Three validation metrics for automated probabilistic image segmentation of brain tumours

Patient-specific temperature distribution prediction in laser interstitial thermal therapy: single-irradiation data-driven method

Laser interstitial thermal therapy (LITT) is popular for treating brain tumours and epilepsy. The strict control of tissue thermal damage extent is crucial for LITT. Temperature prediction is useful for predicting thermal damage extent. Accurately predictingin vivobrain tissue temperature is challenging due to the temperature dependence and the individual variations in tissue properties. Considering these factors is essential for improving the temperature prediction accuracy.Objective. To present a method for predicting patient-specific tissue temperature distribution within a target lesion area in the brain during LITT.Approach. A magnetic resonance temperature imaging (MRTI) data-driven estimation model was constructed and combined with a modified Pennes bioheat transfer equation (PBHE) to predict patient-specific temperature distribution. In the PBHE for temperature prediction, the individual specificity and temperature dependence of thermal tissue properties and blood perfusion, as well as the individual specificity of optical tissue properties were considered. Only MRTI data during one laser irradiation were required in the method. This enables the prediction of patient-specific temperature distribution and the resulting thermal damage region for subsequent ablations.Main results. Patient-specific temperature prediction was evaluated based on clinical data acquired during LITT in the brain, using intraoperative MRTI data as the reference standard. Our method significantly improved the prediction performance of temperature distribution and thermal damage region. The average root mean square error was decreased by 69.54%, the average intraclass correlation coefficient was increased by 37.5%, the average Dice similarity coefficient was increased by 43.14% for thermal damage region prediction.Significance. The proposed method can predict temperature distribution and thermal damage region at an individual patient level during LITT, providing a promising approach to assist in patient-specific treatment planning for LITT in the brain.

A toolkit for stroke infarct volume estimation in rodents

Stroke volume is a key determinant of infarct severity and an important metric for evaluating treatments. However, accurate estimation of stroke volume can be challenging, due to the often confined 2-dimensional nature of available data. Here, we introduce a comprehensive semi-automated toolkit to reliably estimate stroke volumes based on (1) whole brains ex-vivo magnetic resonance imaging (MRI) and (2) brain sections that underwent immunofluorescence staining. We located and quantified infarct areas from MRI three days (acute) and 28 days (chronic) after photothrombotic stroke induction in whole mouse brains. MRI results were compared with measures obtained from immunofluorescent histologic sections of the same brains. We found that infarct volume determined by post-mortem MRI was highly correlated with a deviation of only 6.6 % (acute) and 4.9 % (chronic) to the measurements as determined in the histological brain sections indicating that both methods are capable of accurately assessing brain tissue damage (Pearson r > 0.9, p < 0.001). The Dice similarity coefficient (DC) showed a high degree of coherence (DC > 0.8) between MRI-delineated regions of interest (ROIs) and ROIs obtained from histologic sections at four to six pre-defined landmarks, with histology-based delineation demonstrating higher inter-operator similarity compared to MR images. We further investigated stroke-related scarring and post-ischemic angiogenesis in cortical peri‑infarct regions and described a negative correlation between GFAP+fluorescence intensity and MRI-obtained lesion size.

Measuring intrauterine growth in healthy pregnancies using quantitative magnetic resonance imaging

OBJECTIVE

The aim of this study was to determine in utero fetal-placental growth patterns using in vivo three-dimensional (3D) quantitative magnetic resonance imaging (qMRI).

STUDY DESIGN

Healthy women with singleton pregnancies underwent fetal MRI to measure fetal body, placenta, and amniotic space volumes. The fetal-placental ratio (FPR) was derived using 3D fetal body and placental volumes (PV). Descriptive statistics were used to describe the association of each measurement with increasing gestational age (GA) at MRI.

RESULTS

Fifty-eight (58) women underwent fetal MRI between 16 and 38 completed weeks gestation (mean = 28.12 ± 6.33). PV and FPR varied linearly with GA at MRI (r_PV,GA = 0.83, r_FPR,GA = 0.89, p value < 0.001). Fetal volume varied non-linearly with GA (p value < 0.01).

CONCLUSIONS

We describe in-utero growth trajectories of fetal-placental volumes in healthy pregnancies using qMRI. Understanding healthy in utero development can establish normative benchmarks where departures from normal may identify early in utero placental failure prior to the onset of fetal harm.

Receiver operating characteristic (ROC) curves: equivalences, beta model, and minimum distance estimation

Receiver operating characteristic (ROC) curves are used ubiquitously to evaluate scores, features, covariates or markers as potential predictors in binary problems. We characterize ROC curves from a probabilistic perspective and establish an equivalence between ROC curves and cumulative distribution functions (CDFs). These results support a subtle shift of paradigms in the statistical modelling of ROC curves, which we view as curve fitting. We propose the flexible two-parameter beta family for fitting CDFs to empirical ROC curves and derive the large sample distribution of minimum distance estimators in general parametric settings. In a range of empirical examples the beta family fits better than the classical binormal model, particularly under the vital constraint of the fitted curve being concave.

Automatic pulmonary vessel segmentation on noncontrast chest CT: deep learning algorithm developed using spatiotemporally matched virtual noncontrast images and low-keV contrast-enhanced vessel maps

OBJECTIVES

To develop a deep learning-based pulmonary vessel segmentation algorithm (DLVS) from noncontrast chest CT and to investigate its clinical implications in assessing vascular remodeling of chronic obstructive lung disease (COPD) patients.

METHODS

For development, 104 pulmonary CT angiography scans (49,054 slices) using a dual-source CT were collected, and spatiotemporally matched virtual noncontrast and 50-keV images were generated. Vessel maps were extracted from the 50-keV images. The 3-dimensional U-Net-based DLVS was trained to segment pulmonary vessels (with a vessel map as the output) from virtual noncontrast images (as the input). For external validation, vendor-independent noncontrast CT images (n = 14) and the VESSEL 12 challenge open dataset (n = 3) were used. For each case, 200 points were selected including 20 intra-lesional points, and the probability value for each point was extracted. For clinical validation, we included 281 COPD patients with low-dose noncontrast CTs. The DLVS-calculated volume of vessels with a cross-sectional area < 5 mm2 (PVV5) and the PVV5 divided by total vessel volume (%PVV5) were measured.

RESULTS

DLVS correctly segmented 99.1% of the intravascular points (1,387/1,400) and 93.1% of the extravascular points (1,309/1,400). The areas-under-the receiver-operating characteristic curve (AUROCs) were 0.977 and 0.969 for the two external validation datasets. For the COPD patients, both PPV5 and %PPV5 successfully differentiated severe patients whose FEV1 < 50 (AUROCs; 0.715 and 0.804) and were significantly correlated with the emphysema index (Ps < .05).

