Validation and Development of a New Automatic Algorithm for Time-Resolved Segmentation of the Left Ventricle in Magnetic Resonance Imaging.
ABSTRACT: Manual delineation of the left ventricle is clinical standard for quantification of cardiovascular magnetic resonance images despite being time consuming and observer dependent. Previous automatic methods generally do not account for one major contributor to stroke volume, the long-axis motion. Therefore, the aim of this study was to develop and validate an automatic algorithm for time-resolved segmentation covering the whole left ventricle, including basal slices affected by long-axis motion.Ninety subjects imaged with a cine balanced steady state free precession sequence were included in the study (training set n = 40, test set n = 50). Manual delineation was reference standard and second observer analysis was performed in a subset (n = 25). The automatic algorithm uses deformable model with expectation-maximization, followed by automatic removal of papillary muscles and detection of the outflow tract.The mean differences between automatic segmentation and manual delineation were EDV -11 mL, ESV 1 mL, EF -3%, and LVM 4 g in the test set.The automatic LV segmentation algorithm reached accuracy comparable to interobserver for manual delineation, thereby bringing automatic segmentation one step closer to clinical routine. The algorithm and all images with manual delineations are available for benchmarking.
Project description:BACKGROUND:Efficacy of reperfusion therapy can be assessed as myocardial salvage index (MSI) by determining the size of myocardium at risk (MaR) and myocardial infarction (MI), (MSI?=?1-MI/MaR). Cardiovascular magnetic resonance (CMR) can be used to assess MI by late gadolinium enhancement (LGE) and MaR by either T2-weighted imaging or contrast enhanced SSFP (CE-SSFP). Automatic segmentation algorithms have been developed and validated for MI by LGE as well as for MaR by T2-weighted imaging. There are, however, no algorithms available for CE-SSFP. Therefore, the aim of this study was to develop and validate automatic segmentation of MaR in CE-SSFP. METHODS:The automatic algorithm applies surface coil intensity correction and classifies myocardial intensities by Expectation Maximization to define a MaR region based on a priori regional criteria, and infarct region from LGE. Automatic segmentation was validated against manual delineation by expert readers in 183 patients with reperfused acute MI from two multi-center randomized clinical trials (RCT) (CHILL-MI and MITOCARE) and against myocardial perfusion SPECT in an additional set (n?=?16). Endocardial and epicardial borders were manually delineated at end-diastole and end-systole. Manual delineation of MaR was used as reference and inter-observer variability was assessed for both manual delineation and automatic segmentation of MaR in a subset of patients (n?=?15). MaR was expressed as percent of left ventricular mass (%LVM) and analyzed by bias (mean?±?standard deviation). Regional agreement was analyzed by Dice Similarity Coefficient (DSC) (mean?±?standard deviation). RESULTS:MaR assessed by manual and automatic segmentation were 36?±?10% and 37?±?11%LVM respectively with bias 1?±?6%LVM and regional agreement DSC 0.85?±?0.08 (n?=?183). MaR assessed by SPECT and CE-SSFP automatic segmentation were 27?±?10%LVM and 29?±?7%LVM respectively with bias 2?±?7%LVM. Inter-observer variability was 0?±?3%LVM for manual delineation and -1?±?2%LVM for automatic segmentation. CONCLUSIONS:Automatic segmentation of MaR in CE-SSFP was validated against manual delineation in multi-center, multi-vendor studies with low bias and high regional agreement. Bias and variability was similar to inter-observer variability of manual delineation and inter-observer variability was decreased by automatic segmentation. Thus, the proposed automatic segmentation can be used to reduce subjectivity in quantification of MaR in RCT. CLINICAL TRIAL REGISTRATION:NCT01379261. NCT01374321.
Project description:BACKGROUND: Intensity modulated radiotherapy for head and neck cancer necessitates accurate definition of organs at risk (OAR) and clinical target volumes (CTV). This crucial step is time consuming and prone to inter- and intra-observer variations. Automatic segmentation by atlas deformable registration may help to reduce time and variations. We aim to test a new commercial atlas algorithm for automatic segmentation of OAR and CTV in both ideal and clinical conditions. METHODS: The updated Brainlab automatic head and neck atlas segmentation was tested on 20 patients: 10 cN0-stages (ideal population) and 10 unselected N-stages (clinical population). Following manual delineation of OAR and CTV, automatic segmentation of the same set of structures was performed and afterwards manually corrected. Dice Similarity Coefficient (DSC), Average Surface Distance (ASD) and Maximal Surface Distance (MSD) were calculated for "manual to automatic" and "manual to corrected" volumes comparisons. RESULTS: In both groups, automatic segmentation saved about 40% of the corresponding manual segmentation time. This effect was more pronounced for OAR than for CTV. The edition of the automatically obtained contours significantly improved DSC, ASD and MSD. Large distortions of normal anatomy or lack of iodine contrast were the limiting factors. CONCLUSIONS: The updated Brainlab atlas-based automatic segmentation tool for head and neck Cancer patients is timesaving but still necessitates review and corrections by an expert.
