Accuracy and repeatability of joint sparsity multi-component estimation in MR Fingerprinting.

MR fingerprinting (MRF) is a promising method for quantitative characterization of tissues. Often, voxel-wise measurements are made, assuming a single tissue-type per voxel. Alternatively, the Sparsity Promoting Iterative Joint Non-negative least squares Multi-Component MRF method (SPIJN-MRF) facilitates tissue parameter estimation for identified components as well as partial volume segmentations. The aim of this paper was to evaluate the accuracy and repeatability of the SPIJN-MRF parameter estimations and partial volume segmentations. This was done (1) through numerical simulations based on the BrainWeb phantoms and (2) using in vivo acquired MRF data from 5 subjects that were scanned on the same week-day for 8 consecutive weeks. The partial volume segmentations of the SPIJN-MRF method were compared to those obtained by two conventional methods: SPM12 and FSL. SPIJN-MRF showed higher accuracy in simulations in comparison to FSL- and SPM12-based segmentations: Fuzzy Tanimoto Coefficients (FTC) comparing these segmentations and Brainweb references were higher than 0.95 for SPIJN-MRF in all the tissues and between 0.6 and 0.7 for SPM12 and FSL in white and gray matter and between 0.5 and 0.6 in CSF. For the in vivo MRF data, the estimated relaxation times were in line with literature and minimal variation was observed. Furthermore, the coefficient of variation (CoV) for estimated tissue volumes with SPIJN-MRF were 10.5% for the myelin water, 6.0% for the white matter, 5.6% for the gray matter, 4.6% for the CSF and 1.1% for the total brain volume. CoVs for CSF and total brain volume measured on the scanned data for SPIJN-MRF were in line with those obtained with SPM12 and FSL. The CoVs for white and gray matter volumes were distinctively higher for SPIJN-MRF than those measured with SPM12 and FSL. In conclusion, the use of SPIJN-MRF provides accurate and precise tissue relaxation parameter estimations taking into account intrinsic partial volume effects. It facilitates obtaining tissue fraction maps of prevalent tissues including myelin water which can be relevant for evaluating diseases affecting the white matter.

Assessment of data consistency through cascades of independently recurrent inference machines for fast and robust accelerated MRI reconstruction.

Objective.Machine Learning methods can learn how to reconstruct magnetic resonance images (MRI) and thereby accelerate acquisition, which is of paramount importance to the clinical workflow. Physics-informed networks incorporate the forward model of accelerated MRI reconstruction in the learning process. With increasing network complexity, robustness is not ensured when reconstructing data unseen during training. We aim to embed data consistency (DC) in deep networks while balancing the degree of network complexity. While doing so, we will assess whether either explicit or implicit enforcement of DC in varying network architectures is preferred to optimize performance.Approach.We propose a scheme called Cascades of Independently Recurrent Inference Machines (CIRIM) to assess DC through unrolled optimization. Herein we assess DC both implicitly by gradient descent and explicitly by a designed term. Extensive comparison of the CIRIM to compressed sensing as well as other Machine Learning methods is performed: the End-to-End Variational Network (E2EVN), CascadeNet, KIKINet, LPDNet, RIM, IRIM, and UNet. Models were trained and evaluated on T1-weighted and FLAIR contrast brain data, and T2-weighted knee data. Both 1D and 2D undersampling patterns were evaluated. Robustness was tested by reconstructing 7.5× prospectively undersampled 3D FLAIR MRI data of multiple sclerosis (MS) patients with white matter lesions.Main results.The CIRIM performed best when implicitly enforcing DC, while the E2EVN required an explicit DC formulation. Through its cascades, the CIRIM was able to score higher on structural similarity and PSNR compared to other methods, in particular under heterogeneous imaging conditions. In reconstructing MS patient data, prospectively acquired with a sampling pattern unseen during model training, the CIRIM maintained lesion contrast while efficiently denoising the images.Significance.The CIRIM showed highly promising generalization capabilities maintaining a very fair trade-off between reconstructed image quality and fast reconstruction times, which is crucial in the clinical workflow.

Orientation Prior and Consistent Model Selection Increase Sensitivity of Tract-Based Spatial Statistics in Crossing-Fiber Regions.

