Automated Border Detection Research Articles

Automated Endocardial Border Detection and Left Ventricular Functional Assessment in Echocardiography Using Deep Learning

Endocardial border detection is a key step in assessing left ventricular systolic function in echocardiography. However, this process is still not sufficiently accurate, and manual retracing is often required, causing time-consuming and intra-/inter-observer variability in clinical practice. To address these clinical issues, more accurate and normalized automatic endocardial border detection would be valuable. Here, we develop a deep learning-based method for automated endocardial border detection and left ventricular functional assessment in two-dimensional echocardiographic videos. First, segmentation of the left ventricular cavity was performed in the six representative projections for a cardiac cycle. We employed four segmentation methods: U-Net, UNet++, UNet3+, and Deep Residual U-Net. UNet++ and UNet3+ showed a sufficiently high performance in the mean value of intersection over union and Dice coefficient. The accuracy of the four segmentation methods was then evaluated by calculating the mean value for the estimation error of the echocardiographic indexes. UNet++ was superior to the other segmentation methods, with the acceptable mean estimation error of the left ventricular ejection fraction of 10.8%, global longitudinal strain of 8.5%, and global circumferential strain of 5.8%, respectively. Our method using UNet++ demonstrated the best performance. This method may potentially support examiners and improve the workflow in echocardiography.

Semi-Automated Graphical System for Calculating Pulmonary Vascular Impedances in a Clinical Setting

Goal: Create a semi-automated, graphical, stand-alone application that uses clinically available asynchronous pressure and Doppler velocity captures to rapidly calculate, display, and interpret the pulmonary vascular impedance (PVZ) spectra. Methods: MATLAB-based software was written to analyze PVZ by creating a composite PVZ (cPVZ) spectra comprised of asynchronous screen captures of pulmonary arterial pressure and pulmonary arterial pulsed-wave Doppler velocity waveforms obtained during standard of care procedures. The pressure waveform, Doppler frequency envelopes, and ECG signals were re-digitized via automated border detection. cPVZ of averaged representative beats was calculated in the frequency domain via Fast Fourier Transform, and plotted vs harmonic z. Results: Successful generation of impedance spectra (PVZ(z)), where z is the harmonic, and additional parameters for characteristic impedance (Zc) and stiffness (Zs) were calculated as the mean of PVZ(2-4), and the sum of PVZ (1, 2), respectively. Conclusions: A graphically driven analysis of PVZ, calculated from standard of care right heart catheterization and echocardiography is possible. This system can help characterize both the steady and pulsatile components of right ventricular (RV) afterload in the clinical setting.

P734 Feasibility and accuracy of the new automated software dynamic heart model in an unselected population

Abstract Objective Preliminary studies showed the accuracy of machine learning based automated dynamic quantification of left ventricular (LV) and left atrial (LA) volumes. We aimed to evaluate the feasibility and accuracy of this new Dynamic Heart Model (DHM) software in an unselected population undergoing transthoracic echocardiography (TTE). Methods. We enrolled 91 consecutive unselected patients (80% in sinus rhythm) referred for clinically indicated 2D TTE, who also underwent single 3D TTE image acquisition from the apical 4-chamber view. 2D images were analyzed to measure ejection fraction, LV and LA volumes; 3D images were analyzed using Dynamic Heart Model (DHM) software (Philips Healthcare), which automatically measures chamber volumes throughout the cardiac cycle, resulting in LV and LA volume-time curves. Average time of analysis, feasibility, image quality were recorded and results compared between the 2D and 3D techniques. Results. Quality of the 91 2D TTE images was graded as poor (N = 13), satisfactory (N = 45) and good (N = 33). The use of DHM was feasible in 79/91 cases (87%). The remaining 12 datasets could not be analyzed because of poor images (N = 10) or incorrect automated border detection (N = 2): in these cases, the software did not accurately identify endocardial borders due to LV cavity near obliteration or extreme LA enlargement. When feasible, the boundary position was considered accurate in 61/79 patients (77%), while minor manual correction of the LV/LA borders was needed in the remaining cases. In only 1 case the reconstruction was considered unreliable because it needed major corrections. The overall time required to obtain DHM data was approximately 45 seconds. In all cases in which DHM was used, not only shapes of LV and LA were very well defined, but also functional curves were physiologically plausible. Even in the 13 patients in whom the 2D image was suboptimal, the DHM was not only feasible but also accurate endocardial boundaries in 8 cases, without (N = 5) or with only minimal manual corrections (N = 3). As expected, 3D LV volumes were slightly hige than 2D ones ( EDV 153.9 ± 59.8 vs 121.4 ± 47.3 mL, respectively), while LV EF and LA volumes were similar (EF 58.8 ± 11.8 vs 59 ± 11.8% and LA volume 92 ± 39.3 vs 83.4 ± 32.1 mL, respectively). Conclusions. The new DHM software is quick, feasible and accurate in the majority of unselected patients, including those with suboptimal 2D images or in atrial fibrillation. Introduction of this automated analysis into clinical practice can reduce examination time, while providing reliable information not only on volumes but also on function of the left heart chambers.