CONCLUSIONS

DLVS successfully segmented pulmonary vessels on noncontrast chest CT by utilizing spatiotemporally matched 50-keV images from a dual-source CT scanner and showed promising clinical applicability in COPD.

KEY POINTS

• We developed a deep learning pulmonary vessel segmentation algorithm using virtual noncontrast images and 50-keV enhanced images produced by a dual-source CT scanner. • Our algorithm successfully segmented vessels on diseased lungs. • Our algorithm showed promising results in assessing the loss of small vessel density in COPD patients.

Standardizing human brain parcellations

Using brain atlases to localize regions of interest is a requirement for making neuroscientifically valid statistical inferences. These atlases, represented in volumetric or surface coordinate spaces, can describe brain topology from a variety of perspectives. Although many human brain atlases have circulated the field over the past fifty years, limited effort has been devoted to their standardization. Standardization can facilitate consistency and transparency with respect to orientation, resolution, labeling scheme, file storage format, and coordinate space designation. Our group has worked to consolidate an extensive selection of popular human brain atlases into a single, curated, open-source library, where they are stored following a standardized protocol with accompanying metadata, which can serve as the basis for future atlases. The repository containing the atlases, the specification, as well as relevant transformation functions is available in the neuroparc OSF registered repository or https://github.com/neurodata/neuroparc .

Manual segmentation versus semi-automated segmentation for quantifying vestibular schwannoma volume on MRI

Purpose

Management of vestibular schwannoma (VS) is based on tumour size as observed on T1 MRI scans with contrast agent injection. The current clinical practice is to measure the diameter of the tumour in its largest dimension. It has been shown that volumetric measurement is more accurate and more reliable as a measure of VS size. The reference approach to achieve such volumetry is to manually segment the tumour, which is a time intensive task. We suggest that semi-automated segmentation may be a clinically applicable solution to this problem and that it could replace linear measurements as the clinical standard.

Methods

Using high-quality software available for academic purposes, we ran a comparative study of manual versus semi-automated segmentation of VS on MRI with 5 clinicians and scientists. We gathered both quantitative and qualitative data to compare the two approaches; including segmentation time, segmentation effort and segmentation accuracy.

Results

We found that the selected semi-automated segmentation approach is significantly faster (167 s vs 479 s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p<0.001$$\end{document}p<0.001), less temporally and physically demanding and has approximately equal performance when compared with manual segmentation, with some improvements in accuracy. There were some limitations, including algorithmic unpredictability and error, which produced more frustration and increased mental effort in comparison with manual segmentation.

Conclusion

We suggest that semi-automated segmentation could be applied clinically for volumetric measurement of VS on MRI. In future, the generic software could be refined for use specifically for VS segmentation, thereby improving accuracy.

Electronic supplementary material

The online version of this article (10.1007/s11548-020-02222-y) contains supplementary material, which is available to authorized users.

AUTOMATIC BRAIN DOSE ESTIMATION IN COMPUTED TOMOGRAPHY USING PATIENT DICOM IMAGES

This study aims to develop an Automatic Brain Dose Estimation (ABDE) methodology for head computed tomography examinations. The ABDE is to be applied first to an anthropomorphic Alderson phantom to obtain a Correction factor (Cf) between the ABDE and the direct absorbed brain dose using dosemeters positioned within the anthropomorphic phantom. Then, in order to estimate the correct brain dose for patient, the Cf was multiplied by the mean ABDE values for each patient. Results were compared to those registered with a mathematical simulation phantom using CT-Expo V 2.4 software. Results showed no significant difference between the correct ABDE values and the CT-Expo values with a mean percent difference of 2.54 ± 0.01%. In conclusion, ABDE yields a correct estimation of brain dose, taking into account the size and attenuation of the irradiated region. Thus, it is clinically recommended for accurate patient brain dose assessment.

Tsallis-Entropy Segmentation through MRF and Alzheimer anatomic reference for Brain Magnetic Resonance Parcellation

Quantifying the intracranial tissue volume changes in magnetic resonance imaging (MRI) assists specialists to analyze the effects of natural or pathological changes. Since these changes can be subtle, the accuracy of the automatic compartmentalization method is always criticized by specialists. We propose and then evaluate an automatic segmentation method based on modified q-entropy (Mqe) through a modified Markov Random Field (MMRF) enhanced by Alzheimer anatomic reference (AAR) to provide a high accuracy brain tissues parcellation approach (Mqe-MMRF). We underwent two strategies to evaluate Mqe-MMRF; a simulation of different levels of noise and non-uniformity effect on MRI data (7 subjects) and a set of twenty MRI data available from MRBrainS13 as patient brain tissue segmentation challenge. We accessed eleven quality metrics compared to reference tissues delineations to evaluate Mqe-MMRF. MRI segmentation scores decreased by only 4.6% on quality metrics after noise and non-uniformity simulations of 40% and 9%, respectively. We found significant mean improvements in the metrics of the five training subjects, for whole-brain 0.86%, White Matter 3.20%, Gray Matter 3.99%, and Cerebrospinal Fluid 4.16% (p-values < 0.02) when Mqe-MMRF compared to the other reference methods. We also processed the Mqe-MMRF on 15 evaluation subjects group from MRBrainS13 online challenge, and the results held a higher rank than the reference tools; FreeSurfer, SPM, and FSL. Since the proposed method improved the precision of brain segmentation, specifically, for GM, and thus one can use it in quantitative and morphological brain studies.

Non-monotone transformation of biomarkers to improve diagnostic and screening accuracy in a DNA methylation study with trichotomous phenotypes

We propose a non-monotone transformation to biomarkers in order to improve the diagnostic and screening accuracy. The proposed quadratic transformation only involves modeling the distribution means and variances of the biomarkers and is therefore easy to implement in practice. Mathematical justification was rigorously established to support the validity of the proposed transformation. We conducted extensive simulation studies to assess the performance of the proposed method and compared the new method with the traditional methods. Case studies on real biomedical and epigenetics data were provided to illustrate the proposed transformation. In particular, the proposed method improved the AUC values for a large number of markers in a DNA methylation study and consequently led to the identification of greater number of important biomarkers and biologically meaningful genetic pathways.

In vivo textural and morphometric analysis of placental development in healthy & growth-restricted pregnancies using magnetic resonance imaging

BACKGROUND

The objective of this study was to characterize structural changes in the healthy in vivo placenta by applying morphometric and textural analysis using magnetic resonance imaging (MRI), and to explore features that may be able to distinguish placental insufficiency in fetal growth restriction (FGR).