The goal of this paper is to increase the statistical power of crossing-fiber statistics in voxelwise analyses of diffusion-weighted magnetic resonance imaging (DW-MRI) data. In the proposed framework, a fiber orientation atlas and a model complexity atlas were used to fit the ball-and-sticks model to diffusion-weighted images of subjects in a prospective population-based cohort study. Reproducibility and sensitivity of the partial volume fractions in the ball-and-sticks model were analyzed using TBSS (tract-based spatial statistics) and compared to a reference framework. The reproducibility was investigated on two scans of 30 subjects acquired with an interval of approximately three weeks by studying the intraclass correlation coefficient (ICC). The sensitivity to true biological effects was evaluated by studying the regression with age on 500 subjects from 65 to 90 years old. Compared to the reference framework, the ICC improved significantly when using the proposed framework. Higher t-statistics indicated that regression coefficients with age could be determined more precisely with the proposed framework and more voxels correlated significantly with age. The application of a fiber orientation atlas and a model complexity atlas can significantly improve the reproducibility and sensitivity of crossing-fiber statistics in TBSS.

Reconstructing 3D proton dose distribution using ionoacoustics.

In proton therapy high energy protons are used to irradiate a tumor. Ideally, the delivered proton dose distribution is measured during treatment to ensure patient safety and treatment effectiveness. Here we investigate if we can use the ionoacoustic wave field to monitor the actual proton dose distribution for the two most commonly used proton accelerators; the isochronous cyclotron and the synchrocyclotron. To this end we model the acoustic field generated by the protons when irradiating a heterogeneous cancerous breast with a 89 MeV proton beam. To differentiate between the systems, idealized temporal micro-structures of the beams have been implemented. Results show that by employing model-based inversion we are able to reconstruct the 3D dose distributions from the simulated noisy pressure fields. Good results are obtained for both systems; the absolute error in the position of the maximum amplitude of the dose distribution is 5.0 mm for the isochronous cyclotron and 5.2 mm for the synchrocyclotron. In conclusion, this numerical study suggests that the ionoacoustic wave field may be used to monitor the proton dose distribution during breast cancer treatment.

Automated versus subjective assessment of spatial and temporal MRI small bowel motility in Crohn's disease.

AIM: To investigate whether subjective radiologist grading of motility on magnetic resonance enterography (MRE) is as effective as software quantification, and to determine the combination of motility metrics with the strongest association with symptom severity. MATERIALS AND METHODS: One hundred and five Crohn's disease patients (52 male, 53 female, 16-68 years old, mean age 34 years old) recruited from two sites underwent MRE, including a 20 second breath-hold cine motility sequence. Each subject completed a Harvey-Bradshaw Index (HBI) symptom questionnaire. Five features within normally appearing bowel were scored visually by two experienced radiologists, and then quantified using automated analysis software, including (1) mean motility, (2) spatial motility variation, (3) temporal motility variation, (4) area of motile bowel, (5) intestinal distension. Multivariable linear regression derived the combination of features with the highest association with HBI score. RESULTS: The best automated metric combination was temporal variation (p<0.05) plus area of motile bowel (p<0.05), achieving an R2 adjusted value of 0.036. Spatial variation was also associated with symptoms (p<0.05, R2 adjusted = 0.034); however, when visually assessed by radiologists, none of the features had a significant relationship with the HBI score. CONCLUSION: Software quantified temporal and spatial variability in bowel motility are associated with abdominal symptoms in Crohn's disease. Subjective radiologist assessment of bowel motility is insufficient to detect aberrant motility. Automated analysis of motility patterns holds promise as an objective biomarker for aberrant physiology underlying symptoms in enteric disorders.

Comparison between dynamic gadoxetate-enhanced MRI and 99mTc-mebrofenin hepatobiliary scintigraphy with SPECT for quantitative assessment of liver function.

OBJECTIVES: To compare Gd-EOB-DTPA dynamic hepatocyte-specific contrast-enhanced MRI (DHCE-MRI) with 99mTc-mebrofenin hepatobiliary scintigraphy (HBS) as quantitative liver function tests for the preoperative assessment of patients undergoing liver resection. METHODS: Patients undergoing liver surgery and preoperative assessment of future remnant liver (FRL) function using 99mTc-mebrofenin HBS were included. Patients underwent DHCE-MRI. Total liver uptake function was calculated for both modalities: mebrofenin uptake rate (MUR) and Ki respectively. The FRL was delineated with both SPECT-CT and MRI to calculate the functional share. Blood samples were taken to assess biochemical liver parameters. RESULTS: A total of 20 patients were included. The HBS-derived MUR and the DHCE-MRI-derived mean Ki correlated strongly for both total and FRL function (Pearson r = 0.70, p = 0.001 and r = 0.89, p < 0.001 respectively). There was a strong agreement between the functional share determined with both modalities (ICC = 0.944, 95% CI 0.863-0.978, n = 20). There was a significant negative correlation between liver aminotransferases and bilirubin for both MUR and Ki. CONCLUSIONS: Assessment of liver function with DHCE-MRI is comparable with that of 99mTc-mebrofenin HBS and has the potential to be combined with diagnostic MRI imaging. This can therefore provide a one-stop-shop modality for the preoperative assessment of patients undergoing liver surgery. KEY POINTS: • Quantitative assessment of liver function using hepatobiliary scintigraphy is performed in the preoperative assessment of patients undergoing liver surgery in order to prevent posthepatectomy liver failure. • Gd-EOB-DTPA dynamic hepatocyte-specific contrast-enhanced MRI (DHCE-MRI) is an emerging method to quantify liver function and can serve as a potential alternative to hepatobiliary scintigraphy. • Assessment of liver function with dynamic gadoxetate-enhanced MRI is comparable with that of hepatobiliary scintigraphy and has the potential to be combined with diagnostic MRI imaging.