Automatic media-adventitia IVUS image segmentation based on sparse representation framework and dynamic directional active contour model

Segmentation of the arterial wall boundaries from intravascular ultrasound images is an important image processing task in order to quantify arterial wall characteristics such as shape, area, thickness and eccentricity. Since manual segmentation of these boundaries is a laborious and time consuming procedure, many researchers attempted to develop (semi-) automatic segmentation techniques as a powerful tool for educational and clinical purposes in the past but as yet there is no any clinically approved method in the market. This paper presents a deterministic-statistical strategy for automatic media-adventitia border detection by a fourfold algorithm. First, a smoothed initial contour is extracted based on the classification in the sparse representation framework which is combined with the dynamic directional convolution vector field. Next, an active contour model is utilized for the propagation of the initial contour toward the interested borders. Finally, the extracted contour is refined in the leakage, side branch openings and calcification regions based on the image texture patterns. The performance of the proposed algorithm is evaluated by comparing the results to those manually traced borders by an expert on 312 different IVUS images obtained from four different patients. The statistical analysis of the results demonstrates the efficiency of the proposed method in the media-adventitia border detection with enough consistency in the leakage and calcification regions.

An empirical strategy to detect bacterial transcript structure from directional RNA-seq transcriptome data.

BackgroundAs sequencing costs are being lowered continuously, RNA-seq has gradually been adopted as the first choice for comparative transcriptome studies with bacteria. Unlike microarrays, RNA-seq can directly detect cDNA derived from mRNA transcripts at a single nucleotide resolution. Not only does this allow researchers to determine the absolute expression level of genes, but it also conveys information about transcript structure. Few automatic software tools have yet been established to investigate large-scale RNA-seq data for bacterial transcript structure analysis.ResultsIn this study, 54 directional RNA-seq libraries from Salmonella serovar Typhimurium (S. Typhimurium) 14028s were examined for potential relationships between read mapping patterns and transcript structure. We developed an empirical method, combined with statistical tests, to automatically detect key transcript features, including transcriptional start sites (TSSs), transcriptional termination sites (TTSs) and operon organization. Using our method, we obtained 2,764 TSSs and 1,467 TTSs for 1331 and 844 different genes, respectively. Identification of TSSs facilitated further discrimination of 215 putative sigma 38 regulons and 863 potential sigma 70 regulons. Combining the TSSs and TTSs with intergenic distance and co-expression information, we comprehensively annotated the operon organization in S. Typhimurium 14028s.ConclusionsOur results show that directional RNA-seq can be used to detect transcriptional borders at an acceptable resolution of ±10-20 nucleotides. Technical limitations of the RNA-seq procedure may prevent single nucleotide resolution. The automatic transcript border detection methods, statistical models and operon organization pipeline that we have described could be widely applied to RNA-seq studies in other bacteria. Furthermore, the TSSs, TTSs, operons, promoters and unstranslated regions that we have defined for S. Typhimurium 14028s may constitute valuable resources that can be used for comparative analyses with other Salmonella serotypes.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-015-1555-8) contains supplementary material, which is available to authorized users.

