNLABELLED: We have developed a biventricular dynamic physical cardiac phantom to test gated blood-pool (GBP) SPECT image-processing algorithms. Such phantoms provide absolute values against which to assess accuracy of both right and left computed ventricular volume and ejection fraction (EF) measurements. METHODS: Two silicon-rubber chambers driven by 2 piston pumps simulated crescent-shaped right ventricles wrapped partway around ellopsoid left ventricles. Twenty experiments were performed at Ghent University, for which right and left ventricular true volume and EF ranges were 65-275 mL and 55-165 mL and 7%-49% and 12%-69%, respectively. Resulting 64 x 64 simulated GBP SPECT images acquired at 16 frames per R-R interval were sent to Columbia University, where 2 observers analyzed images independently of each other, without knowledge of true values. Algorithms automatically segmented right ventricular activity volumetrically from left ventricular activity. Automated valve planes, midventricular planes, and segmentation regions were presented to observers, who accepted these choices or modified them as necessary. One observer repeated measurements >1 mo later without reference to previous determinations. RESULTS: Linear correlation coefficients (r) of the mean of the 3 GBP SPECT observations versus true values for right and left ventricles were 0.80 and 0.94 for EF and 0.94 and 0.95 for volumes, respectively. Correlations for right and left ventricles were 0.97 and 0.97 for EF and 0.96 and 0.89 for volumes, respectively, for interobserver agreement and 0.97 and 0.98 for EF and 0.96 and 0.90 for volumes, respectively, for intraobserver agreement. No trends were detected, though volumes and right ventricular EFs were significantly higher than true values. CONCLUSION: Overall, GBP SPECT measurements correlated strongly with true values. The phantom evaluated shows considerable promise for helping to guide algorithm developments for improved GBP SPECT accuracy.

Computer Simulation↗

PubMed2006

Automated interpretation of regional left ventricular wall motion from cardiac magnetic resonance images.

UNLABELLED: Magnetic resonance (MR) diagnosis of regional left ventricular (LV) dysfunction relies on visual interpretation of cine images that suffers from wide inter-observer variability, especially when performed by readers not specifically trained in the assessment of LV wall motion. Quantitative analysis tools, though widely available, are rarely used because they provide large amounts of detailed information, the interpretation of which requires additional time-consuming processing. We tested the feasibility of fast automated interpretation of regional LV function using computer analysis of this wall motion information. METHODS: Dynamic, ECG-gated, steady-state free precession short-axis images were obtained in 6-10 slices in 28 subjects (10 normal volunteers; 18 patients). Images were reviewed by an expert cardiologist who provided "gold standard" grades (normal, abnormal) for regional wall motion and, independently, by four radiologists. Same images were then analyzed using custom software. Regional fractional area changes computed in normal volunteers were used to obtain the optimal segment- and slice-specific threshold values for automated classification of regional wall motion for each patient. The levels of agreement with the "gold standard" grades were compared between the radiologists and the automated interpretation. RESULTS: While the visual interpretation required 2-5 minute per patient, the automated interpretation required < 1 sec, after endocardial border detection was complete. The automated interpretation resulted in higher sensitivity, specificity, and accuracy (84%, 77%, 79%, respectively) than the radiologists' grades (80%, 76%, 77%, respectively) and eliminated the high interobserver variability. CONCLUSION: Once the endocardial boundaries are defined, computer analysis of the regional wall motion information allows accurate, fully automated, immediate, objective and experience-independent interpretation of regional LV function.

Automation↗

PubMed2002

Computerized detection of pulmonary embolism in spiral CT angiography based on volumetric image analysis.

A fully automated method for computerized detection of pulmonary embolism in spiral computed tomography angiography was developed based on volumetric image analysis. The detection method is based on segmentation of pulmonary vessels to limit the search space, and analysis of several three-dimensional features inside segmented vessel volume. The features utilized are vascular size, local contrast based on mathematical morphology, degree of curvilinearity based on second derivatives, and geometric features such as volume and length. Detection results were obtained for 19 clinical data sets and the performance of the method was evaluated. Using the number and locations of thrombi diagnosed by radiologists as the gold standard, 100% sensitivity was achieved with 7.7 false positives per case, and 85% sensitivity was obtained with 2.6 false positives. For identification of all the positive cases as positive, i.e., detection of at least one thrombus per positive case, 1.9 false positives per case were obtained. These preliminary results suggest that the method has potential for fully automated detection of pulmonary embolism.

Algorithms↗

PubMed2007

Performance of the FearID earprint identification system.

The Forensic Ear Identification (FearID) research project was started in order to study the strength of evidence of earprints found on crime scenes. For this purpose, a sample of earprints from 1229 donors over three countries was collected. From each donor three left and three right earprints were gathered. On the one hand, operators denoted contours of the earprints to facilitate segmentation of the images, on the other anthropological specialists denoted anatomically specific locations. On the basis of this, methods for automated classification were developed and used for training of a system that classifies pairs of prints as 'matching' or 'non-matching'. Comparing lab quality prints, the system has an equal error rate of 4%. Starting from a reference database containing two prints per ear, hitlist behaviour is such that in 90% of all query searches the best hit is in the top 0.1% of the list. The results become less favourable (equal error rate of 9%) for print/mark comparisons.

Databases, Factual↗

PubMed1999

Calcified plaque cross-sectional area in human arteries: correlation between intravascular ultrasound and undecalcified histology.