A three-dimensional analysis of osteoarthritic changes in the thumb carpometacarpal joint.

The purpose of this study is to gain a better understanding of the changes due to osteoarthritis (OA) occurring in the thumb carpometacarpal (CMC) joint by comparing quantitative geometrical measurements in computed tomography scans of healthy and pathological joints in various stages of OA. The measurements were (1) the subluxation of the metacarpal on the trapezium, (2) distance from the scaphoid centre to the metacarpal base, and (3) distance from the metacarpal base to the articulating surface of the trapezium. The three-dimensional position of three characteristic points on the metacarpal, trapezium, and scaphoid were detected in each of the 90 wrists we scanned. The distances between the points were compared by statistical analysis. With high accuracy, we have been able to confirm and quantify that subluxation occurs in the dorso-radial direction. A significant difference in trapezium height and joint space width was found between the OA and control groups. The results indicate how to restore the centre of rotation in surgical treatment of OA with total joint arthroplasty, but the clinical relevance of these findings has to be tested in further clinical studies.

WYSIWYP: What You See Is What You Pick.

Scientists, engineers and physicians are used to analyze 3D data with slice-based visualizations. Radiologists for example are trained to read slices of medical imaging data. Despite the numerous examples of sophisticated 3D rendering techniques, domain experts, who still prefer slice-based visualization do not consider these to be very useful. Since 3D renderings have the advantage of providing an overview at a glance, while 2D depictions better serve detailed analyses, it is of general interest to better combine these methods. Recently there have been attempts to bridge this gap between 2D and 3D renderings. These attempts include specialized techniques for volume picking in medical imaging data that result in repositioning slices. In this paper, we present a new volume picking technique called WYSIWYP ("what you see is what you pick") that, in contrast to previous work, does not require pre-segmented data or metadata and thus is more generally applicable. The positions picked by our method are solely based on the data itself, the transfer function, and the way the volumetric rendering is perceived by the user. To demonstrate the utility of the proposed method, we apply it to automated positioning of slices in volumetric scalar fields from various application areas. Finally, we present results of a user study in which 3D locations selected by users are compared to those resulting from WYSIWYP. The user study confirms our claim that the resulting positions correlate well with those perceived by the user.

Event-based progression detection strategies using scanning laser polarimetry images of the human retina.

Monitoring glaucoma patients and ensuring optimal treatment requires accurate and precise detection of progression. Many glaucomatous progression detection strategies may be formulated for Scanning Laser Polarimetry (SLP) data of the local nerve fiber thickness. In this paper, several strategies, all based on repeated GDx VCC SLP measurements, are tested to identify the optimal one for clinical use. The parameters of the methods were adapted to yield a set specificity of 97.5% on real image series. For a fixed sensitivity of 90%, the minimally detectable loss was subsequently determined for both localized and diffuse loss. Due to the large size of the required data set, a previously described simulation method was used for assessing the minimally detectable loss. The optimal strategy was identified and was based on two baseline visits and two follow-up visits, requiring two-out-of-four positive tests. Its associated minimally detectable loss was 5-12 µm, depending on the reproducibility of the measurements.

Reducing the oral contrast dose in CT colonography: evaluation of faecal tagging quality and patient acceptance.