Project description:Blood flow measurements in the ascending aorta and pulmonary artery from phase-contrast magnetic resonance images require accurate time-resolved vessel segmentation over the cardiac cycle. Current semi-automatic segmentation methods often involve time-consuming manual correction, relying on user experience for accurate results. The purpose of this study was to develop a semi-automatic vessel segmentation algorithm with shape constraints based on manual vessel delineations for robust segmentation of the ascending aorta and pulmonary artery, to evaluate the proposed method in healthy volunteers and patients with heart failure and congenital heart disease, to validate the method in a pulsatile flow phantom experiment, and to make the method freely available for research purposes. Algorithm shape constraints were extracted from manual reference delineations of the ascending aorta (n = 20) and pulmonary artery (n = 20) and were included in a semi-automatic segmentation method only requiring manual delineation in one image. Bias and variability (bias ± SD) for flow volume of the proposed algorithm versus manual reference delineations were 0·0 ± 1·9 ml in the ascending aorta (n = 151; seven healthy volunteers; 144 heart failure patients) and -1·7 ± 2·9 ml in the pulmonary artery (n = 40; 25 healthy volunteers; 15 patients with atrial septal defect). Interobserver bias and variability were lower (P = 0·008) for the proposed semi-automatic method (-0·1 ± 0·9 ml) compared to manual reference delineations (1·5 ± 5·1 ml). Phantom validation showed good agreement between the proposed method and timer-and-beaker flow volumes (0·4 ± 2·7 ml). In conclusion, the proposed semi-automatic vessel segmentation algorithm can be used for efficient analysis of flow and shunt volumes in the aorta and pulmonary artery.
Project description:PURPOSE:To evaluate the uncertainty of radiomics features from contrast-enhanced breath-hold helical CT scans of non-small cell lung cancer for both manual and semi-automatic segmentation due to intra-observer, inter-observer, and inter-software reliability. METHODS:Three radiation oncologists manually delineated lung tumors twice from 10 CT scans using two software tools (3D-Slicer and MIM Maestro). Additionally, three observers without formal clinical training were instructed to use two semi-automatic segmentation tools, Lesion Sizing Toolkit (LSTK) and GrowCut, to delineate the same tumor volumes. The accuracy of the semi-automatic contours was assessed by comparison with physician manual contours using Dice similarity coefficients and Hausdorff distances. Eighty-three radiomics features were calculated for each delineated tumor contour. Informative features were identified based on their dynamic range and correlation to other features. Feature reliability was then evaluated using intra-class correlation coefficients (ICC). Feature range was used to evaluate the uncertainty of the segmentation methods. RESULTS:From the initial set of 83 features, 40 radiomics features were found to be informative, and these 40 features were used in the subsequent analyses. For both intra-observer and inter-observer reliability, LSTK had higher reliability than GrowCut and the two manual segmentation tools. All observers achieved consistently high ICC values when using LSTK, but the ICC value varied greatly for each observer when using GrowCut and the manual segmentation tools. For inter-software reliability, features were not reproducible across the software tools for either manual or semi-automatic segmentation methods. Additionally, no feature category was found to be more reproducible than another feature category. Feature ranges of LSTK contours were smaller than those of manual contours for all features. CONCLUSION:Radiomics features extracted from LSTK contours were highly reliable across and among observers. With semi-automatic segmentation tools, observers without formal clinical training were comparable to physicians in evaluating tumor segmentation.
