Standard Post-processing Methods Research Articles

Deep PACO: combining statistical models with deep learning for exoplanet detection and characterization in direct imaging at high contrast

ABSTRACT Direct imaging is an active research topic in astronomy for the detection and the characterization of young substellar objects. The very high contrast between the host star and its companions makes the observations particularly challenging. In this context, post-processing methods combining several images recorded with the pupil tracking mode of telescope are needed. In previous works, we have presented a data-driven algorithm, PACO, capturing locally the spatial correlations of the data with a multivariate Gaussian model. PACO delivers better detection sensitivity and confidence than the standard post-processing methods of the field. However, there is room for improvement due to the approximate fidelity of the PACO statistical model to the time evolving observations. In this paper, we propose to combine the statistical model of PACO with supervised deep learning. The data are first pre-processed with the PACO framework to improve the stationarity and the contrast. A convolutional neural network (CNN) is then trained in a supervised fashion to detect the residual signature of synthetic sources. Finally, the trained network delivers a detection map. The photometry of detected sources is estimated by a second CNN. We apply the proposed approach to several data sets from the VLT/SPHERE instrument. Our results show that its detection stage performs significantly better than baseline methods (cADI and PCA), and leads to a contrast improvement up to half a magnitude compared to PACO. The characterization stage of the proposed method performs on average on par with or better than the comparative algorithms (PCA and PACO) for angular separation above 0.5 arcsec.

Detecting schizophrenia with 3D structural brain MRI using deep learning

Schizophrenia is a chronic neuropsychiatric disorder that causes distinct structural alterations within the brain. We hypothesize that deep learning applied to a structural neuroimaging dataset could detect disease-related alteration and improve classification and diagnostic accuracy. We tested this hypothesis using a single, widely available, and conventional T1-weighted MRI scan, from which we extracted the 3D whole-brain structure using standard post-processing methods. A deep learning model was then developed, optimized, and evaluated on three open datasets with T1-weighted MRI scans of patients with schizophrenia. Our proposed model outperformed the benchmark model, which was also trained with structural MR images using a 3D CNN architecture. Our model is capable of almost perfectly (area under the ROC curve = 0.987) distinguishing schizophrenia patients from healthy controls on unseen structural MRI scans. Regional analysis localized subcortical regions and ventricles as the most predictive brain regions. Subcortical structures serve a pivotal role in cognitive, affective, and social functions in humans, and structural abnormalities of these regions have been associated with schizophrenia. Our finding corroborates that schizophrenia is associated with widespread alterations in subcortical brain structure and the subcortical structural information provides prominent features in diagnostic classification. Together, these results further demonstrate the potential of deep learning to improve schizophrenia diagnosis and identify its structural neuroimaging signatures from a single, standard T1-weighted brain MRI.

REXPACO: An algorithm for high contrast reconstruction of the circumstellar environment by angular differential imaging

Context.Direct imaging is a method of choice for probing the close environment of young stars. Even with the coupling of adaptive optics and coronagraphy, the direct detection of off-axis sources such as circumstellar disks and exoplanets remains challenging due to the required high contrast and small angular resolution. Angular differential imaging (ADI) is an observational technique that introduces an angular diversity to help disentangle the signal of off-axis sources from the residual signal of the star in a post-processing step.Aims.While various detection algorithms have been proposed in the last decade to process ADI sequences and reach high contrast for the detection of point-like sources, very few methods are available to reconstruct meaningful images of extended features such as circumstellar disks. The purpose of this paper is to describe a new post-processing algorithm dedicated to the reconstruction of the spatial distribution of light (total intensity) received from off-axis sources, in particular from circumstellar disks.Methods.Built on the recentPACOalgorithm dedicated to the detection of point-like sources, the proposed method is based on the local learning of patch covariances capturing the spatial fluctuations of the stellar leakages. From this statistical modeling, we develop a regularized image reconstruction algorithm (REXPACO) following an inverse problems approach based on a forward image formation model of the off-axis sources in the ADI sequences.Results.Injections of fake circumstellar disks in ADI sequences from the VLT/SPHERE-IRDIS instrument show that both the morphology and the photometry of the disks are better preserved byREXPACOcompared to standard post-processing methods such as cADI. In particular, the modeling of the spatial covariances proves useful in reducing typical ADI artifacts and in better disentangling the signal of these sources from the residual stellar contamination. The application to stars hosting circumstellar disks with various morphologies confirms the ability ofREXPACOto produce images of the light distribution with reduced artifacts. Finally, we show howREXPACOcan be combined withPACOto disentangle the signal of circumstellar disks from the signal of candidate point-like sources.Conclusions.REXPACOis a novel post-processing algorithm for reconstructing images of the circumstellar environment from high contrast ADI sequences. It produces numerically deblurred images and exploits the spatial covariances of the stellar leakages and of the noise to efficiently eliminate this nuisance term. The processing is fully unsupervised, all tuning parameters being directly estimated from the data themselves.