METHODS

Women with healthy pregnancies or pregnancies complicated by FGR underwent MRI between 20 and 40 weeks gestation. Measures of placental morphometry (volume, elongation, depth) and digital texture (voxel-wise geometric and signal-intensity analysis) were calculated from T2W MR images.

RESULTS

We studied 66 pregnant women (32 healthy controls, 34 FGR); during the study period, placentas undergo significant increases in size; signal intensity remains relatively constant, however there is increasing variation in spatial arrangements, suggestive of progressive microstructural heterogeneity. In FGR, placental size is smaller, with great homogeneity of signal intensity and spatial arrangements.

CONCLUSION

We report quantitative textural and morphometric changes in the in vivo placenta in healthy controls over the second half of pregnancy. These MRI features demonstrate important differences in placental development in the setting of placental insufficiency that relate to onset and severity of FGR, as well as neonatal outcome.

Automated Volumetric Assessment of Hepatocellular Carcinoma Response to Sorafenib: A Pilot Study

PURPOSE

This pilot study evaluates the feasibility of automated volumetric quantification of hepatocellular carcinoma (HCC) as an imaging biomarker to assess treatment response for sorafenib.

METHODS

In this institutional review board-approved, Health Insurance Portability and Accountability Act-compliant retrospective study, a training database of manually labeled background liver, enhancing and nonenhancing tumor tissue was established using pretherapy and first posttherapy multiphasic computed tomography images from a registry of 13 HCC patients. For each patient, Hounsfield density and geometry-based feature images were generated from registered multiphasic computed tomography data sets and used as the input for a random forest-based classifier of enhancing and nonenhancing tumor tissue. Leave-one-out cross-validation of the dice similarity measure was applied to quantify the classifier accuracy. A Cox regression model was used to confirm volume changes as predictors of time to progression (TTP) of target lesions for both manual and automatic methods.

RESULTS

When compared with manual labels, an overall classification accuracy of dice similarity coefficient of 0.71 for pretherapy and 0.66 posttherapy enhancing tumor labels and 0.45 for pretherapy and 0.59 for posttherapy nonenhancing tumor labels was observed. Automated methods for quantifying volumetric changes in the enhancing lesion agreed with manual methods and were observed as a significant predictor of TTP.

CONCLUSIONS

Automated volumetric analysis was determined to be feasible for monitoring HCC response to treatment. The information extracted using automated volumetrics is likely to reproduce labor-intensive manual data and provide a good predictor for TTP. Further work will extend these studies to additional treatment modalities and larger patient populations.

A novel voxel-wise lesion segmentation technique on 3.0-T diffusion MRI of hyperacute focal cerebral ischemia at 1 h after permanent MCAO in rats

To assess hyperacute focal cerebral ischemia in rats on 3.0-Tesla diffusion-weighted imaging (DWI), we developed a novel voxel-wise lesion segmentation technique that overcomes intra- and inter-subject variation in apparent diffusion coefficient (ADC) distribution. Our novel technique involves the following: (1) intensity normalization including determination of the optimal type of region of interest (ROI) and its intra- and inter-subject validation, (2) verification of focal cerebral ischemic lesions at 1 h with gross and high-magnification light microscopy of hematoxylin-eosin (H&E) pathology, (3) voxel-wise segmentation on ADC with various thresholds, and (4) calculation of dice indices (DIs) to compare focal cerebral ischemic lesions at 1 h defined by ADC and matching H&E pathology. The best coefficient of variation was the mode of the left hemisphere after normalization using whole left hemispheric ROI, which showed lower intra- (2.54 ± 0.72%) and inter-subject (2.67 ± 0.70%) values than the original. Focal ischemic lesion at 1 h after middle cerebral artery occlusion (MCAO) was confirmed on both gross and microscopic H&E pathology. The 83 relative threshold of normalized ADC showed the highest mean DI (DI = 0.820 ± 0.075). We could evaluate hyperacute ischemic lesions at 1 h more reliably on 3-Tesla DWI in rat brains.

Orbital shape in intentional skull deformations and adult sagittal craniosynostoses

Intentional cranial deformations are the result of external mechanical forces exerted on the skull vault that modify the morphology of various craniofacial structures such as the skull base, the orbits and the zygoma. In this controlled study, we investigated the 3D shape of the orbital inner mould and the orbital volume in various types of intentional deformations and in adult non-operated scaphocephaly - the most common type of craniosynostosis - using dedicated morphometric methods. CT scans were performed on 32 adult skulls with intentional deformations, 21 adult skull with scaphocephaly and 17 non-deformed adult skulls from the collections of the Muséum national d'Histoire naturelle in Paris, France. The intentional deformations group included six skulls with Toulouse deformations, eight skulls with circumferential deformations and 18 skulls with antero-posterior deformations. Mean shape models were generated based on a semi-automatic segmentation technique. Orbits were then aligned and compared qualitatively and quantitatively using colour-coded distance maps and by computing the mean absolute distance, the Hausdorff distance, and the Dice similarity coefficient. Orbital symmetry was assessed after mirroring, superimposition and Dice similarity coefficient computation. We showed that orbital shapes were significantly and symmetrically modified in intentional deformations and scaphocephaly compared with non-deformed control skulls. Antero-posterior and circumferential deformations demonstrated a similar and severe orbital deformation pattern resulting in significant smaller orbital volumes. Scaphocephaly and Toulouse deformations had similar deformation patterns but had no effect on orbital volumes. This study showed that intentional deformations and scaphocephaly significantly interact with orbital growth. Our approach was nevertheless not sufficient to identify specific modifications caused by the different types of skull deformations or by scaphocephaly.

Autodelineation of cervical cancers using multiparametric magnetic resonance imaging and machine learning

BACKGROUND

Tumour delineation is a challenging, time-consuming and complex part of radiotherapy planning. In this study, an automatic method for delineating locally advanced cervical cancers was developed using a machine learning approach.

MATERIALS AND METHODS

A method for tumour segmentation based on image voxel classification using Fisher?s Linear Discriminant Analysis (LDA) was developed. This was applied to magnetic resonance (MR) images of 78 patients with locally advanced cervical cancer. The segmentation was based on multiparametric MRI consisting of T2- weighted (T2w), T1-weighted (T1w) and dynamic contrast-enhanced (DCE) sequences, and included intensity and spatial information from the images. The model was trained and assessed using delineations made by two radiologists.

