Detecting Cocircular Subsets of a Spherical Set of Points.

Given a spherical set of points, we consider the detection of cocircular subsets of the data. We distinguish great circles from small circles, and develop algorithms for detecting cocircularities of both types. The suggested approach is an extension of the Hough transform. We address the unique parameter-space quantization issues arising due to the spherical geometry, present quantization schemes, and evaluate the quantization-induced errors. We demonstrate the proposed algorithms by detecting cocircular cities and airports on Earth's spherical surface. These results facilitate the detection of great and small circles in spherical images.

Anisotropic neural deblurring for MRI acceleration.

PURPOSE: MRI has become the tool of choice for brain imaging, providing unrivalled contrast between soft tissues, as well as a wealth of information about anatomy, function, and neurochemistry. Image quality, in terms of spatial resolution and noise, is strongly dependent on acquisition duration. A typical brain MRI scan may last several minutes, with total protocol duration often exceeding 30 minutes. Long scan duration leads to poor patient experience, long waiting time for appointments, and high costs. Therefore, shortening MRI scans is crucial. In this paper, we investigate the enhancement of low-resolution (LR) brain MRI scanning, to enable shorter acquisition times without compromising the diagnostic value of the images. METHODS: We propose a novel fully convolutional neural enhancement approach. It is optimized for accelerated LR MRI acquisitions obtained by reducing the acquisition matrix size only along phase encoding direction. The network is trained to transform the LR acquisitions into corresponding high-resolution (HR) counterparts in an end-to-end manner. In contrast to previous neural-based MRI enhancement algorithms, such as DAGAN, the LR images used for training are real acquisitions rather than smoothed, downsampled versions of the HR images. RESULTS: The proposed method is validated qualitatively and quantitatively for an acceleration factor of 4. Favourable comparison is demonstrated against the state-of-the-art DeblurGAN and DAGAN algorithms in terms of PSNR and SSIM scores. The result was further confirmed by an image quality rating experiment performed by four senior neuroradiologists. CONCLUSIONS: The proposed method may become a valuable tool for scan time reduction in brain MRI. In continuation of this research, the validation should be extended to larger datasets acquired for different imaging protocols, and considering several MRI machines produced by different vendors.

Dataset Growth in Medical Image Analysis Research.

Medical image analysis research requires medical image datasets. Nevertheless, due to various impediments, researchers have been described as "data starved". We hypothesize that implicit evolving community standards require researchers to use ever-growing datasets. In Phase I of this research, we scanned the MICCAI (Medical Image Computing and Computer-Assisted Intervention) conference proceedings from 2011 to 2018. We identified 907 papers involving human MRI, CT or fMRI datasets and extracted their sizes. The median dataset size had grown by 3-10 times from 2011 to 2018, depending on imaging modality. Statistical analysis revealed exponential growth of the geometric mean dataset size with an annual growth of 21% for MRI, 24% for CT and 31% for fMRI. Thereupon, we had issued a forecast for dataset sizes in MICCAI 2019 well before the conference. In Phase II of this research, we examined the MICCAI 2019 proceedings and analyzed 308 relevant papers. The MICCAI 2019 statistics compare well with the forecast. The revised annual growth rates of the geometric mean dataset size are 27% for MRI, 30% for CT and 32% for fMRI. We predict the respective dataset sizes in the MICCAI 2020 conference (that we have not yet analyzed) and the future MICCAI 2021 conference.

Motion magnification analysis of microscopy videos of biological cells.

It is well recognized that isolated cardiac muscle cells beat in a periodic manner. Recently, evidence indicates that other, non-muscle cells, also perform periodic motions that are either imperceptible under conventional lab microscope lens or practically not easily amenable for analysis of oscillation amplitude, frequency, phase of movement and its direction. Here, we create a real-time video analysis tool to visually magnify and explore sub-micron rhythmic movements performed by biological cells and the induced movements in their surroundings. Using this tool, we suggest that fibroblast cells perform small fluctuating movements with a dominant frequency that is dependent on their surrounding substrate and its stiffness.