BACKGROUND: The purpose of this investigation was to quantify the amount of intralesional calcium detected by intravascular ultrasound (IVUS) compared with undecalcified histology in human arteries. This method preserves intralesional calcium and reduces sectioning artifacts, thereby providing an accurate measure of calcium plaque morphology. METHODS AND RESULTS: Ten arterial segments (5 coronary, 5 iliac) were obtained at autopsy. IVUS imaging was performed with a 4.9F catheter at an automated pullback rate of 1.0 mm/s. The undecalcified arteries were dehydrated in ascending alcohol and polymerized in glycol methylmethacrylate. The arteries were cut into 200-microm sections with an Isomet low-speed saw and stained with Goldner's trichrome. The lumen cross-sectional area, the calcium plaque cross-sectional area, the calcium plaque depth, length, and angle of arc of calcified plaque were measured from the IVUS images and histologic sections. In 24 selected cross sections, there were 38 separate calcium plaques. An independent observer correctly identified 34 of 38 calcified plaques for a sensitivity of 89% and specificity of 97%. The total mean calcified plaque cross-sectional area measured from histology was 4.6 +/- 4.1 mm2 compared with 2.8 +/- 2.3 mm2 by IVUS (P =.002). Plaque depth measured by histology was 1.2 +/- 0.4 mm versus 0.7 +/- 0.2 mm by IVUS (P =.001). The length of calcium plaques measured by histology was 3.6 +/- 1.78 mm versus 3.6 +/- 1.5 mm for IVUS (r = 0.79). CONCLUSIONS: IVUS accurately depicts circumferential calcified lesions with high sensitivity (89%) and specificity (97%). However, IVUS underestimates the total calcified plaque cross-sectional area by 39%. This is mainly because of the inability of the ultrasound to penetrate intralesional calcium, which leads to an underestimation of the depth of calcium by 45%.

Adult↗

PubMed1995

Optic nerve axon count and axon diameter in patients with ocular hypertension and normal visual fields.

BACKGROUND: At postmortem examination, the authors obtained eight eyes of five individuals with elevated intraocular pressure and normal visual fields to study the axon count and mean axon diameter. METHODS: Automated image analysis was used to calculate the total axon count and mean axon diameter per nerve and per nerve segment for each eye. The authors applied the method of identification analysis to compare each study eye with a corresponding normal eye of patients of the same age. RESULTS: There was no statistically significant difference compared with control subjects for total axon count or segmental axon count for any of the eyes. Two eyes showed a statistically significant difference for mean axon diameter for the whole nerve but not for individual segments of the nerve. CONCLUSIONS: Some eyes subjected to varying duration and magnitude of intraocular pressure elevation with normal visual fields may maintain normal axon counts and mean axon diameters.

Aged↗

PubMed2005

Toward automatic phenotyping of developing embryos from videos.

We describe a trainable system for analyzing videos of developing C. elegans embryos. The system automatically detects, segments, and locates cells and nuclei in microscopic images. The system was designed as the central component of a fully automated phenotyping system. The system contains three modules 1) a convolutional network trained to classify each pixel into five categories: cell wall, cytoplasm, nucleus membrane, nucleus, outside medium; 2) an energy-based model, which cleans up the output of the convolutional network by learning local consistency constraints that must be satisfied by label images; 3) a set of elastic models of the embryo at various stages of development that are matched to the label images.

Algorithms↗

PubMed2005

Macular segmentation with optical coherence tomography.

PURPOSE: To develop a software algorithm to perform automated segmentation of retinal layer structures on linear macular optical coherence tomography (StratusOCT; Carl Zeiss Meditec, Inc., Dublin, CA) scan images and to test its performance in discriminating normal from glaucomatous eyes in comparison with conventional circumpapillary nerve fiber layer (cpNFL) thickness measurement. METHODS: Four layer structures within the retina were defined: the macular nerve fiber layer (mNFL), the inner retinal complex (IRC; retinal ganglion cell [RGC] layer + inner plexiform and nuclear layers), outer plexiform layer (OPL), and outer retinal complex (ORC; outer nuclear layer + photoreceptor layer). Normal and glaucomatous eyes underwent fast macular map and fast NFL OCT scans. Linear macular images were analyzed using the developed algorithm, and the results were compared with the cpNFL thickness measurement. RESULTS: Forty-seven subjects (23 normal and 24 with glaucoma) were analyzed. mNFL, cpNFL, IRC, and the total retinal thicknesses were significantly greater in normal than in glaucomatous eyes (P < or = 0.0002; Wilcoxon), whereas OPL thickness did not show a significant difference (P = 0.46). ORC thickness was significantly greater in glaucomatous than normal eyes (P = 0.035). Areas under the receiver operator characteristic curve (AROCs) for discriminating normal from glaucomatous eyes were highest with mNFL + IRC (0.97) and lowest with OPL (0.56). AROCs for OPL and ORC were significantly smaller than those for mNFL, IRC, mNFL+IRC, and cpNFL (P < or = 0.01). AROCs for IRC, mNFL + IRC, and cpNFL were significantly larger than for retinal thickness (P < or = 0.049). Among the best-performing parameters (mNFL, IRC, mNFL + IRC, and cpNFL) there was no significant difference in AROCs (P > or = 0.15). CONCLUSIONS: The newly developed macular segmentation algorithm described herein demonstrated its ability to quantify objectively the glaucomatous damage to RGCs and NFL and to discriminate between glaucomatous and normal eyes. Further algorithm refinement and improvements in resolution and image quality may yield a more powerful methodology for clinical glaucoma evaluation.

Algorithms↗