An ensemble classification approach for melanoma diagnosis

Malignant melanoma is the deadliest form of skin cancer, and has, among cancer types, one of the most rapidly increasing incidence rates in the world. Early diagnosis is crucial, since if detected early, its cure is simple. In this paper, we present an effective approach to melanoma identification from dermoscopic images of skin lesions based on ensemble classification. First, we perform automatic border detection to segment the lesion from the background skin. Based on the extracted border, we extract a series of colour, texture and shape features. The derived features are then employed in a pattern classification stage for which we employ a novel, dedicated ensemble learning approach to address the class imbalance in the training data and to yield improved classification performance. Our classifier committee trains individual classifiers on balanced subspaces, removes redundant predictors based on a diversity measure and combines the remaining classifiers using a neural network fuser. Experimental results on a large dataset of dermoscopic skin lesion images show our approach to work well, to provide both high sensitivity and specificity, and our presented classifier ensemble to lead to statistically better recognition performance compared to other dedicated classification algorithms.

Important advances in technology: echocardiography.

Echocardiography has evolved over the past 45 years from a simple M-mode tracing to an array of technologies that include two-dimensional imaging, pulsed and continuous wave spectral Doppler, color flow and tissue Doppler, and transesophageal echocardiography. Together, these modalities provide a comprehensive anatomic and functional evaluation of cardiac chambers and valves, pericardium, and ascending and descending aorta. The switch from analog to digital signal processing revolutionized the field of ultrasound, resulting in improved image resolution, smaller instrumentation that allows bedside evaluation and diagnosis of patients, and digital image storage for more accurate quantification and comparison with previous studies. It also opened the door for new advances such as harmonic imaging, automated border detection and quantification, 3-dimensional imaging, and speckle tracking. This article offers an overview of some newer developments in echocardiography and their promising applications.

Diferencias en la infraestimación de los volúmenes del ventrículo izquierdo entre 2 sistemas de ecocardiografía 3D. Comparación con resonancia magnética

Introduction3-D echocardiography (3DE) is superior to the 2-D technique in assessing left ventricular (LV) volumes and ejection fraction (EF). This study compares two 3DE systems with cardiac magnetic resonance imaging (cMRI): single-beat 3DE with automated endocardiac border detection (Siemens SC2000;A) and 3DE with multi-beat acquisition and semi-automated endocardiac borders detection (Philips iE33;B). MethodsA 3DE with single-beat acquisition and automated border detection (A) and 3DE with multi-beat acquisition and semi-automated border detection (B) was performed on 22patients referred for a study using cMRI. ResultsThe mean volumes and EF measured by cMRI were: DTV: 211.07±85.4mL; STV: 119.2±89.25mL; EF: 48.67±20.07%. The correlation between cMRI and the two 3DE systems was excellent for the DTV (rhoA 0.912; rhoB 0.907; P<.0001), STV (rhoA 0.877; rhoB 0.945; P<.0001) and EF (rhoA 0.974; rhoB 0.951; P<.0001). The Bland-Altman analysis showed an underestimation of the LV volumes with both 3DE systems, with the difference being lower in systemA than in systemB (–40±50.2 vs –82.2±40.9mL, respectively for DTV; –16±44.4 vs –43±43.5mL, respectively for STV; P=.001). There were no differences in the measurement of EF between the 3imaging techniques. Conclusion3DE underestimates LV volumes. In our series, this underestimate was lower in the system with single-beat acquisition and automated border detection.