AIM: To evaluate the minimal iodine contrast medium load necessary for an optimal computed tomography colonography tagging quality. MATERIALS AND METHODS: Faecal occult blood test positive patients were randomly selected for one of three iodine bowel preparations: (1) 3 × 50 ml meglumine ioxithalamate (45 g iodine), (2) 4 × 25 ml meglumine ioxithalamate (30 g iodine); or (3) 3 × 25 ml (22.5 g iodine) meglumine ioxithalamate. Two experienced readers assessed the tagging quality per colonic segment on a five-point scale and the presence of adherent stool. Also semi-automatic homogeneity measurements were performed. Patient acceptance was assessed with questionnaires. RESULTS: Of 70 eligible patients, 45 patients participated (25 males, mean age 62 years). Each preparation group contained 15 patients. The quality of tagging was insufficient (score 1-2) in 0% of segments in group 1; 4% in group 2 (p<0.01 versus group 1); and 5% in group 3 (p=0.06 versus group 1). In group 1 in 11% of the segments adherent stool was present compared with 49% in group 2 and 41% in group 3 (p<0.01, group 2 and 3 versus group 1). Homogeneity was 85, 102 (p<0.01), and 91 SD HU (p=0.26) in groups 1, 2, and 3, respectively. In group 1 two patients experienced no burden after contrast agent ingestion compared to one patient in group 2 and nine patients in group 3 (p=0.017). CONCLUSION: A dose of 3 × 50 ml meglumine ioxithalamate is advisable for an optimal tagging quality despite beneficial effects on the patient acceptance in patients receiving a lower dose.

Protrusion method for automated estimation of polyp size on CT colonography.

OBJECTIVE: The purpose of this study was to assess the accuracy and measurement variability of automated lesion measurement on CT colonography in comparison with manual 2D and 3D techniques under varying scanning conditions. MATERIALS AND METHODS: The study included phantoms (23 phantom objects) and patients (16 polyps). Measurement with sliding calipers served as the reference for the phantom data. The mean of two independent colonoscopic measurements was the reference for the polyps. The automated measurement was developed for a computer-aided detection scheme, and the size of any detected object was obtained from measurement of its largest diameter. The automated measurement was compared with manual 2D and 3D measurements by two experienced observers. RESULTS: For phantom data, the measurement variability of the automated method was significantly less than that of the two observers (p < 0.05), except for the 3D measurement by observer 1, as follows: automated, 0.86 mm; observer 1, 1.76 mm (2D), 0.96 (3D); observer 2, 1.34 mm (2D), 1.45 mm (3D). The variability of the automated method did not differ significantly from that of manual methods in measurement with patient data. The automated method had a systematic error for phantom data (1.9 mm). CONCLUSION: For phantoms, the automated method has less measurement variability than manual 2D and 3D techniques. For true polyps, the measurement variability of the automated method is comparable with that of manual methods. The automated method does not suffer from intraobserver variability. Because systematic error can be calibrated, automated size measurement may contribute to a practical evaluation strategy.

Split bundle detection in polarimetric images of the human retinal nerve fiber layer.

OBJECTIVES: One method for assessing pathological retinal nerve fiber layer (NFL) appearance is by comparing the NFL to normative values, derived from healthy subjects. These normative values will be more specific when normal physiological differences are taken into account. One common variation is a split bundle. This paper describes a method to automatically detect these split bundles. METHODS: The thickness profile along the NFL bundle is described by a non-split and a split bundle model. Based on these two fits, statistics are derived and used as features for two non-parametric classifiers (Parzen density based and k nearest neighbor). Features were selected by forward feature selection. Three hundred and nine superior and 324 inferior bundles were used to train and test this method. RESULTS: The prevalence of split superior bundles was 68% and the split inferior bundles' prevalence was 13%. The resulting estimated error of the Parzen density- based classifier was 12.5% for the superior bundle and 10.2% for the inferior bundle. The k nearest neighbor classifier errors were 11.7% and 9.2%. CONCLUSIONS: The classification error of automated detection of split inferior bundles is not much smaller than its prevalence, thereby limiting the usefulness of separate cut-off values for split and non-split inferior bundles. For superior bundles, however, the classification error was low compared to the prevalence. Application of specific cut-off values, selected by the proposed classification system, may therefore increase the specificity and sensitivity of pathological NFL detection.

Shaving diffusion tensor images in discriminant analysis: a study into schizophrenia.

A technique called 'shaving' is introduced to automatically extract the combination of relevant image regions in a comparative study. No hypothesis is needed, as in conventional pre-defined or expert selected region of interest (ROI)-analysis. In contrast to traditional voxel based analysis (VBA), correlations within the data can be modeled using principal component analysis (PCA) and linear discriminant analysis (LDA). A study into schizophrenia using diffusion tensor imaging (DTI) serves as an application. Conventional VBA found a decreased fractional anisotropy (FA) in a part of the genu of the corpus callosum and an increased FA in larger parts of white matter. The proposed method reproduced the decrease in FA in the corpus callosum and found an increase in the posterior limb of the internal capsule and uncinate fasciculus. A correlation between the decrease in the corpus callosum and the increase in the uncinate fasciculus was demonstrated.