Neural Segmentation of Seeding ROIs (sROIs) for Pre-Surgical Brain Tractography.

White matter tractography mapping is an important tool for neuro-surgical planning and navigation. It relies on the accurate manual delineation of anatomical seeding ROIs (sROIs) by neuroanatomy experts. Stringent pre-operative time-constraints and limited availability of experts suggest that automation tools are strongly needed for the task. In this article, we propose and compare several multi-modal fully convolutional network architectures for segmentation of sROIs. Inspired by their manual segmentation practice, anatomical information from T1w maps is fused by the network with directionally encoded color (DEC) maps to compute the segmentation. Qualitative and quantitative validation was performed on image data from 75 real tumor resection candidates for the sROIs of the motor tract, the arcuate fasciculus, and optic radiation. Favorable comparison was also obtained with state-of-the-art methods for the tumor dataset as well as the ISMRM 2017 traCED challenge dataset. The proposed networks showed promising results, indicating they may significantly improve the efficiency of pre-surgical tractography mapping, without compromising its quality.

Classification of contrast-enhanced spectral mammography (CESM) images.

PURPOSE: Contrast-enhanced spectral mammography (CESM) is a recently developed breast imaging technique. CESM relies on dual-energy acquisition following contrast agent injection to improve mammography sensitivity. CESM is comparable to contrast-enhanced MRI in terms of sensitivity, at a fraction of the cost. However, since lesion variability is large, even with the improved visibility provided by CESM, differentiation between benign and malignant enhancement is not accurate and a biopsy is usually performed for final assessment. Breast biopsies can be stressful to the patient and are expensive to healthcare systems. Moreover, as the biopsies results are most of the time benign, a specificity improvement in the radiologist diagnosis is required. This work presents a deep learning-based decision support system, which aims at improving the specificity of breast cancer diagnosis by CESM without affecting sensitivity. METHODS: We compare two analysis approaches, fine-tuning a pretrained network and fully training a convolutional neural network, for classification of CESM breast mass as benign or malignant. Breast Imaging Reporting and Data Systems (BIRADS) is a radiological lexicon, used with breast images, to categorize lesions. We improve each classification network by incorporating BIRADS textual features as an additional input to the network. We evaluate two ways of BIRADS fusion as network input: feature fusion and decision fusion. This leads to multimodal network architectures. At classification, we also exploit information from apparently normal breast tissue in the CESM of the considered patient, leading to a patient-specific classification. RESULTS: We evaluate performance using fivefold cross-validation, on 129 randomly selected breast lesions annotated by an experienced radiologist. Each annotation includes a contour of the mass in the image, biopsy-proven label of benign or malignant lesion and BIRADS descriptors. At 100% sensitivity, specificity of 66% was achieved using a multimodal network, which combines inputs at feature level and patient-specific classification. CONCLUSIONS: The presented multimodal network may significantly reduce benign biopsies, without compromising sensitivity.

3-D Neural denoising for low-dose Coronary CT Angiography (CCTA).

CCTA has become an important tool for coronary arteries assessment in low and medium risk patients. However, it exposes the patient to significant radiation doses, resulting from high image quality requirements and acquisitions at multiple cardiac phases. For widespread use of CCTA for coronary assessment, significant reduction of radiation exposure with minimal image quality loss is still needed. A neural denoising scheme, relying on a fully convolutional neural network (FCNN) architecture, is developed and applied to noisy CCTA. In contrast to previously published methods, the proposed FCNN is trained directly on 3-D CT data patches (blocks), implementing 3-D convolutions. Considering that anatomy is inherently tridimensional, the proposed 3-D approach may better capture and enforce inter-slice continuity of tiny structures. While training is performed on individual blocks, whole input scans can be fed and denoised in one piece, thus leveraging the fully convolutional architecture to maximize processing speed. The proposed method is compared to state-of-the-art denoising algorithms on a dataset of 45 CCTA scans. Low-dose scans are simulated by synthetic Poisson noise applied to the sinogram corresponding to a 90% reduction in radiation dose. The average feature similarity score (0.864) and the peak signal-to-noise ratio (41.47) obtained for the proposed algorithm outperformed the compared methods while requiring significantly shorter processing time. A set of 2-D FCNNs, structurally similar to the proposed 3-D network, are also implemented to demonstrate contribution of the additional dimension to the improved denoising. For further validation of the method coronary reconstruction using the Intellispace cardiac tool (Philips, Holland) is performed both on a real noisy CCTA scan and on the denoised scan using the proposed method. It is shown that the cardiac tool succeeds in reconstructing more coronaries using the scan denoised by the proposed method. The obtained results suggest the proposed method provides an efficient and powerful approach to low-dose CCTA denoising.