Automatic Border Detection of the Left Ventricle in Parasternal Short Axis View of Echocardiogram

Automatic Border Detection of the Left Ventricle in Parasternal Short Axis View of Echocardiogram

Unified approach for lesion border detection based on mixture modeling and local entropy thresholding

Computer-aided design (CAD) methods are highly valuable for the analysis of skin lesions using digital dermoscopy due to low rate of diagnostic accuracy of expert dermatologist. In computerized diagnostic methods, automatic border detection is the first and crucial step. In this study, a novel unified approach is proposed for automatic border detection (ABD). A preprocessing step is performed by normalized smoothing filter (NSF) to reduce background noise. Mixture models technique is then utilized to initially segment the lesion area roughly. Afterward, local entropy thresholding is performed to extract the lesion candidate pixels and the lesion border is smoothed using morphological reconstruction. The overall ABD system is tested on a set of 100 dermoscopy images with ground truth. A comparative study was conducted with the other three state-of-the-art methods using statistical metrics. This ABD technique has the minimal average error probability rate of 5%, true detection of 92.10% and false positive rate of 6.41%. Results demonstrate that the proposed method segments the lesion area accurately. Sample dataset and execute software are available online and can be downloaded from: http://cs.ntu.edu.pk/research.

Association of age with left ventricular volumes, ejection fraction and concentricity: the Framingham heart study

Background Left ventricular (LV) parameters may vary with aging, but concomitant disease can obscure relationships between age and LV size, mass and function. CMR provides reproducible measures of LV volumes, mass, and ejection fraction (EF), but manual analyses are time consuming. Automated border detection (ABD) methods may decrease analytical burden. We sought to examine the effect of greater age on LV volumes, mass, EF and concentricity in a large cohort of healthy adults free of clinical cardiovascular disease and hypertension using ABDassisted methods.

A perceptually oriented method for contrast enhancement and segmentation of dermoscopy images

Dermoscopy images often suffer from low contrast caused by different light conditions, which reduces the accuracy of lesion border detection. Accordingly for lesion recognition, automatic melanoma border detection (MBD) is an initial as well as crucial task. In this article, a novel perceptually oriented approach for MBD is presented by combing region and edge-based segmentation techniques. The MBD system for color contrast and segmentation improvement consists of four main steps: first, the RGB dermoscopy image is transformed to CIE L*a*b* color space, lesion contrast is then enhanced by adjusting and mapping the intensity values of the lesion pixels in the specified range using the three channels of CIE L*a*b*, a hill-climbing algorithm is used later to detect region-of-interest (ROI) map in a perceptually oriented color space using color channels (L*,a*,b*) and finally, an adaptive thresholding is applied to determine the optimal lesion border. Manually drawn borders obtained from an experienced dermatologist are utilized as a ground truth for performance evaluation. The proposed MBD method is tested on a total of 100 dermoscopy images. A comparative study with three state-of-the-art color and texture-based segmentation techniques (JSeg, dermatologists-like tumor area extraction: DTEA and region-based active contours: RAC), is also conducted to show the effectiveness of our MBD method using measures of true positive rate (TPR), false positive rate (FPR), and error probability (EP). Among different algorithms, our MBD algorithm achieved TPR of 94.25%, FPR of 3.56%, and EP of 4%. The proposed MBD approach is highly accurate to detect the lesion border area. The MBD software and sample of dermoscopy images can be downloaded at http://cs.ntu.edu.pk/research.php.

Reproducibility of IVUS border detection for carotid atherosclerotic plaque assessment

Plaque composition is a potentially important diagnostic feature for carotid artery stenting (CAS). The purpose of this investigation is to evaluate the reproducibility of manual border correction in intravascular ultrasound with virtual histology (VH IVUS) images. Three images each were obtained from 51 CAS datasets on which automatic border detection was corrected manually by two trained observers. Plaque was classified using the definitions from the CAPITAL (Carotid Artery Plaque Virtual Histology Evaluation) study, listed in order from least to most pathological: no plaque, pathological intimal thickening, fibroatheroma, fibrocalcific, calcified fibroatheroma, thin-cap fibroatheroma, and calcified thin-cap fibroatheroma. Inter-observer variability was quantified using both weighted and unweighted Kappa statistics. Bland–Altman analysis was used to compare the cross-sectional areas of the vessel and lumen. Agreement using necrotic core percentage as the criterion was evaluated using the unweighted Kappa statistic. Agreement between classifications of plaque type was evaluated using the weighted Kappa statistic. There was substantial agreement between the observers based on necrotic core percentage (κ=0.63), while the agreement was moderate (κquadratic=0.60) based on plaque classification. Due to the time-consuming nature of manual border detection, an improved automatic border detection algorithm is necessary for using VH IVUS as a diagnostic tool for assessing the suitability of patients with carotid artery occlusive disease for CAS.

