In representation learning, Convolutional Sparse Coding (CSC) enables unsupervised learning of features by jointly optimising both an $\ell _{2}$ -norm fidelity term and a sparsity enforcing penalty. This work investigates using a regularisation term derived from an assumed Cauchy prior for the coefficients of the feature maps of a CSC generative model. The sparsity penalty term resulting from this prior is solved via its proximal operator, which is then applied iteratively, element-wise, on the coefficients of the feature maps to optimise the CSC cost function. The performance of the proposed Iterative Cauchy Thresholding (ICT) algorithm in reconstructing natural images is compared against algorithms based on minimising standard penalty functions via soft and hard thresholding as well as against the Iterative Log-Thresholding (ILT) method. ICT outperforms the Iterative Hard Thresholding (IHT), Iterative Soft Thresholding (IST), and ILT algorithms in most of our reconstruction experiments across various datasets, with an average Peak Signal to Noise Ratio (PSNR) of up to 11.30 dB, 7.04 dB, and 7.74 dB over IST, IHT, and ILT respectively. The source code for the implementation of the proposed approach is publicly available at https://github.com/p-mayo/cauchycsc

Purpose Many methods are available to segment structural magnetic resonance (MR) images of the brain into different tissue types. These have generally been developed for research purposes but there is some clinical use in the diagnosis of neurodegenerative diseases such as dementia. The potential exists for computed tomography (CT) segmentation to be used in place of MRI segmentation, but this will require a method to verify the accuracy of CT processing, particularly if algorithms developed for MR are used, as MR has notably greater tissue contrast. Methods To investigate these issues we have created a three‐dimensional (3D) printed brain with realistic Hounsfield unit (HU) values based on tissue maps segmented directly from an individual T1 MRI scan of a normal subject. Several T1 MRI scans of normal subjects from the ADNI database were segmented using SPM12 and used to create stereolithography files of different tissues for 3D printing. The attenuation properties of several material blends were investigated, and three suitable formulations were used to print an object expected to have realistic geometry and attenuation properties. A skull was simulated by coating the object with plaster of Paris impregnated bandages. Using two CT scanners, the realism of the phantom was assessed by the measurement of HU values, SPM12 segmentation and comparison with the source data used to create the phantom. Results Realistic relative HU values were measured although a subtraction of 60 was required to obtain equivalence with the expected values (gray matter 32.9–35.8 phantom, 29.9–34.2 literature). Segmentation of images acquired at different kVps/mAs showed excellent agreement with the source data (Dice Similarity Coefficient 0.79 for gray matter). The performance of two scanners with two segmentation methods was compared, with the scanners found to have similar performance and with one segmentation method clearly superior to the other. Conclusion The ability to use 3D printing to create a realistic (in terms of geometry and attenuation properties) head phantom has been demonstrated and used in an initial assessment of CT segmentation accuracy using freely available software developed for MRI.

In the machine learning era, sparsity continues to attract significant interest due to the benefits it provides to learning models. Algorithms aiming to optimise the \(\ell_0\)- and \(\ell_1\)-norm are the common choices to achieve sparsity. In this work, an alternative algorithm is proposed, which is derived based on the assumption of a Cauchy distribution characterising the coefficients in sparse domains. The Cauchy distribution is known to be able to capture heavy-tails in the data, which are linked to sparse processes. We begin by deriving the Cauchy proximal operator and subsequently propose an algorithm for optimising a cost function which includes a Cauchy penalty term. We have coined our contribution as Iterative Cauchy Thresholding (ICT). Results indicate that sparser solutions can be achieved using ICT in conjunction with a fixed over-complete discrete cosine transform dictionary under a sparse coding methodology.