Project description:Growing interest has developed in hippocampal subfield volumetry over the past few years and an increasing number of studies use the automatic segmentation algorithm implemented in FreeSurfer. However, this approach has not been validated on standard resolution T1-weighted magnetic resonance (MR) as used in most studies. We aimed at comparing hippocampal subfield segmentation using FreeSurfer on standard T1-weighted images versus manual delineation on dedicated high-resolution hippocampal scans. Hippocampal subfields were segmented in 133 individuals including 98 cognitively normal controls aged 19-84 years, 17 mild cognitive impairment and 18 Alzheimer's disease (AD) patients using both methods. Intraclass correlation coefficients (ICC) and Bland-Altman plots were computed to assess the consistency between both methods, and the effects of age and diagnosis were assessed from both measures. Low to moderate ICC (0.31-0.74) were found for the subiculum and other subfields as well as for the whole hippocampus, and the correlations were very low for cornu ammonis (CA)1 (<0.1). FreeSurfer CA1 volume estimates were found to be much lower than those obtained from manual segmentation, and this bias was proportional to the volume of this structure so that no effect of age or AD could be detected on FreeSurfer CA1 volumes. This study points to the differences in the anatomic definition of the subfields between FreeSurfer and manual delineation, especially for CA1, and provides clue for improvement of this automatic technique for potential clinical application on standard T1-weighted MR.
Project description:BACKGROUND: Two-dimensional echocardiography (2D-echo) allows the evaluation of cardiac structures and their movements. A wide range of clinical diagnoses are based on the performance of the left ventricle. The evaluation of myocardial function is typically performed by manual segmentation of the ventricular cavity in a series of dynamic images. This process is laborious and operator dependent. The automatic segmentation of the left ventricle in 4-chamber long-axis images during diastole is troublesome, because of the opening of the mitral valve. METHODS: This work presents a method for segmentation of the left ventricle in dynamic 2D-echo 4-chamber long-axis images over the complete cardiac cycle. The proposed algorithm is based on classic image processing techniques, including time-averaging and wavelet-based denoising, edge enhancement filtering, morphological operations, homotopy modification, and watershed segmentation. The proposed method is semi-automatic, requiring a single user intervention for identification of the position of the mitral valve in the first temporal frame of the video sequence. Image segmentation is performed on a set of dynamic 2D-echo images collected from an examination covering two consecutive cardiac cycles. RESULTS: The proposed method is demonstrated and evaluated on twelve healthy volunteers. The results are quantitatively evaluated using four different metrics, in a comparison with contours manually segmented by a specialist, and with four alternative methods from the literature. The method's intra- and inter-operator variabilities are also evaluated. CONCLUSIONS: The proposed method allows the automatic construction of the area variation curve of the left ventricle corresponding to a complete cardiac cycle. This may potentially be used for the identification of several clinical parameters, including the area variation fraction. This parameter could potentially be used for evaluating the global systolic function of the left ventricle.
Project description:OBJECTIVE:To evaluate the accuracy of a deep learning-based automated segmentation of the left ventricle (LV) myocardium using cardiac CT. MATERIALS AND METHODS:To develop a fully automated algorithm, 100 subjects with coronary artery disease were randomly selected as a development set (50 training / 20 validation / 30 internal test). An experienced cardiac radiologist generated the manual segmentation of the development set. The trained model was evaluated using 1000 validation set generated by an experienced technician. Visual assessment was performed to compare the manual and automatic segmentations. In a quantitative analysis, sensitivity and specificity were calculated according to the number of pixels where two three-dimensional masks of the manual and deep learning segmentations overlapped. Similarity indices, such as the Dice similarity coefficient (DSC), were used to evaluate the margin of each segmented masks. RESULTS:The sensitivity and specificity of automated segmentation for each segment (1-16 segments) were high (85.5-100.0%). The DSC was 88.3 ± 6.2%. Among randomly selected 100 cases, all manual segmentation and deep learning masks for visual analysis were classified as very accurate to mostly accurate and there were no inaccurate cases (manual vs. deep learning: very accurate, 31 vs. 53; accurate, 64 vs. 39; mostly accurate, 15 vs. 8). The number of very accurate cases for deep learning masks was greater than that for manually segmented masks. CONCLUSION:We present deep learning-based automatic segmentation of the LV myocardium and the results are comparable to manual segmentation data with high sensitivity, specificity, and high similarity scores.