Determination of acrolein-associated T1 and T2 relaxation times and noninvasive detection using nuclear magnetic resonance and magnetic resonance spectroscopy.

An estimated 3.3 million people are living with a traumatic brain injury (TBI)-associated morbidity. Currently, only invasive and sacrificial methods exist to study neurochemical alterations following TBI. Nuclear magnetic resonance methods-magnetic resonance imaging (MRI) and spectroscopy (MRS)-are powerful tools which may be used non-invasively to diagnose a range of medical issues. These methods can be utilized to explore brain functionality, connectivity, and biochemistry. Unfortunately, many of the commonly studied brain metabolites (e.g., N-acetyl-aspartate, choline, creatine) remain relatively stable following mild to moderate TBI and may not be suitable for longitudinal assessment of injury severity and location. Therefore, a critical need exists to investigate alternative biomarkers of TBI, such as acrolein. Acrolein is a byproduct of lipid peroxidation and accumulates following damage to neuronal tissue. Acrolein has been shown to increase in post-mortem rat brain tissue following TBI. However, no methods exist to noninvasively quantify acrolein in vivo. Currently, we have characterized the T1 and T2 of acrolein via NMR saturation recovery and Carr-Purcell-Meiboom-Gill experiments, accordingly, to maximize the signal-to-noise ratio of acrolein obtained with MRS. Additionally, we have quantified acrolein in water and whole-brain phantom using PRESS MRS and standard post-processing methods. With this potential novel biomarker for assessing TBI, we can investigate methods for predicting acute and chronic neurological dysfunction in humans and animal models. By quantifying and localizing acrolein with MRS, and investigating neurological outcomes associated with in vivo measures, patient-specific interventions could be developed to decrease TBI-associated morbidity and improve quality of life.

Exploring the origins of EEG motion artefacts during simultaneous fMRI acquisition: Implications for motion artefact correction

Motion artefacts (MAs) are induced within EEG data collected simultaneously with fMRI when the subject's head rotates relative to the magnetic field. The effects of these artefacts have generally been ameliorated by removing periods of data during which large artefact voltages appear in the EEG traces. However, even when combined with other standard post-processing methods, this strategy does not remove smaller MAs which can dominate the neuronal signals of interest. A number of methods are therefore being developed to characterise the MA by measuring reference signals and then using these in artefact correction. These methods generally assume that the head and EEG cap, plus any attached sensors, form a rigid body which can be characterised by a standard set of six motion parameters. Here we investigate the motion of the head/EEG cap system to provide a better understanding of MAs. We focus on the reference layer artefact subtraction (RLAS) approach, as this allows measurement of a separate reference signal for each electrode that is being used to measure brain activity.Through a series of experiments on phantoms and subjects, we find that movement of the EEG cap relative to the phantom and skin on the forehead is relatively small and that this non-rigid body movement does not appear to cause considerable discrepancy in artefacts between the scalp and reference signals. However, differences in the amplitude of these signals is observed which may be due to differences in geometry of the system from which the reference signals are measured compared with the brain signals. In addition, we find that there is non-rigid body movement of the skull and skin which produces an additional MA component for a head shake, which is not present for a head nod. This results in a large discrepancy in the amplitude and temporal profile of the MA measured on the scalp and reference layer, reducing the efficacy of MA correction based on the reference signals.Together our data suggest that the efficacy of the correction of MA using any reference-based system is likely to differ for different types of head movement with head shake being the hardest to correct. This provides new information to inform the development of hardware and post-processing methods for removing MAs from EEG data acquired simultaneously with fMRI data.

Design Process Control for Improved Surface Finish of Metal Additive Manufactured Parts of Complex Build Geometry

Metal additive manufacturing (AM) is increasingly used to create complex 3D components at near net shape. However, the surface finish (SF) of the metal AM part is uneven, with surface roughness being variable over the facets of the design. Standard post-processing methods such as grinding and linishing often meet with major challenges in finishing parts of complex shape. This paper reports on research that demonstrated that mass finishing (MF) processes are able to deliver high-quality surface finishes (Ra and Sa) on AM-generated parts of a relatively complex geometry (both internal features and external facets) under select conditions. Four processes were studied in this work: stream finishing, high-energy (HE) centrifuge, drag finishing and disc finishing. Optimisation of the drag finishing process was then studied using a structured design of experiments (DOE). The effects of a range of finishing parameters were evaluated and optimal parameters and conditions were determined. The study established that the proposed method can be successfully applied in drag finishing to optimise the surface roughness in an industrial application and that it is an economical way of obtaining the maximum amount of information in a short period of time with a small number of tests. The study has also provided an important step in helping understand the requirements of MF to deliver AM-generated parts to a target quality finish and cycle time.