RESULTS

Segmentation based on T2w or T1w images resulted in mean sensitivity and specificity of 94% and 52%, respectively. Including DCE-MR images improved the segmentation model?s performance significantly, giving mean sensitivity and specificity of 85?93%. Comparisons with radiologists? tumour delineations gave Dice similarity coefficients of up to 0.44.

CONCLUSION

Voxel classification using a machine learning approach is a flexible and fully automatic method for tumour delineation. Combining all relevant MR image series resulted in high sensitivity and specificity. Moreover, the presented method can be extended to include additional imaging modalities.

Referenceless magnetic resonance temperature imaging using Gaussian process modeling

PURPOSE

During magnetic resonance (MR)-guided thermal therapies, water proton resonance frequency shift (PRFS)-based MR temperature imaging can quantitatively monitor tissue temperature changes. It is widely known that the PRFS technique is easily perturbed by tissue motion, tissue susceptibility changes, magnetic field drift, and modality-dependent applicator-induced artifacts. Here, a referenceless Gaussian process modeling (GPM)-based estimation of the PRFS is investigated as a methodology to mitigate unwanted background field changes. The GPM offers a complementary trade-off between data fitting and smoothing and allows prior information to be used. The end result being the GPM provides a full probabilistic prediction and an estimate of the uncertainty.

METHODS

GPM was employed to estimate the covariance between the spatial position and MR phase measurements. The mean and variance provided by the statistical model extrapolated background phase values from nonheated neighboring voxels used to train the model. MR phase predictions in the heating ROI are computed using the spatial coordinates as the test input. The method is demonstrated in ex vivo rabbit liver tissue during focused ultrasound heating with manually introduced perturbations (n = 6) and in vivo during laser-induced interstitial thermal therapy to treat the human brain (n = 1) and liver (n = 1).

RESULTS

Temperature maps estimated using the GPM referenceless method demonstrated a RMS error of <0.8°C with artifact-induced reference-based MR thermometry during ex vivo heating using focused ultrasound. Nonheated surrounding areas were <0.5°C from the artifact-free MR measurements. The GPM referenceless MR temperature values and thermally damaged regions were within the 95% confidence interval during in vivo laser ablations.

CONCLUSIONS

A new approach to estimation for referenceless PRFS temperature imaging is introduced that allows for an accurate probabilistic extrapolation of the background phase. The technique demonstrated reliable temperature estimates in the presence of the background phase changes and was demonstrated useful in the in vivo brain and liver ablation scenarios presented.

A model evaluation study for treatment planning of laser-induced thermal therapy

A cross-validation analysis evaluating computer model prediction accuracy for a priori planning magnetic resonance-guided laser-induced thermal therapy (MRgLITT) procedures in treating focal diseased brain tissue is presented. Two mathematical models are considered. (1) A spectral element discretisation of the transient Pennes bioheat transfer equation is implemented to predict the laser-induced heating in perfused tissue. (2) A closed-form algorithm for predicting the steady-state heat transfer from a linear superposition of analytic point source heating functions is also considered. Prediction accuracy is retrospectively evaluated via leave-one-out cross-validation (LOOCV). Modelling predictions are quantitatively evaluated in terms of a Dice similarity coefficient (DSC) between the simulated thermal dose and thermal dose information contained within N = 22 MR thermometry datasets. During LOOCV analysis, the transient model's DSC mean and median are 0.7323 and 0.8001 respectively, with 15 of 22 DSC values exceeding the success criterion of DSC ≥ 0.7. The steady-state model's DSC mean and median are 0.6431 and 0.6770 respectively, with 10 of 22 passing. A one-sample, one-sided Wilcoxon signed-rank test indicates that the transient finite element method model achieves the prediction success criteria, DSC ≥ 0.7, at a statistically significant level.

Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool

BACKGROUND

Medical Image segmentation is an important image processing step. Comparing images to evaluate the quality of segmentation is an essential part of measuring progress in this research area. Some of the challenges in evaluating medical segmentation are: metric selection, the use in the literature of multiple definitions for certain metrics, inefficiency of the metric calculation implementations leading to difficulties with large volumes, and lack of support for fuzzy segmentation by existing metrics.

RESULT

First we present an overview of 20 evaluation metrics selected based on a comprehensive literature review. For fuzzy segmentation, which shows the level of membership of each voxel to multiple classes, fuzzy definitions of all metrics are provided. We present a discussion about metric properties to provide a guide for selecting evaluation metrics. Finally, we propose an efficient evaluation tool implementing the 20 selected metrics. The tool is optimized to perform efficiently in terms of speed and required memory, also if the image size is extremely large as in the case of whole body MRI or CT volume segmentation. An implementation of this tool is available as an open source project.

CONCLUSION

We propose an efficient evaluation tool for 3D medical image segmentation using 20 evaluation metrics and provide guidelines for selecting a subset of these metrics that is suitable for the data and the segmentation task.

Computer Vision Techniques for Transcatheter Intervention

Minimally invasive transcatheter technologies have demonstrated substantial promise for the diagnosis and the treatment of cardiovascular diseases. For example, transcatheter aortic valve implantation is an alternative to aortic valve replacement for the treatment of severe aortic stenosis, and transcatheter atrial fibrillation ablation is widely used for the treatment and the cure of atrial fibrillation. In addition, catheter-based intravascular ultrasound and optical coherence tomography imaging of coronary arteries provides important information about the coronary lumen, wall, and plaque characteristics. Qualitative and quantitative analysis of these cross-sectional image data will be beneficial to the evaluation and the treatment of coronary artery diseases such as atherosclerosis. In all the phases (preoperative, intraoperative, and postoperative) during the transcatheter intervention procedure, computer vision techniques (e.g., image segmentation and motion tracking) have been largely applied in the field to accomplish tasks like annulus measurement, valve selection, catheter placement control, and vessel centerline extraction. This provides beneficial guidance for the clinicians in surgical planning, disease diagnosis, and treatment assessment. In this paper, we present a systematical review on these state-of-the-art methods. We aim to give a comprehensive overview for researchers in the area of computer vision on the subject of transcatheter intervention. Research in medical computing is multi-disciplinary due to its nature, and hence, it is important to understand the application domain, clinical background, and imaging modality, so that methods and quantitative measurements derived from analyzing the imaging data are appropriate and meaningful. We thus provide an overview on the background information of the transcatheter intervention procedures, as well as a review of the computer vision techniques and methodologies applied in this area.

Objective detection of retinal vessel pulsation

Purpose

Retinal venous pulsation detection is a subjective sign, which varies in elevated intracranial pressure, venous obstruction and glaucoma. To date no method can objectively measure and identify pulsating regions.