AbstractBackgroundPrognosis for Alzheimer’s disease is difficult, with rates of disease progression varying widely. Despite the extensive use of clinical dementia severity assessment scales to determine dementia diagnosis and to monitor progression, there is no consensus on which imaging factors may accurately predict future decline trajectories on these measures. Determination of baseline imaging patterns that are related to slower, or faster, decline rates could be used to identify those at highest risk of worse outcomes and inform care planning.MethodDecline trajectories were estimated from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset which contains clinical, imaging, and biological assessment records for 1,736 patients followed‐up over 8 years at 3, 6, or 12 month intervals. Clusters of similar decline trajectories were identified for patients with relevant baseline fluorodeoxyglucose positron emission tomography (FDG‐PET) imaging data (N=530) using mini‐mental state examination (MMSE), clinical dementia rating sum of boxes (CDR‐SB), Alzheimer’s disease assessment scale (ADAS‐13), and functional activities questionnaire (FAQ) assessment scores. Decline trajectories allocated to slow, intermediate, and fast clusters were further analysed to find significant associations with PET imaging features using statistical parametric mapping (SPM12).ResultWhen compared to cognitively normal non‐decliners, slow, intermediate and fast cluster groups showed similar topographic patterns of hypometabolism on FDG‐PET for all four clinical assessments (p<0.05 family‐wise error corrected). Slow progressors showed focal hippocampal hypometabolism, which extended into parietotemporal, inferior parietal and precuneal regions for intermediate progressors. Fast progressors showed more extensive hypometabolism in similar regions with additional superior frontal hypometabolism. Although locational deficits were similar on cognitive measures for decline groups, CDR‐SB showed the greatest hypometabolism extent for slow progressors, followed by FAQ, then ADAS‐13, with MMSE showing the least.ConclusionOur results suggest a single pattern of pathological progression of functional degeneration in Alzheimer’s disease, with slow, intermediate and fast progressors at different time points on the same decline trajectory. This progression begins with focal medial temporal regional abnormalities, extending to parietotemporal, inferior parietal and precuneus regions, before finally involving the superior frontal lobe. CDR‐SB was the most sensitive in detecting slow progressor abnormality, with MMSE being the least sensitive.

Diagnosis of brain diseases is considered one of the most challenging medical tasks to perform, even for medical experts who rely on high-resolution anatomical images to identify signs of abnormalities by visual inspection. However, new computational tools which assist to automate this diagnosis have the potential to significantly improve the speed and accuracy of this process. This work presents a model to aid in the task of classification of structural Magnetic Resonance Imaging scans. The classification is performed using a Support Vector Machine, whilst the features to analyze belong to a dictionary space. Such space was mainly built from a dictionary learning perspective, although a predefined one was also assessed. The results indicate that features learnt from the data of interest lead to improved classification performance. The proposed framework was tested on the ADNI dataset stage I.

Perfusion brain single photon emission computed tomography (SPECT) imaging is commonly used to aid dementia diagnosis; however the relationship between specific imaging patterns and longitudinal functional decline in unselected patient cohorts has not been fully explored. This study aimed to investigate whether perfusion SPECT imaging patterns could provide useful information to aid dementia prognosis. 117 individuals with cognitive complaints who underwent perfusion SPECT imaging due to diagnostic doubt were consented to the Brain Imaging in Dementia (BraIID) study, a two-year longitudinal follow up study aimed to improve dementia imaging reporting. The clinical dementia rating (CDR) scale was completed at baseline and 6 monthly time points to assess patient decline (mean follow up 435.8 days, s.d. 203.9 days). Using a decline outcome of 2 points on CDR Sum of Boxes or one global CDR category increase, patients were grouped into decliners (N=60) and non-decliners (N=57). Decliners were further divided into fast (reached outcome in <200 days, N=26), moderate (200-400 days, N=23) and slow (>400 days, N=11) decline groups. Images were pre-processed using statistical parametric mapping (SPM) and t-test comparisons were performed comparing decliner groups against normal controls (N=34), and all decliners against non-decliners. Fast decliners were also compared against slow/moderate decline groups. A p<0.05 FWE corrected threshold was used. The deficit patterns were markedly different for the three decliner subgroups when compared to controls, involving frontal and parietal lobes bilaterally in the fast decliners, parietal and superior frontal regions for moderate decliners, and a posterior pattern involving both occipital and parietal lobes for the slow decliners. Decliners showed significant perfusion deficits in bilateral angular gyrus, precuneus and bilateral superior frontal lobe regions when compared to non-decliners. When comparing decliner groups, perfusion was significantly reduced in the left orbitofrontal and superior frontal gyrus for the fast decliner group compared to the moderate and slow groups. Regional perfusion SPECT abnormality provides prognostic information that can aid clinician reporting, and quantify the likely time course of decline. Patients with frontal hypoperfusion are at higher risk when compared to other cognitively impaired individuals. Those with posterior only patterns may decline more slowly.