Automated detection of wedge-shaped defects in polarimetric images of the retinal nerve fibre layer.

PURPOSE: Automated glaucoma detection in images obtained by scanning laser polarimetry is currently insensitive to local abnormalities, impairing its performance. The purpose of this investigation was to test and validate a recently proposed algorithm for detecting wedge-shaped defects. METHODS: In all, 31 eyes of healthy subjects and 37 eyes of glaucoma patients were imaged with a GDx. Each image was classified by two experts in one of four classes, depending on how clear any wedge could be identified. The detection algorithm itself aimed at detecting and combining the edges of the wedge. The performance of both the experts and the algorithm were evaluated. RESULTS: The interobserver correlation, expressed as ICC(3,1), was 0.77. For the clearest cases, the algorithm yielded a sensitivity of 80% at a specificity of 93%, with an area under the ROC of 0.95. Including less obvious cases by the experts resulted in a sensitivity of 55% at a specificity of 95%, with an area under the ROC of 0.89. CONCLUSIONS: It is possible to automatically detect many wedge-shaped defects at a fairly low rate of false-positives. Any detected wedge defect is presented in a user-friendly way, which may assist the clinician in making a diagnosis.

Interactive 3D segmentation using connected orthogonal contours.

This paper describes a new method for interactive segmentation that is based on cross-sectional design and 3D modelling. The method represents a 3D model by a set of connected contours that are planar and orthogonal. Planar contours overlayed on image data are easily manipulated and linked contours reduce the amount of user interaction.1 This method solves the contour-to-contour correspondence problem and can capture extrema of objects in a more flexible way than manual segmentation of a stack of 2D images. The resulting 3D model is guaranteed to be free of geometric and topological errors. We show that manual segmentation using connected orthogonal contours has great advantages over conventional manual segmentation. Furthermore, the method provides effective feedback and control for creating an initial model for, and control and steering of, (semi-)automatic segmentation methods.

Inter-frame motion correction for MR thermometry.

Noninvasive temperature measurement is feasible with MRI to monitor changes in thermal therapy. Phase shift based MR thermometry gives an estimate of the relative temperature variation between thermal and baseline images. This technique is limited, however, when applied on targets under inter-frame motion. Simple image registration and subtraction are not adequate to recover the temperature properly since the phase shift due to temperature changes is corrupted by an unwanted phase shift. In this work, the unwanted phase shift is predicted from the raw registered phase shift map itself. To estimate the unwanted phase shift, a thin plate smoothing spline is fitted to the values outside the heated region. The spline value in the heated area serves as an estimate for the offset. The estimation result is applied to correct errors in the temperature maps of an ex-vivo experiment.

Segmentation and size measurement of polyps in CT colonography.

Virtual colonoscopy is a relatively new method for the detection of colonic polyps. Their size, which is measured from reformatted CT images, mainly determines diagnosis. We present an automatic method for measuring the polyp size. The method is based on a robust segmentation method that grows a surface patch over the entire polyp surface starting from a seed. Projection of the patch points along the polyp axis yields a 2D point set to which we fit an ellipse. The long axis of the ellipse denotes the size of the polyp. We evaluate our method by comparing the automated size measurement with those of two radiologists using scans of a colon phantom. We give data for inter-observer and intra-observer variability of radiologists and our method as well as the accuracy and precision.

A model based method for retinal blood vessel detection.

Retinal blood vessels are important structures in ophthalmological images. Many detection methods are available, but the results are not always satisfactory. In this paper, we present a novel model based method for blood vessel detection in retinal images. It is based on a Laplace and thresholding segmentation step, followed by a classification step to improve performance. The last step assures incorporation of the inner part of large vessels with specular reflection. The method gives a sensitivity of 92% with a specificity of 91%. The method can be optimized for the specific properties of the blood vessels in the image and it allows for detection of vessels that appear to be split due to specular reflection.

Detecting glaucomatous wedge shaped defects in polarimetric images.

Wedge shaped defects of the retinal nerve fiber layer (RNFL) may occur in glaucoma. Currently, automatic detection of wedge shaped defects in Scanning Laser Polarimetry (SLP) images of the RNFL is not available. An automatic classification is currently based only on global parameters, thereby ignoring important local information. Our method works by a modified dynamic programming technique that searches for locally strong edges with a preference for straight edges. These edges are initially classified based on their strength and next combined into wedge shaped defects. Our method yields a sensitivity of 73% and a specificity of 90% on a limited set of 65 images.