Method

Using high resolution video-recordings of the optic disk and retina we measured fluctuating light absorption by haemoglobin during pulsation. Pulsation amplitude was calculated from all regions of the retinal image video-frames in a raster pattern. Segmented retinal images were formed by objectively selecting regions with amplitudes above a range of threshold values. These were compared to two observers manually drawing an outline of the pulsating areas while viewing video-clips in order to generate receiver operator characteristics.

Results

216,515 image segments were analysed from 26 eyes in 18 research participants. Using data from each eye, the median area under the receiver operator curve (AU-ROC) was 0.95. With all data analysed together the AU-ROC was 0.89. We defined the ideal threshold amplitude for detection of any pulsating segment being that with maximal sensitivity and specificity. This was 5 units (95% confidence interval 4.3 to 6.0) compared to 12 units before any regions were missed. A multivariate model demonstrated that ideal threshold amplitude increased with increased variation in video-sequence illumination (p = 0.0119), but between the two observers (p = 0.0919) or other variables.

Conclusion

This technique demonstrates accurate identification of retinal vessel pulsating regions with no areas identified manually being missed with the objective technique. The amplitude values are derived objectively and may be a significant advance upon subjective ophthalmodynamometric threshold techniques.

Morphometry-based measurements of the structural response to whole-brain radiation

PURPOSE

Morphometry techniques were applied to quantify the normal tissue therapy response in patients receiving whole-brain radiation for intracranial malignancies.

METHODS

Pre- and Post-irradiation magnetic resonance imaging (MRI) data sets were retrospectively analyzed in N = 15 patients. Volume changes with respect to pre-irradiation were quantitatively measured in the cerebrum and ventricles. Measurements were correlated with the time interval from irradiation. Criteria for inclusion included craniospinal irradiation, pre-irradiation MRI, at least one follow-up MRI, and no disease progression. The brain on each image was segmented to remove the skull and registered to the initial pre-treatment scan. Average volume changes were measured using morphometry analysis of the deformation Jacobian and direct template registration-based segmentation of brain structures.

RESULTS

An average cerebral volume atrophy of -0.2 and -3% 3% was measured for the deformation morphometry and direct segmentation methods, respectively. An average ventricle volume dilation of 21 and 20% was measured for the deformation morphometry and direct segmentation methods, respectively.

CONCLUSION

The presented study has developed an image processing pipeline for morphometric monitoring of brain tissue volume changes as a response to radiation therapy. Results indicate that quantitative morphometric monitoring is feasible and may provide additional information in assessing response.

A non-parametric permutation method for assessing agreement for distance matrix observations

Distance matrix data are occurring ever more frequently in medical research, particularly in fields such as genetics, DNA research, and image analysis. We propose a non-parametric permutation method for assessing agreement when the data under study are distance matrices. We apply agglomerative hierarchical clustering and accompanying dendrograms to visualize the internal structure of the matrix observations. The accompanying test is based on random permutations of the elements within individual matrix observations and the corresponding matrix mean of these permutations. We compare the within-matrix element sum of squares (WMESS) for the observed mean against the WMESS for the permutation means. The methodology is exemplified using simulations and real data from magnetic resonance imaging.

Automated identification of brain tumors from single MR images based on segmentation with refined patient-specific priors

Brain tumors can have different shapes or locations, making their identification very challenging. In functional MRI, it is not unusual that patients have only one anatomical image due to time and financial constraints. Here, we provide a modified automatic lesion identification (ALI) procedure which enables brain tumor identification from single MR images. Our method rests on (A) a modified segmentation-normalization procedure with an explicit “extra prior” for the tumor and (B) an outlier detection procedure for abnormal voxel (i.e., tumor) classification. To minimize tissue misclassification, the segmentation-normalization procedure requires prior information of the tumor location and extent. We therefore propose that ALI is run iteratively so that the output of Step B is used as a patient-specific prior in Step A. We test this procedure on real T1-weighted images from 18 patients, and the results were validated in comparison to two independent observers' manual tracings. The automated procedure identified the tumors successfully with an excellent agreement with the manual segmentation (area under the ROC curve = 0.97 ± 0.03). The proposed procedure increases the flexibility and robustness of the ALI tool and will be particularly useful for lesion-behavior mapping studies, or when lesion identification and/or spatial normalization are problematic.

Optimal thresholds by maximizing or minimizing various metrics via ROC-type analysis

RATIONALE AND OBJECTIVES

Based on imaging features, the optimal thresholds are typically determined as cutoff points to dichotomize the corresponding measurement scales.

MATERIALS AND METHODS

Five metrics (ie, the Youden index, Euclidian distance, percent of correct diagnosis, kappa statistic, and mutual information) are individually maximized or minimized to derive the corresponding optimal threshold. These optimal thresholds are estimated under the parametric binormal assumption. Monte Carlo simulation studies are conducted to compare the performances of these different methods. A published radiological example on the choice of treatment outcomes following ureteral stones is used to illustrate and compare the estimated thresholds both empirically and parametrically.

RESULTS

The optimal threshold can be a "moving target" because it would depend on modeling assumptions, metrics, and variability in the data. Even with large samples, disease prevalence has an impact on the robustness of the metrics.

CONCLUSIONS

It is recommended that researchers compare different optimal cutoff points using several metrics and select one that is most clinically relevant. The ultimate goal is to maximize diagnostic performances that are clinically meaningful to achieve improved global health.

A survey of MRI-based medical image analysis for brain tumor studies

MRI-based medical image analysis for brain tumor studies is gaining attention in recent times due to an increased need for efficient and objective evaluation of large amounts of data. While the pioneering approaches applying automated methods for the analysis of brain tumor images date back almost two decades, the current methods are becoming more mature and coming closer to routine clinical application. This review aims to provide a comprehensive overview by giving a brief introduction to brain tumors and imaging of brain tumors first. Then, we review the state of the art in segmentation, registration and modeling related to tumor-bearing brain images with a focus on gliomas. The objective in the segmentation is outlining the tumor including its sub-compartments and surrounding tissues, while the main challenge in registration and modeling is the handling of morphological changes caused by the tumor. The qualities of different approaches are discussed with a focus on methods that can be applied on standard clinical imaging protocols. Finally, a critical assessment of the current state is performed and future developments and trends are addressed, giving special attention to recent developments in radiological tumor assessment guidelines.