The Alderson striatal phantom is frequently used to assess 123I-FP-CIT (Ioflupane) image quality and to test semi-quantification software. However, its design is associated with a number of limitations, in particular: unrealistic image appearances and inflexibility. A new physical phantom approach is proposed on the basis of subresolution phantom technology. The design incorporates thin slabs of attenuating material generated through additive manufacturing, and paper sheets with radioactive ink patterns printed on their surface, created with a conventional inkjet printer. The paper sheets and attenuating slabs are interleaved before scanning. Use of thin layers ensures that they cannot be individually resolved on reconstructed images. An investigation was carried out to demonstrate the performance of such a phantom in producing simplified 123I-FP-CIT uptake patterns. Single photon emission computed tomography imaging was carried out on an assembled phantom designed to mimic a healthy patient. Striatal binding ratio results and linear striatal dimensions were calculated from the reconstructed data and compared with that of 22 clinical patients without evidence of Parkinsonian syndrome, determined from clinical follow-up. Striatal binding ratio results for the fully assembled phantom were: 3.1, 3.3, 2.9 and 2.6 for the right caudate, left caudate, right putamen and right caudate, respectively. All were within two SDs of results derived from a cohort of clinical patients. Medial–lateral and anterior–posterior dimensions of the simulated striata were also within the range of values seen in clinical data. This work provides the foundation for the generation of a range of more clinically realistic, physical phantoms.

Background In the UK, dementia affects 5% of the population aged over 65 years and 25% of those over 85 years. Frontotemporal dementia (FTD) represents one subtype and is thought to account for up to 16% of all degenerative dementias. Although the core of the diagnostic process in dementia rests firmly on clinical and cognitive assessments, a wide range of investigations are available to aid diagnosis. Regional cerebral blood flow (rCBF) single-photon emission computed tomography (SPECT) is an established clinical tool that uses an intravenously injected radiolabelled tracer to map blood flow in the brain. In FTD the characteristic pattern seen is hypoperfusion of the frontal and anterior temporal lobes. This pattern of blood flow is different to patterns seen in other subtypes of dementia and so can be used to differentiate FTD. DTA 24 Regional Cerebral Blood Flow Single Photon Emission Computed Tomography for detection of Frontotemporal...

PURPOSE To make an adaptable, head shaped radionuclide phantom to simulate molecular imaging of the brain using clinical acquisition and reconstruction protocols. This will allow the characterization and correction of scanner characteristics, and improve the accuracy of clinical image analysis, including the application of databases of normal subjects. METHODS A fused deposition modeling 3D printer was used to create a head shaped phantom made up of transaxial slabs, derived from a simulated MRI dataset. The attenuation of the printed polylactide (PLA), measured by means of the Hounsfield unit on CT scanning, was set to match that of the brain by adjusting the proportion of plastic filament and air (fill ratio). Transmission measurements were made to verify the attenuation of the printed slabs. The radionuclide distribution within the phantom was created by adding (99m)Tc pertechnetate to the ink cartridge of a paper printer and printing images of gray and white matter anatomy, segmented from the same MRI data. The complete subresolution sandwich phantom was assembled from alternate 3D printed slabs and radioactive paper sheets, and then imaged on a dual headed gamma camera to simulate an HMPAO SPECT scan. RESULTS Reconstructions of phantom scans successfully used automated ellipse fitting to apply attenuation correction. This removed the variability inherent in manual application of attenuation correction and registration inherent in existing cylindrical phantom designs. The resulting images were assessed visually and by count profiles and found to be similar to those from an existing elliptical PMMA phantom. CONCLUSIONS The authors have demonstrated the ability to create physically realistic HMPAO SPECT simulations using a novel head-shaped 3D printed subresolution sandwich method phantom. The phantom can be used to validate all neurological SPECT imaging applications. A simple modification of the phantom design to use thinner slabs would make it suitable for use in PET.