Feasibility of Single-Beat Full-Volume Capture Real-Time Three-Dimensional Echocardiography and Auto-Contouring Algorithm for Quantification of Left Ventricular Volume: Validation with Cardiac Magnetic Resonance Imaging

With recent developments in echocardiographic technology, a new system using real-time three-dimensional echocardiography (RT3DE) that allows single-beat acquisition of the entire volume of the left ventricle and incorporates algorithms for automated border detection has been introduced. Provided that these techniques are acceptably reliable, three-dimensional echocardiography may be much more useful for clinical practice. The aim of this study was to evaluate the feasibility and accuracy of left ventricular (LV) volume measurements by RT3DE using the single-beat full-volume capture technique. One hundred nine consecutive patients scheduled for cardiac magnetic resonance imaging and RT3DE using the single-beat full-volume capture technique on the same day were recruited. LV end-systolic volume, end-diastolic volume, and ejection fraction were measured using an auto-contouring algorithm from data acquired on RT3DE. The data were compared with the same measurements obtained using cardiac magnetic resonance imaging. Volume measurements on RT3DE with single-beat full-volume capture were feasible in 84% of patients. Both interobserver and intraobserver variability of three-dimensional measurements of end-systolic and end-diastolic volumes showed excellent agreement. Pearson's correlation analysis showed a close correlation of end-systolic and end-diastolic volumes between RT3DE and cardiac magnetic resonance imaging (r= 0.94 and r= 0.91, respectively, P < .0001 for both). Bland-Altman analysis showed reasonable limits of agreement. After application of the auto-contouring algorithm, the rate of successful auto-contouring (cases requiring minimal manual corrections) was <50%. RT3DE using single-beat full-volume capture is an easy and reliable technique to assess LV volume and systolic function in clinical practice. However, the image quality and low frame rate still limit its application for dilated left ventricles, and the automated volume analysis program needs more development to make it clinically efficacious.

ACCURACY AND REPRODUCIBILITY OF A NOVEL IN VIVO AUTOMATIC BORDER DETECTION SOFTWARE FOR THE ANALYSIS OF STENT NEOINTIMAL RESPONSE COMPARED TO HISTOLOGY IN A FAMILIAL HYPERCHOLESTEROLEMIC SWINE MODEL

Results: 26 OCT frames were analyzed per operator. Lumen area showed high correlation (r=0.96, p < 0.001) and no signiicant difference between automatic (A; 4.93±2 mm2) manual (M; 4.94±2 mm2) or histological assessment (H; 5.27±2.4 mm2, p 0.58). Similar results were obtained on neointimal thickness (A=0.5±0.2mm; M=0.5±0.2mm and H=0.6±0.3mm, p=0.39). The OCT-histology correlation for neointimal thickness was also very high (r=0.95, p<0.001). Stent area show no statistical difference between both OCT techniques (A=9.3±1.4 mm2 vs M=9.3±1.3 mm2, p=0.99). However, the OCT-histology correlation for stent area was lower (OCT= 9.3±1.3 mm2 vs histology= 10.6±1.3mm2, r=0.89, p < 0.001) than seen in the other analyzed variables. Conclusions: The morphometric evaluation of neointimal response using a novel in-vivo automatic border detection software highly correlated with the data obtained by histology. Although still high, the evaluation of stent areas by OCT appears to have a lower degree of correlation in comparison to the other analyzed variables.