Applications of ROC analysis in medical research: recent developments and future directions

With the growing focus on comparative effectiveness research and personalized medicine, receiver-operating characteristic analysis can continue to play an important role in health care decision making. Specific applications of receiver-operating characteristic analysis include predictive model assessment and validation, biomarker diagnostics, responder analysis in patient-reported outcomes, and comparison of alternative treatment options. The authors present a survey of the potential applications of the method and briefly review several relevant extensions. Given the level of attention paid to biomarker validation, personalized medicine and comparative effectiveness research, it is highly likely that the receiver-operating characteristic analysis will remain an important visual and analytic tool for medical research and evidence-based medicine in the foreseeable future.

Accuracy and reliability of automated gray matter segmentation pathways on real and simulated structural magnetic resonance images of the human brain

Automated gray matter segmentation of magnetic resonance imaging data is essential for morphometric analyses of the brain, particularly when large sample sizes are investigated. However, although detection of small structural brain differences may fundamentally depend on the method used, both accuracy and reliability of different automated segmentation algorithms have rarely been compared. Here, performance of the segmentation algorithms provided by SPM8, VBM8, FSL and FreeSurfer was quantified on simulated and real magnetic resonance imaging data. First, accuracy was assessed by comparing segmentations of twenty simulated and 18 real T1 images with corresponding ground truth images. Second, reliability was determined in ten T1 images from the same subject and in ten T1 images of different subjects scanned twice. Third, the impact of preprocessing steps on segmentation accuracy was investigated. VBM8 showed a very high accuracy and a very high reliability. FSL achieved the highest accuracy but demonstrated poor reliability and FreeSurfer showed the lowest accuracy, but high reliability. An universally valid recommendation on how to implement morphometric analyses is not warranted due to the vast number of scanning and analysis parameters. However, our analysis suggests that researchers can optimize their individual processing procedures with respect to final segmentation quality and exemplifies adequate performance criteria.

Task-based evaluation of segmentation algorithms for diffusion-weighted MRI without using a gold standard

In many studies, the estimation of the apparent diffusion coefficient (ADC) of lesions in visceral organs in diffusion-weighted (DW) magnetic resonance images requires an accurate lesion-segmentation algorithm. To evaluate these lesion-segmentation algorithms, region-overlap measures are used currently. However, the end task from the DW images is accurate ADC estimation, and the region-overlap measures do not evaluate the segmentation algorithms on this task. Moreover, these measures rely on the existence of gold-standard segmentation of the lesion, which is typically unavailable. In this paper, we study the problem of task-based evaluation of segmentation algorithms in DW imaging in the absence of a gold standard. We first show that using manual segmentations instead of gold-standard segmentations for this task-based evaluation is unreliable. We then propose a method to compare the segmentation algorithms that does not require gold-standard or manual segmentation results. The no-gold-standard method estimates the bias and the variance of the error between the true ADC values and the ADC values estimated using the automated segmentation algorithm. The method can be used to rank the segmentation algorithms on the basis of both the ensemble mean square error and precision. We also propose consistency checks for this evaluation technique.

Kalman filtered MR temperature imaging for laser induced thermal therapies

The feasibility of using a stochastic form of Pennes bioheat model within a 3-D finite element based Kalman filter (KF) algorithm is critically evaluated for the ability to provide temperature field estimates in the event of magnetic resonance temperature imaging (MRTI) data loss during laser induced thermal therapy (LITT). The ability to recover missing MRTI data was analyzed by systematically removing spatiotemporal information from a clinical MR-guided LITT procedure in human brain and comparing predictions in these regions to the original measurements. Performance was quantitatively evaluated in terms of a dimensionless L(2) (RMS) norm of the temperature error weighted by acquisition uncertainty. During periods of no data corruption, observed error histories demonstrate that the Kalman algorithm does not alter the high quality temperature measurement provided by MR thermal imaging. The KF-MRTI implementation considered is seen to predict the bioheat transfer with RMS error < 4 for a short period of time, ∆t < 10 s, until the data corruption subsides. In its present form, the KF-MRTI method currently fails to compensate for consecutive for consecutive time periods of data loss ∆t > 10 sec.

Development of image-processing software for automatic segmentation of brain tumors in MR images

Most of the commercially available software for brain tumor segmentation have limited functionality and frequently lack the careful validation that is required for clinical studies. We have developed an image-analysis software package called ‘Prometheus,’ which performs neural system–based segmentation operations on MR images using pre-trained information. The software also has the capability to improve its segmentation performance by using the training module of the neural system. The aim of this article is to present the design and modules of this software. The segmentation module of Prometheus can be used primarily for image analysis in MR images. Prometheus was validated against manual segmentation by a radiologist and its mean sensitivity and specificity was found to be 85.71±4.89% and 93.2±2.87%, respectively. Similarly, the mean segmentation accuracy and mean correspondence ratio was found to be 92.35±3.37% and 0.78±0.046, respectively.

Quantitative comparison of thermal dose models in normal canine brain

PURPOSE

Minimally invasive thermal ablative therapies as alternatives to conventional surgical management of solid tumors and other pathologies is increasing owing to the potential benefits of performing these procedures in an outpatient setting with reduced complications and comorbidity. Magnetic resonance temperature imaging (MRTI) measurement allows existing thermal dose models to use the spatiotemporal temperature history to estimate the thermal damage to tissue. However, the various thermal dose models presented in the literature employ different parameters and thresholds, affecting the reliability of thermal dosimetry. In this study, the authors quantitatively compared three thermal dose models (Arrhenius rate process, CEM43, and threshold temperature) using the dice similarity coefficient (DSC).

METHODS

The DSC was used to compare the spatial overlap between the region of thermal damage as predicted by the models for in vivo normal canine brain during thermal therapy to the region of thermal damage as revealed by contrast-enhanced T1-weighted images acquired immediately after therapy (< 20 min). The outer edge of the hyperintense rim of the ablation region was used as the surrogate marker for the limits of thermal coagulation. The DSC was also used to investigate the impact of varying the thresholds on each models' ability to predict the zone of thermal necrosis.

RESULTS

At previously reported thresholds, the authors found that all three models showed good agreement (defined as DSC > 0.7) with post-treatment imaging. All three models examined across the range of commonly applied thresholds consistently showed highly accurate spatial overlap, low variability, and little dependence on temperature uncertainty. DSC values corresponding to cited thresholds were not significantly different from peak DSC values.

CONCLUSIONS

Thus, the authors conclude that the all three thermal dose models can be used as a reliable surrogate for postcontrast tissue damage verification imaging in rapid ablation procedures and can also be used to enhance the capability of MRTI to control thermal therapy in real time.