BACKGROUNDIn the UK, dementia affects 5% of the population aged over 65 years and 25% of those over 85 years. Frontotemporal dementia (FTD) represents one subtype and is thought to account for up to 16% of all degenerative dementias. Although the core of the diagnostic process in dementia rests firmly on clinical and cognitive assessments, a wide range of investigations are available to aid diagnosis.Regional cerebral blood flow (rCBF) single-photon emission computed tomography (SPECT) is an established clinical tool that uses an intravenously injected radiolabelled tracer to map blood flow in the brain. In FTD the characteristic pattern seen is hypoperfusion of the frontal and anterior temporal lobes. This pattern of blood flow is different to patterns seen in other subtypes of dementia and so can be used to differentiate FTD.It has been proposed that a diagnosis of FTD, (particularly early stage), should be made not only on the basis of clinical criteria but using a combination of other diagnostic findings, including rCBF SPECT. However, more extensive testing comes at a financial cost, and with a potential risk to patient safety and comfort.OBJECTIVESTo determine the diagnostic accuracy of rCBF SPECT for diagnosing FTD in populations with suspected dementia in secondary/tertiary healthcare settings and in the differential diagnosis of FTD from other dementia subtypes.SEARCH METHODSOur search strategy used two concepts: (a) the index test and (b) the condition of interest. We searched citation databases, including MEDLINE (Ovid SP), EMBASE (Ovid SP), BIOSIS (Ovid SP), Web of Science Core Collection (ISI Web of Science), PsycINFO (Ovid SP), CINAHL (EBSCOhost) and LILACS (Bireme), using structured search strategies appropriate for each database. In addition we searched specialised sources of diagnostic test accuracy studies and reviews including: MEDION (Universities of Maastricht and Leuven), DARE (Database of Abstracts of Reviews of Effects) and HTA (Health Technology Assessment) database.We requested a search of the Cochrane Register of Diagnostic Test Accuracy Studies and used the related articles feature in PubMed to search for additional studies. We tracked key studies in citation databases such as Science Citation Index and Scopus to ascertain any further relevant studies. We identified 'grey' literature, mainly in the form of conference abstracts, through the Web of Science Core Collection, including Conference Proceedings Citation Index and Embase. The most recent search for this review was run on the 1 June 2013.Following title and abstract screening of the search results, full-text papers were obtained for each potentially eligible study. These papers were then independently evaluated for inclusion or exclusion.SELECTION CRITERIAWe included both case-control and cohort (delayed verification of diagnosis) studies. Where studies used a case-control design we included all participants who had a clinical diagnosis of FTD or other dementia subtype using standard clinical diagnostic criteria. For cohort studies, we included studies where all participants with suspected dementia were administered rCBF SPECT at baseline. We excluded studies of participants from selected populations (e.g. post-stroke) and studies of participants with a secondary cause of cognitive impairment.DATA COLLECTION AND ANALYSISTwo review authors extracted information on study characteristics and data for the assessment of methodological quality and the investigation of heterogeneity. We assessed the methodological quality of each study using the QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies) tool. We produced a narrative summary describing numbers of studies that were found to have high/low/unclear risk of bias as well as concerns regarding applicability. To produce 2 x 2 tables, we dichotomised the rCBF SPECT results (scan positive or negative for FTD) and cross-tabulated them against the results for the reference standard. These tables were then used to calculate the sensitivity and specificity of the index test. Meta-analysis was not performed due to the considerable between-study variation in clinical and methodological characteristics.MAIN RESULTSEleven studies (1117 participants) met our inclusion criteria. These consisted of six case-control studies, two retrospective cohort studies and three prospective cohort studies. Three studies used single-headed camera SPECT while the remaining eight used multiple-headed camera SPECT. Study design and methods varied widely. Overall, participant selection was not well described and the studies were judged as having either high or unclear risk of bias. Often the threshold used to define a positive SPECT result was not predefined and the results were reported with knowledge of the reference standard. Concerns regarding applicability of the studies to the review question were generally low across all three domains (participant selection, index test and reference standard).Sensitivities and specificities for differentiating FTD from non-FTD ranged from 0.73 to 1.00 and from 0.80 to 1.00, respectively, for the three multiple-headed camera studies. Sensitivities were lower for the two single-headed camera studies; one reported a sensitivity and specificity of 0.40 (95% confidence interval (CI) 0.05 to 0.85) and 0.95 (95% CI 0.90 to 0.98), respectively, and the other a sensitivity and specificity of 0.36 (95% CI 0.24 to 0.50) and 0.92 (95% CI 0.88 to 0.95), respectively.Eight of the 11 studies which used SPECT to differentiate FTD from Alzheimer's disease used multiple-headed camera SPECT. Of these studies, five used a case-control design and reported sensitivities of between 0.52 and 1.00, and specificities of between 0.41 and 0.86. The remaining three studies used a cohort design and reported sensitivities of between 0.73 and 1.00, and specificities of between 0.94 and 1.00. The three studies that used single-headed camera SPECT reported sensitivities of between 0.40 and 0.80, and specificities of between 0.61 and 0.97.AUTHORS' CONCLUSIONSAt present, we would not recommend the routine use of rCBF SPECT in clinical practice because there is insufficient evidence from the available literature to support this.Further research into the use of rCBF SPECT for differentiating FTD from other dementias is required. In particular, protocols should be standardised, study populations should be well described, the threshold for 'abnormal' scans predefined and clear details given on how scans are analysed. More prospective cohort studies that verify the presence or absence of FTD during a period of follow up should be undertaken.