Fully automatic detection of renal cysts in abdominal CT scans.

PURPOSE: Simple renal cysts are a common benign finding in abdominal CT scans. However, since they may evolve in time, simple cysts need to be reported. With an ever-growing number of slices per CT scan, cysts are easily overlooked by the overloaded radiologist. In this paper, we address the detection of simple renal cysts as an incidental finding in a real clinical setting. METHODS: We propose a fully automatic framework for renal cyst detection, supported by a robust segmentation of the kidneys performed by a fully convolutional neural network. A combined 3D distance map of the kidneys and surrounding fluids provides initial candidates for cysts. Eventually, a second convolutional neural network classifies the candidates as cysts or non-cyst objects. RESULTS: Performance was evaluated on 52 abdominal CT scans selected at random in a real radiological workflow and containing over 70 cysts annotated by an experienced radiologist. Setting the minimal cyst diameter to 10 mm, the algorithm detected 59/70 cysts (true-positive rate = 84.3%) while producing an average of 1.6 false-positive per case. CONCLUSIONS: The obtained results suggest the proposed framework is a promising approach for the automatic detection of renal cysts as incidental findings of abdominal CT scans.

Patient-specific image denoising for ultra-low-dose CT-guided lung biopsies.

PURPOSE: Low-dose CT screening of the lungs is becoming a reality, triggering many more CT-guided lung biopsies. During these biopsies, the patient is submitted to repeated guiding scans with substantial cumulated radiation dose. Extension of the dose reduction to the biopsy procedure is therefore necessary. We propose an image denoising algorithm that specifically addresses the setup of CT-guided lung biopsies. It minimizes radiation exposure while keeping the image quality appropriate for navigation to the target lesion. METHODS: A database of high-SNR CT patches is used to filter noisy pixels in a non-local means framework, while explicitly enforcing local spatial consistency in order to preserve fine image details and structures. The patch database may be created from a multi-patient set of high-SNR lung scans. Alternatively, the first scan, acquired at high-SNR right before the needle insertion, can provide a convenient patient-specific patch database. RESULTS: The proposed algorithm is compared to state-of-the-art denoising algorithms for a dataset of 43 real CT-guided biopsy scans. Ultra-low-dose scans were simulated by synthetic noise addition to the sinogram, equivalent to a 96% reduction in radiation dose. The feature similarity score for the proposed algorithm outperformed the compared methods for all the scans in the dataset. The benefit of the patient-specific patch database over the multi-patient one is demonstrated in terms of recovered contrast for a tiny porcine lung nodule, following denoising with both approaches. CONCLUSIONS: The proposed method provides a promising approach to the denoising of ultra-low-dose CT-guided biopsy images.

Optimization of two-compartment-exchange-model analysis for dynamic contrast-enhanced mri incorporating bolus arrival time.