PET-guided delineation of radiation therapy treatment volumes: a survey of image segmentation techniques

Historically, anatomical CT and MR images were used to delineate the gross tumour volumes (GTVs) for radiotherapy treatment planning. The capabilities offered by modern radiation therapy units and the widespread availability of combined PET/CT scanners stimulated the development of biological PET imaging-guided radiation therapy treatment planning with the aim to produce highly conformal radiation dose distribution to the tumour. One of the most difficult issues facing PET-based treatment planning is the accurate delineation of target regions from typical blurred and noisy functional images. The major problems encountered are image segmentation and imperfect system response function. Image segmentation is defined as the process of classifying the voxels of an image into a set of distinct classes. The difficulty in PET image segmentation is compounded by the low spatial resolution and high noise characteristics of PET images. Despite the difficulties and known limitations, several image segmentation approaches have been proposed and used in the clinical setting including thresholding, edge detection, region growing, clustering, stochastic models, deformable models, classifiers and several other approaches. A detailed description of the various approaches proposed in the literature is reviewed. Moreover, we also briefly discuss some important considerations and limitations of the widely used techniques to guide practitioners in the field of radiation oncology. The strategies followed for validation and comparative assessment of various PET segmentation approaches are described. Future opportunities and the current challenges facing the adoption of PET-guided delineation of target volumes and its role in basic and clinical research are also addressed.

Statistical evaluations of the reproducibility and reliability of 3-tesla high resolution magnetization transfer brain images: a pilot study on healthy subjects

Magnetization transfer imaging (MT) may have considerable promise for early detection and monitoring of subtle brain changes before they are apparent on conventional magnetic resonance images. At 3 Tesla (T), MT affords higher resolution and increased tissue contrast associated with macromolecules. The reliability and reproducibility of a new high-resolution MT strategy were assessed in brain images acquired from 9 healthy subjects. Repeated measures were taken for 12 brain regions of interest (ROIs): genu, splenium, and the left and right hemispheres of the hippocampus, caudate, putamen, thalamus, and cerebral white matter. Spearman's correlation coefficient, coefficient of variation, and intraclass correlation coefficient (ICC) were computed. Multivariate mixed-effects regression models were used to fit the mean ROI values and to test the significance of the effects due to region, subject, observer, time, and manual repetition. A sensitivity analysis of various model specifications and the corresponding ICCs was conducted. Our statistical methods may be generalized to many similar evaluative studies of the reliability and reproducibility of various imaging modalities.

Evaluation of uterine cervix segmentations using ground truth from multiple experts

This work is focused on the generation and utilization of a reliable ground truth (GT) segmentation for a large medical repository of digital cervicographic images (cervigrams) collected by the National Cancer Institute (NCI). NCI invited twenty experts to manually segment a set of 939 cervigrams into regions of medical and anatomical interest. Based on this unique data, the objectives of the current work are to: (1) Automatically generate a multi-expert GT segmentation map; (2) Use the GT map to automatically assess the complexity of a given segmentation task; (3) Use the GT map to evaluate the performance of an automated segmentation algorithm. The multi-expert GT map is generated via the STAPLE (Simultaneous Truth and Performance Level Estimation) algorithm, which is a well-known method to generate a GT segmentation from multiple observations. A new measure of segmentation complexity, which relies on the inter-observer variability within the GT map, is defined. This measure is used to identify images that were found difficult to segment by the experts and to compare the complexity of different segmentation tasks. An accuracy measure, which evaluates the performance of automated segmentation algorithms is presented. Two algorithms for cervix boundary detection are compared using the proposed accuracy measure. The measure is shown to reflect the actual segmentation quality achieved by the algorithms. The methods and conclusions presented in this work are general and can be applied to different images and segmentation tasks. Here they are applied to the cervigram database including a thorough analysis of the available data.

Concurrent multimodality image segmentation by active contours for radiotherapy treatment planning

Multimodality imaging information is regularly used now in radiotherapy treatment planning for cancer patients. The authors are investigating methods to take advantage of all the imaging information available for joint target registration and segmentation, including multimodality images or multiple image sets from the same modality. In particular, the authors have developed variational methods based on multivalued level set deformable models for simultaneous 2D or 3D segmentation of multimodality images consisting of combinations of coregistered PET, CT, or MR data sets. The combined information is integrated to define the overall biophysical structure volume. The authors demonstrate the methods on three patient data sets, including a nonsmall cell lung cancer case with PET/CT, a cervix cancer case with PET/CT, and a prostate patient case with CT and MRI. CT, PET, and MR phantom data were also used for quantitative validation of the proposed multimodality segmentation approach. The corresponding Dice similarity coefficient (DSC) was 0.90 +/- 0.02 (p < 0.0001) with an estimated target volume error of 1.28 +/- 1.23% volume. Preliminary results indicate that concurrent multimodality segmentation methods can provide a feasible and accurate framework for combining imaging data from different modalities and are potentially useful tools for the delineation of biophysical structure volumes in radiotherapy treatment planning.

Validation of phalanx bone three-dimensional surface segmentation from computed tomography images using laser scanning

OBJECTIVE

To examine the validity of manually defined bony regions of interest from computed tomography (CT) scans.

MATERIALS AND METHODS

Segmentation measurements were performed on the coronal reformatted CT images of the three phalanx bones of the index finger from five cadaveric specimens. Two smoothing algorithms (image-based and Laplacian surface-based) were evaluated to determine their ability to represent accurately the anatomic surface. The resulting surfaces were compared with laser surface scans of the corresponding cadaveric specimen.

RESULTS

The average relative overlap between two tracers was 0.91 for all bones. The overall mean difference between the manual unsmoothed surface and the laser surface scan was 0.20 mm. Both image-based and Laplacian surface-based smoothing were compared; the overall mean difference for image-based smoothing was 0.21 mm and 0.20 mm for Laplacian smoothing.

CONCLUSIONS

This study showed that manual segmentation of high-contrast, coronal, reformatted, CT datasets can accurately represent the true surface geometry of bones. Additionally, smoothing techniques did not significantly alter the surface representations. This validation technique should be extended to other bones, image segmentation and spatial filtering techniques.