In order to pool neuroimaging data from different imaging centres using different PET scanners it is usual to compare imaging performance using the Hoffman and Jaszczak phantoms. PET data submitted to the Alzheimer's Disease Neuroimaging Initiative (ADNI) is corrected using an algorithm derived from Hoffman phantom scans at participating centres. Recent publications have identified that superior corrections could be obtained via the use of head-shaped phantoms of different sizes which would more accurately reflect scatter and attenuation in human subjects. The sharing of patient and control data between scanners without increasing scanner-related variance is vital in order to maximize sensitivity to early Alzheimer's diseaseand to avoid artefacts that may lead to misdiagnosis. In order to simulate normal cerebral FDG uptake we created printout templates based on MRI scans segmented into grey and white matter. These templates were printed using an ink-18F mixture at an axial spacing of 2mm. The printed sheets were placed in an elliptical Perspex subresolution sandwich method (SSM) phantom and scanned using a GE 690 Discovery PET/CT scanner. Scans were compared to a database of FDG scans of normal subjects from ADNI using the statistical parametric mapping (SPM8) software and were also analyzed using Neurostat. We have also constructed head-shaped SSM phantoms using RepRap Mendel 3D printers (3DP) using both tissue and bone equivalent plastics. After variation of the printed grey to white matter ratio scans of the elliptical phantom have been obtained that appear realistic visually and when compared to human control scans using SPM8 and Neurostat. This initial work has demonstrated the feasibility of creating highly realistic FDG PET brain simulations by combining 3DP and SSM. Further work is required to refine the realism of scans using by incorporating recently constructed head-shaped phantoms with more realistic scatter and attenuation properties. By simulating both normal and AD FDG uptake this work will provide realistic ground truth for the optimization and validation of analysis methods utilizing image standardization and comparison with control databases. These methods have the potential to significantly augment visual interpretation of scan data.

Traditional interpretation of rCBF SPECT data is of a qualitative nature and is dependent on the observer's understanding of the normal distribution of the tracer. The use of a normal database in quantitative regional analysis facilitates the detection of functional abnormality in individual and group studies by accounting for inter-subject variability. The ability to simulate realistic images would allow various important areas related to the use of normal databases to be studied. These include the optimisation of the detection of abnormal blood flow and the portability of normal databases between gamma camera systems. To investigate this further we have constructed a hardware phantom and scanned various configurations of radioactive brain patterns and simulated skull configurations.Methods: A subresolution sandwich phantom was constructed with a simulated skull which was assembled using a high-resolution segmented MR scan printed with a (TcO4)-Tc-99m- mixture and scanned using a double-headed gamma camera with parallel-hole collimators. Various different grey-to-white matter (GM:WM) ratios and aluminium simulated skull configurations were used. A single difference measure between the phantom data and a control database mean image was used for optimisation. The realism of phantom data was assessed using statistical parametric mapping (SPM) and ROI analysis.Results: Optimisation was achieved with a range of WM:GM ratios from 1.9 to 2.4:1 with various simulated skull configurations.Conclusion: The ability to simulate realistic HMPAO SPECT scans has been demonstrated using a subresolution sandwich phantom. Further work, involving scanning the optimised phantom on different gamma camera systems and comparison with camera-specific normal databases should further refine the phantom configuration. (c) 2013 Elsevier Inc. All rights reserved.