PURPOSE: To optimize the analysis of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) under the two-compartment-exchange-model (2CXM) and to incorporate voxelwise bolus-arrival-time (BAT). MATERIALS AND METHODS: The accuracy of the pharmacokinetic (PK) parameters, extracted from 3T DCE-MRI using 2CXM, was tested under several conditions: eight algorithms for data estimation; correction for BAT; using model selection; different temporal resolution and scan duration. Comparisons were performed on simulated data. The best algorithm was applied to seven patients with brain tumors or following stroke. The extracted perfusion parameters were compared to those of dynamic susceptibility contrast MRI (DSC-MRI). RESULTS: ACoPeD (AIF-corrected-perfusion-DCE-MRI), an analysis using a 2nd derivative regularized-spline and incorporating BAT, achieved the most accurate estimation in simulated data, mean-relative-error: Fp , F, vp , ve : 24.8%, 41.7%, 26.4%, 27.2% vs. 76.5%, 190.8%, 78.8%, 82.39% of the direct four parameters estimation (one-sided two-sample t-test, P < 0.001). Correction for BAT increased the estimation accuracy of the PK parameters by more than 30% and provided a supertemporal resolution estimation of the BAT (higher than the acquired resolution, mean-absolute-error 0.2 sec). High temporal resolution (∼2 sec) is required to avoid biased estimation of PK parameters, and long scan duration (∼20 min) is important for reliable permeability but not for perfusion estimations, mean-error-reduction: E: ∼12%, ve : ∼6%. Using ACoPeD, PK values from normal-appearing white matter, gray matter, and lesion were extracted from patients. Preliminary results showed significant voxelwise correlations to DSC-MRI, between flow values in a patient following stroke (r = 0.49, P < 0.001), and blood volume in a patient with a brain tumor (r = 0.62, P < 0.001). CONCLUSION: This study proposes an optimized analysis method, ACoPeD, for tissue perfusion and permeability estimation using DCE-MRI, to be used in clinical settings. LEVEL OF EVIDENCE: 1 J. Magn. Reson. Imaging 2017;45:237-249.

Computerized assessment of body image in anorexia nervosa and bulimia nervosa: comparison with standardized body image assessment tool.

Body image disturbances are a prominent feature of eating disorders (EDs). Our aim was to test and evaluate a computerized assessment of body image (CABI), to compare the body image disturbances in different ED types, and to assess the factors affecting body image. The body image of 22 individuals undergoing inpatient treatment with restricting anorexia nervosa (AN-R), 22 with binge/purge AN (AN-B/P), 20 with bulimia nervosa (BN), and 41 healthy controls was assessed using the Contour Drawing Rating Scale (CDRS), the CABI, which simulated the participants' self-image in different levels of weight changes, and the Eating Disorder Inventory-2-Body Dissatisfaction (EDI-2-BD) scale. Severity of depression and anxiety was also assessed. Significant differences were found among the three scales assessing body image, although most of their dimensions differentiated between patients with EDs and controls. Our findings support the use of the CABI in the comparison of body image disturbances in patients with EDs vs. CONTROLS: Moreover, the use of different assessment tools allows for a better understanding of the differences in body image disturbances in different ED types.

A novel technique for infant length measurement based on stereoscopic vision.

BACKGROUND: Monitoring infant growth is essential for evaluation of development and is an important indicator of health and illness. Length is an essential indicator of infant growth, however, length measurement methods suffer from limitations which restrict their use. OBJECTIVE: To improve infant length measurement by development of a novel, accurate, precise and practical measurement technique. METHODS: A new system based on stereoscopic vision was developed. The system is comprised of two digital still cameras combined with software that calculates the infant's length from two simultaneously taken pictures. Length measurements of 54 healthy newborns were performed using a standard length board and the stereoscopic system. The two measurement methods were compared. RESULTS: Mean infant length was 473.1 (SD=29.1) mm versus 473.3 (SD=29.3) mm by length board and by the stereoscopic system, respectively. The mean difference between measurements was 0.2 (SD=2.5) mm and the mean of the absolute values of differences was 2.0 (SD=1.4) mm. Bland-Altman analysis showed good agreement between the two measurement methods. Precision of the new technique was demonstrated by a technical error of measurement of 2.57 mm. CONCLUSIONS: The stereoscopic system is accurate, reliable, easy to use, and involves less handling and discomfort to the newborns. It has the potential to measure premature infants or sick neonates through incubators.