A hierarchical 3D segmentation method and the definition of vertebral body coordinate systems for QCT of the lumbar spine

We have developed a new hierarchical 3D technique to segment the vertebral bodies in order to measure bone mineral density (BMD) with high trueness and precision in volumetric CT datasets. The hierarchical approach starts with a coarse separation of the individual vertebrae, applies a variety of techniques to segment the vertebral bodies with increasing detail and ends with the definition of an anatomic coordinate system for each vertebral body, relative to which up to 41 trabecular and cortical volumes of interest are positioned. In a pre-segmentation step constraints consisting of Boolean combinations of simple geometric shapes are determined that enclose each individual vertebral body. Bound by these constraints viscous deformable models are used to segment the main shape of the vertebral bodies. Volume growing and morphological operations then capture the fine details of the bone-soft tissue interface. In the volumes of interest bone mineral density and content are determined. In addition, in the segmented vertebral bodies geometric parameters such as volume or the length of the main axes of inertia can be measured. Intra- and inter-operator precision errors of the segmentation procedure were analyzed using existing clinical patient datasets. Results for segmented volume, BMD, and coordinate system position were below 2.0%, 0.6%, and 0.7%, respectively. Trueness was analyzed using phantom scans. The bias of the segmented volume was below 4%; for BMD it was below 1.5%. The long-term goal of this work is improved fracture prediction and patient monitoring in the field of osteoporosis. A true 3D segmentation also enables an accurate measurement of geometrical parameters that may augment the clinical value of a pure BMD analysis.

Quantitative evaluation of automated skull-stripping methods applied to contemporary and legacy images: effects of diagnosis, bias correction, and slice location

Performance of automated methods to isolate brain from nonbrain tissues in magnetic resonance (MR) structural images may be influenced by MR signal inhomogeneities, type of MR image set, regional anatomy, and age and diagnosis of subjects studied. The present study compared the performance of four methods: Brain Extraction Tool (BET; Smith [2002]: Hum Brain Mapp 17:143-155); 3dIntracranial (Ward [1999] Milwaukee: Biophysics Research Institute, Medical College of Wisconsin; in AFNI); a Hybrid Watershed algorithm (HWA, Segonne et al. [2004] Neuroimage 22:1060-1075; in FreeSurfer); and Brain Surface Extractor (BSE, Sandor and Leahy [1997] IEEE Trans Med Imag 16:41-54; Shattuck et al. [2001] Neuroimage 13:856-876) to manually stripped images. The methods were applied to uncorrected and bias-corrected datasets; Legacy and Contemporary T1-weighted image sets; and four diagnostic groups (depressed, Alzheimer's, young and elderly control). To provide a criterion for outcome assessment, two experts manually stripped six sagittal sections for each dataset in locations where brain and nonbrain tissue are difficult to distinguish. Methods were compared on Jaccard similarity coefficients, Hausdorff distances, and an Expectation-Maximization algorithm. Methods tended to perform better on contemporary datasets; bias correction did not significantly improve method performance. Mesial sections were most difficult for all methods. Although AD image sets were most difficult to strip, HWA and BSE were more robust across diagnostic groups compared with 3dIntracranial and BET. With respect to specificity, BSE tended to perform best across all groups, whereas HWA was more sensitive than other methods. The results of this study may direct users towards a method appropriate to their T1-weighted datasets and improve the efficiency of processing for large, multisite neuroimaging studies.

Validation of bone segmentation and improved 3-D registration using contour coherency in CT data

A method is presented to validate the segmentation of computed tomography (CT) image sequences, and improve the accuracy and efficiency of the subsequent registration of the three-dimensional surfaces that are reconstructed from the segmented slices. The method compares the shapes of contours extracted from neighborhoods of slices in CT stacks of tibias. The bone is first segmented by an automatic segmentation technique, and the bone contour for each slice is parameterized as a one-dimensional function of normalized arc length versus inscribed angle. These functions are represented as vectors within a K-dimensional space comprising the first K amplitude coefficients of their Fourier Descriptors. The similarity or coherency of neighboring contours is measured by comparing statistical properties of their vector representations within this space. Experimentation has demonstrated this technique to be very effective at identifying low-coherency segmentations. Compared with experienced human operators, in a set of 23 CT stacks (1,633 slices), the method correctly detected 87.5% and 80% of the low-coherency and 97.7% and 95.5% of the high coherency segmentations, respectively from two different automatic segmentation techniques. Removal of the automatically detected low-coherency segmentations also significantly improved the accuracy and time efficiency of the registration of 3-D bone surface models. The registration error was reduced by over 500% (i.e., a factor of 5) and 280%, and the computational performance was improved by 540% and 791% for the two respective segmentation methods.

Reproducibility of functional MR imaging: preliminary results of prospective multi-institutional study performed by Biomedical Informatics Research Network

PURPOSE

To prospectively investigate the factors--including subject, brain hemisphere, study site, field strength, imaging unit vendor, imaging run, and examination visit--affecting the reproducibility of functional magnetic resonance (MR) imaging activations based on a repeated sensory-motor (SM) task.

MATERIALS AND METHODS

The institutional review boards of all participating sites approved this HIPAA-compliant study. All subjects gave informed consent. Functional MR imaging data were repeatedly acquired from five healthy men aged 20-29 years who performed the same SM task at 10 sites. Five 1.5-T MR imaging units, four 3.0-T units, and one 4.0-T unit were used. The subjects performed bilateral finger tapping on button boxes with a 3-Hz audio cue and a reversing checkerboard. In a block design, 15-second epochs of alternating baseline and tasks yielded 85 acquisitions per run. Functional MR images were acquired with block-design echo-planar or spiral gradient-echo sequences. Brain activation maps standardized in a unit-sphere for the left and right hemispheres of each subject were constructed. Areas under the receiver operating characteristic curve, intraclass correlation coefficients, multiple regression analysis, and paired Student t tests were used for statistical analyses.

RESULTS

Significant factors were subject (P < .005), k-space (P < .005), and field strength (P = .02) for sensitivity and subject (P = .03) and k-space (P = .05) for specificity. At 1.5-T MR imaging, mean sensitivities ranged from 7% to 32% and mean specificities were higher than 99%. At 3.0 T, mean sensitivities and specificities ranged from 42% to 85% and from 96% to 99%, respectively. At 4.0 T, mean sensitivities and specificities ranged from 41% to 73% and from 95% to 99%, respectively. Mean areas under the receiver operating characteristic curve (+/- their standard errors) were 0.77 +/- 0.05 at 1.5 T, 0.90 +/- 0.09 at 3.0 T, and 0.95 +/- 0.02 at 4.0 T, with significant differences between the 1.5- and 3.0-T examinations and between the 1.5- and 4.0-T examinations (P < .01 for both comparisons). Intraclass correlation coefficients ranged from 0.49 to 0.71.

CONCLUSION

MR imaging at 3.0- and 4.0-T yielded higher reproducibility across sites and significantly better results than 1.5-T imaging. The effects of subject, k-space, and field strength on examination reproducibility were significant.