Project description:BACKGROUND:PET-based tumor delineation is an error prone and labor intensive part of image analysis. Especially for patients with advanced disease showing bulky tumor FDG load, segmentations are challenging. Reducing the amount of user-interaction in the segmentation might help to facilitate segmentation tasks especially when labeling bulky and complex tumors. Therefore, this study reports on segmentation workflows/strategies that may reduce the inter-observer variability for large tumors with complex shapes with different levels of user-interaction. METHODS:Twenty PET images of bulky tumors were delineated independently by six observers using four strategies: (I) manual, (II) interactive threshold-based, (III) interactive threshold-based segmentation with the additional presentation of the PET-gradient image and (IV) the selection of the most reasonable result out of four established semi-automatic segmentation algorithms (Select-the-best approach). The segmentations were compared using Jaccard coefficients (JC) and percentage volume differences. To obtain a reference standard, a majority vote (MV) segmentation was calculated including all segmentations of experienced observers. Performed and MV segmentations were compared regarding positive predictive value (PPV), sensitivity (SE), and percentage volume differences. RESULTS:The results show that with decreasing user-interaction the inter-observer variability decreases. JC values and percentage volume differences of Select-the-best and a workflow including gradient information were significantly better than the measurements of the other segmentation strategies (p-value<0.01). Interactive threshold-based and manual segmentations also result in significant lower and more variable PPV/SE values when compared with the MV segmentation. CONCLUSIONS:FDG PET segmentations of bulky tumors using strategies with lower user-interaction showed less inter-observer variability. None of the methods led to good results in all cases, but use of either the gradient or the Select-the-best workflow did outperform the other strategies tested and may be a good candidate for fast and reliable labeling of bulky and heterogeneous tumors.
Project description:Total brain white matter lesion (WML) volume is the most widely established magnetic resonance imaging (MRI) outcome measure in studies of multiple sclerosis (MS). To estimate WML volume, there are a number of automatic segmentation methods available, yet manual delineation remains the gold standard approach. Automatic approaches often yield a probability map to which a threshold is applied to create lesion segmentation masks. Unfortunately, few approaches systematically determine the threshold employed; many methods use a manually selected threshold, thus introducing human error and bias into the automated procedure. In this study, we propose and validate an automatic thresholding algorithm, Thresholding Approach for Probability Map Automatic Segmentation in Multiple Sclerosis (TAPAS), to obtain subject-specific threshold estimates for probability map automatic segmentation of T2-weighted (T2) hyperintense WMLs. Using multimodal MRI, the proposed method applies an automatic segmentation algorithm to obtain probability maps. We obtain the true subject-specific threshold that maximizes the Sørensen-Dice similarity coefficient (DSC). Then the subject-specific thresholds are modeled on a naive estimate of volume using a generalized additive model. Applying this model, we predict a subject-specific threshold in data not used for training. We ran a Monte Carlo-resampled split-sample cross-validation (100 validation sets) using two data sets: the first obtained from the Johns Hopkins Hospital (JHH) on a Philips 3 Tesla (3T) scanner (n = 94) and a second collected at the Brigham and Women's Hospital (BWH) using a Siemens 3T scanner (n = 40). By means of the proposed automated technique, in the JHH data we found an average reduction in subject-level absolute error of 0.1 mL per one mL increase in manual volume. Using Bland-Altman analysis, we found that volumetric bias associated with group-level thresholding was mitigated when applying TAPAS. The BWH data showed similar absolute error estimates using group-level thresholding or TAPAS likely since Bland-Altman analyses indicated no systematic biases associated with group or TAPAS volume estimates. The current study presents the first validated fully automated method for subject-specific threshold prediction to segment brain lesions.
Project description:Due to advances in the acquisition and analysis of medical imaging, it is currently possible to quantify the tumor phenotype. The emerging field of Radiomics addresses this issue by converting medical images into minable data by extracting a large number of quantitative imaging features. One of the main challenges of Radiomics is tumor segmentation. Where manual delineation is time consuming and prone to inter-observer variability, it has been shown that semi-automated approaches are fast and reduce inter-observer variability. In this study, a semiautomatic region growing volumetric segmentation algorithm, implemented in the free and publicly available 3D-Slicer platform, was investigated in terms of its robustness for quantitative imaging feature extraction. Fifty-six 3D-radiomic features, quantifying phenotypic differences based on tumor intensity, shape and texture, were extracted from the computed tomography images of twenty lung cancer patients. These radiomic features were derived from the 3D-tumor volumes defined by three independent observers twice using 3D-Slicer, and compared to manual slice-by-slice delineations of five independent physicians in terms of intra-class correlation coefficient (ICC) and feature range. Radiomic features extracted from 3D-Slicer segmentations had significantly higher reproducibility (ICC?=?0.85±0.15, p?=?0.0009) compared to the features extracted from the manual segmentations (ICC?=?0.77±0.17). Furthermore, we found that features extracted from 3D-Slicer segmentations were more robust, as the range was significantly smaller across observers (p?=?3.819e-07), and overlapping with the feature ranges extracted from manual contouring (boundary lower: p?=?0.007, higher: p?=?5.863e-06). Our results show that 3D-Slicer segmented tumor volumes provide a better alternative to the manual delineation for feature quantification, as they yield more reproducible imaging descriptors. Therefore, 3D-Slicer can be employed for quantitative image feature extraction and image data mining research in large patient cohorts.