Lesion border detection in dermoscopy images using dynamic programming

Automated border detection is an important and challenging task in the computerized analysis of dermoscopy images. However, dermoscopic images often contain artifacts such as illumination, dermoscopic gel, and outline (hair, skin lines, ruler markings, and blood vessels). As a result, there is a need for robust methods to remove artifacts and detect lesion borders in dermoscopy images. This automated method consists of three main steps: (1) preprocessing, (2) edge candidate point detection, and (3) tumor outline delineation. First, algorithms to reduce artifacts were used. Second, a least-squares method (LSM) was performed to acquire edge points. Third, dynamic programming (DP) technique was used to find the optimal boundary of the lesion. Statistical measures based on dermatologist-drawn borders were utilized as ground-truth to evaluate the performance of the proposed method. The method is tested on a total of 240 dermoscopic images: 30 benign melanocytic, 50 malignant melanomas, 50 basal cell carcinomas, 20 Merkel cell carcinomas, 60 seborrheic keratosis, and 30 atypical naevi. We obtained mean border detection error of 8.6%, 5.04%, 9.0%, 7.02%, 2.01%, and 3.24%, respectively. The results demonstrate that border detection combined with artifact removal increases sensitivity and specificity for segmentation of lesions in dermoscopy images.

Real-Time Volumetric Bi-Plane Ultrasound Imaging And Quantification

Quantified measures of a volumetric object in the body can be made ultrasonically by acquiring concurrent biplane images of two different image planes of the object. Corresponding borders of the volumetric object are traced using automatic border detection. The border tracings are used in their planar spatial relationship to compute a graphical model of the volumetric object. The volume of the graphical model may be computed by the rule of disks, and a graphical or numerical display of the changing volume with time displayed. A user interface comprises both real time biplane images, the real time graphical model, and the quantified measures.

Aortic Root Volume and Geometry: Ready for Clinical Application?

Aortic Root Volume and Geometry: Ready for Clinical Application?

Analysis of density based and fuzzy c-means clustering methods on lesion border extraction in dermoscopy images

BackgroundComputer-aided segmentation and border detection in dermoscopic images is one of the core components of diagnostic procedures and therapeutic interventions for skin cancer. Automated assessment tools for dermoscopy images have become an important research field mainly because of inter- and intra-observer variations in human interpretation. In this study, we compare two approaches for automatic border detection in dermoscopy images: density based clustering (DBSCAN) and Fuzzy C-Means (FCM) clustering algorithms. In the first approach, if there exists enough density –greater than certain number of points- around a point, then either a new cluster is formed around the point or an existing cluster grows by including the point and its neighbors. In the second approach FCM clustering is used. This approach has the ability to assign one data point into more than one cluster.ResultsEach approach is examined on a set of 100 dermoscopy images whose manually drawn borders by a dermatologist are used as the ground truth. Error rates; false positives and false negatives along with true positives and true negatives are quantified by comparing results with manually determined borders from a dermatologist. The assessments obtained from both methods are quantitatively analyzed over three accuracy measures: border error, precision, and recall. ConclusionAs well as low border error, high precision and recall, visual outcome showed that the DBSCAN effectively delineated targeted lesion, and has bright future; however, the FCM had poor performance especially in border error metric.

Border detection in dermoscopy images using hybrid thresholding on optimized color channels

Automated border detection is one of the most important steps in dermoscopy image analysis. Although numerous border detection methods have been developed, few studies have focused on determining the optimal color channels for border detection in dermoscopy images. This paper proposes an automatic border detection method which determines the optimal color channels and performs hybrid thresholding to detect the lesion borders. The color optimization process is tested on a set of 30 dermoscopy images with four sets of dermatologist-drawn borders used as the ground truth. The hybrid border detection method is tested on a set of 85 dermoscopy images with two sets of ground truth using various metrics including accuracy, precision, sensitivity, specificity, and border error. The proposed method, which is comprised of two stages, is designed to increase specificity in the first stage and sensitivity in the second stage. It is shown to be highly competitive with three state-of-the-art border detection methods and potentially faster, since it mainly involves scalar processing as opposed to vector processing performed in the other methods. Furthermore, it is shown that our method is as good as, and in some cases more effective than a dermatology registrar.