Objective: Acute kidney injury (AKI) post-cardiac surgery is associated with mortality rates approaching 20%. The development of effective treatments is hindered by the poor homology between rodent models, the mainstay of research into AKI, and that which occurs in humans. This pilot study aims to characterise post-cardiopulmonary bypass (CPB) AKI in an animal model with potentially greater homology to cardiac surgery patients. Methods and results: Adult pigs, weighing 50-75 kg, underwent 2.5 h of CPB. Pigs undergoing saphenous vein grafting procedures served as controls. Pre-CPB measures of porcine renal function were within normal ranges for adult humans. The effect of CPB on renal function; a 25% reduction in Cr-51-EDTA clearance (p = 0.068), and a 33% reduction in creatinine clearance (p = 0.043), was similar to those reported in clinical studies. CPB resulted in tubular epithelial injury (median NAG/creatinine ratio 2.6 u mmol(-1) (interquartile range (IQR): 0.81-5.43) post-CPB vs 0.48 u mmol(-1) (IQR: 0.37-0.97) pre-CPB, p = 0.043) as well as glomerular and/or proximal tubular injury (median albumin/creatinine ratio 6.8 mg mmol(-1) (IQR: 5.45-13.06) post-CPB vs 1.10 mg mmol(-1) (IQR: 0.05-2.00) pre-CPB, p = 0.080). Tubular injury scores were significantly higher in kidneys post-CPB (median score 2.0 (IQR: 1.0-2.0) relative to vein graft controls (median score 1.0 (IQR 1.0-1.0), p = 0.019). AKI was associated with endothelial injury and activation, as demonstrated by reduced DBA (dolichos biflorus agglutinin) lectin and increased endothelin-1 and vascular cell adhesion molecule (VCAM) staining. Conclusions: The porcine model of post-CPB AKI shows significant homology to AKI in cardiac surgical patients. It links functional, urinary and histological measures of kidney injury and may offer novel insights into the mechanisms underlying post-CPB AKI. (C) 2009 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.

Wavelet-based cluster analysis (WCA) is a technique that can be used to separate fMRI and phMRI data into different clusters, where no model of the neural response is known a priori, based on the similarity of decomposed time courses at certain temporal scales. Here we extend this iteratively as an interactive step in a data-driven analysis. It works by removing voxels from further analysis by examining how they are clustered at temporal scales which may be different to that of any neural response under investigation. This is in contrast to existing techniques that suppress artefactual effects through excessive smoothing which may also suppress localised drug responses in phMRI. The method is demonstrated here on an auditory fMRI experiment. We conclude that it will be a useful step in the preprocessing and further analysis of phMRI data.

Introduction The analysis of pharmacological MRI (phMRI) traditionally depends upon the use of an appropriate input function, usually derived from blood plasma concentrations of the drug used in the experiment. There are a number of problems with this approach including the relationship between plasma and brain concentrations and the longer term effects of receptor activation. Because of this a number of data-driven approaches have been used where no model of the neural response is known a priori such as independent component analysis and wavelet cluster analysis [1]. Here we explore the use of a measure of signal complexity known as the Renyi entropy to discover voxels of interest in a data-driven manner using a dataset known to show reduced perfusion in the hippocampus.

Subject motion is a common problem in all forms of BOLD MRI and it is accepted that conventional motion correction does not remove all of this variance from the data. Hypothesis-driven approaches use the shifts from these corrections as unwanted-effects regressors, whereas data-driven analysis techniques often cannot, and so are more susceptible to this form of artefact. Modified temporal cluster analysis (MTCA) is a data-driven approach to the examination of fMRI and phMRI data. It is, however, susceptible to motion artefact, which we use here as an advantage and introduce the dual MTCA method as a means to discover when and where in the brain is being affected by motion which is uncorrected for, so that those slices may be suppressed before further data-driven analysis. This is particularly useful in experiments with long repetition times since only a subset of slices will require further correction. We apply this to placebo data from a phMRI study and conclude that it will be a useful step in the preprocessing and further analysis of such experiments.