On symmetry, perspectivity, and level-set-based segmentation.

We introduce a novel variational method for the extraction of objects with either bilateral or rotational symmetry in the presence of perspective distortion. Information on the symmetry axis of the object and the distorting transformation is obtained as a by--product of the segmentation process. The key idea is the use of a flip or a rotation of the image to segment as if it were another view of the object. We call this generated image the symmetrical counterpart image. We show that the symmetrical counterpart image and the source image are related by planar projective homography. This homography is determined by the unknown planar projective transformation that distorts the object symmetry. The proposed segmentation method uses a level-set-based curve evolution technique. The extraction of the object boundaries is based on the symmetry constraint and the image data. The symmetrical counterpart of the evolving level-set function provides a dynamic shape prior. It supports the segmentation by resolving possible ambiguities due to noise, clutter, occlusions, and assimilation with the background. The homography that aligns the symmetrical counterpart to the source level-set is recovered via a registration process carried out concurrently with the segmentation. Promising segmentation results of various images of approximately symmetrical objects are shown.

Atlas-based indexing of brain sections via 2-D to 3-D image registration.

A 2-D to 3-D nonlinear intensity-based registration method is proposed in which the alignment of histological brain sections with a volumetric brain atlas is performed. First, sparsely cut brain sections were linearly matched with an oblique slice automatically extracted from the atlas. Second, a planar-to-curved surface alignment was employed in order to match each section with its corresponding image overlaid on a curved-surface within the atlas. For the latter, a PDE-based registration technique was developed that is driven by a local normalized-mutual-information similarity measure. We demonstrate the method and evaluate its performance with simulated and real data experiments. An atlas-guided segmentation of mouse brains' hippocampal complex, retrieved from the Mouse Brain Library (MBL) database, is demonstrated with the proposed algorithm.

Deblurring of color images corrupted by impulsive noise.

We consider the problem of restoring a multichannel image corrupted by blur and impulsive noise (e.g., salt-and-pepper noise). Using the variational framework, we consider the L1 fidelity term and several possible regularizers. In particular, we use generalizations of the Mumford-Shah (MS) functional to color images and gamma-convergence approximations to unify deblurring and denoising. Experimental comparisons show that the MS stabilizer yields better results with respect to Beltrami and total variation regularizers. Color edge detection is a beneficial by-product of our methods.

Semi-blind image restoration via Mumford-Shah regularization.

Image restoration and segmentation are both classical problems, that are known to be difficult and have attracted major research efforts. This paper shows that the two problems are tightly coupled and can be successfully solved together. Mutual support of image restoration and segmentation processes within a joint variational framework is theoretically motivated, and validated by successful experimental results. The proposed variational method integrates semi-blind image deconvolution (parametric blur-kernel), and Mumford-Shah segmentation. The functional is formulated using the T-convergence approximation and is iteratively optimized via the alternate minimization method. While the major novelty of this work is in the unified treatment of the semi-blind restoration and segmentation problems, the important special case of known blur is also considered and promising results are obtained.

The point spread function of spiral CT.

Computed tomography (CT) scanners are usually described by their in-plane resolution and slice-sensitivity profile (SSP). Other imaging systems are characterized by their point spread function (PSF). The PSF is an excellent basis for the analysis, design and enhancement of imaging systems. The 3D PSF of CT systems has rarely been considered, and has usually been approximated by a 3D Gaussian. We present mathematical analysis of the PSF of single-slice and multi-slice fan-beam and cone-beam CT, for major reconstruction algorithms. We show that the PSF has a complicated, non-separable 3D shape. It is anisotropic in the xy plane and twisted in the z direction. Furthermore, the PSF is space variant in all three axes. In particular, it rotates as the input impulse function moves in the z direction. The PSF may also have effective discontinuities that can lead to streaking artefacts. Indirect measurements of the PSF can be misleading. We support the theoretical results by direct experimental measurements of the PSF.