How much are clinical fMRI reports influenced by standard postprocessing methods? An investigation of normalization and region of interest effects in the medial temporal lobe

Recent evidence has indicated that standard postprocessing methods such as template-based region of interest (ROI) definition and normalization of individual brains to a standard template may influence final outcome of functional magnetic resonance imaging investigations. Here, we provide the first comprehensive investigation into whether ROI definition and normalization may also change the clinical interpretation of patient data. A series of medial temporal lobe epilepsy patients were investigated with a clinical memory paradigm and individually delineated as well as template-based ROIs. Different metrics for activation quantification were applied. Results show that the application of template-based ROIs can significantly change the clinical interpretation of individual patient data. This relates to sensitivity for brain activation and hemispheric dominance. We conclude that individual ROIs should be defined on nontransformed functional data and that use of more than one metric for activation quantification is beneficial.

Can a Multimodel SuperEnsemble technique be used for precipitation forecasts?

Abstract. The Multimodel SuperEnsemble technique is a postprocessing method for the estimation of weather forecast parameters reducing direct model output errors. It differs from other ensemble analysis techniques by the use of an adequate weighting of the input forecast models in order to obtain a combined estimation of meteorological parameters. Weights are calculated by least-square minimization of the differences between the model and the observed field during a so-called training period. Although it can be applied successfully on continuous parameters like temperature, relative humidity, wind speed and mean sea level pressure, the Multimodel SuperEnsemble also gives good results when applied on the precipitation, a parameter quite difficult to handle with standard post-processing methods. Here we present a methodology for the Multimodel precipitation forecasts with a careful ensemble dressing via the precipitation PDF estimation.

Multimodel SuperEnsemble technique for quantitative precipitation forecasts in Piemonte region

Abstract. The Multimodel SuperEnsemble technique is a powerful post-processing method for the estimation of weather forecast parameters reducing direct model output errors. It has been applied to real time NWP, TRMM-SSM/I based multi-analysis, Seasonal Climate Forecasts and Hurricane Forecasts. The novelty of this approach lies in the methodology, which differs from ensemble analysis techniques used elsewhere. Several model outputs are put together with adequate weights to obtain a combined estimation of meteorological parameters. Weights are calculated by least-square minimization of the difference between the model and the observed field during a so-called training period. Although it can be applied successfully on the continuous parameters like temperature, humidity, wind speed and mean sea level pressure, the Multimodel SuperEnsemble gives good results also when applied on the precipitation, a parameter quite difficult to handle with standard post-processing methods. Here we present our methodology for the Multimodel precipitation forecasts, involving a new accurate statistical method for bias correction and a wide spectrum of results over Piemonte very dense non-GTS weather station network.

Comparison of 180° and 360° acquisition for myocardial perfusion SPECT with compensation for attenuation, detector response, and scatter: Monte Carlo and mathematical observer results

The optimal projection data acquisition strategy for myocardial perfusion (MP) single photon emission computed tomography (SPECT) remains controversial. We compared MP SPECT using 180 degrees and 360 degrees projection data obtained with the same acquisition time, reconstructed either with filtered back projection (FBP) or the iterative ordered-subsets expectation maximization (OS-EM) algorithm with various combinations of attenuation, detector response, and scatter compensation using mathematical observers and a myocardial defect detection task. We used Monte Carlo-simulated projection data from a population of 3-dimensional nurbs-based cardiac-torso (NCAT) phantoms with ranges of variability in patient anatomy, organ uptake, defect location, defect size, and noise level based on clinical data. Projection data from 180 degrees and 360 degrees acquisitions were generated by assuming the same acquisition time. After iterative or FBP reconstruction, standard postprocessing methods were applied. For each acquisition and reconstruction method, we optimized the number of iterations and cut-off frequency of the Butterworth filter using the Channelized Hotelling Observer methodology. The optimum set of parameters was that which gave the maximum area under the curve. For both acquisition protocols, OS-EM with compensations provided better performance than FBP or OS-EM without compensation. For FBP, the optimized 180 degrees acquisition provided a statistically significant increase in AUC as compared with optimized 360 degrees acquisition. For OS-EM, the AUCs for 180 degrees were slightly larger than for 360 degrees acquisitions when comparing images reconstructed with the same compensations. However, the differences were smaller and not statistically significant. With optimized reconstruction and filtering parameters, 180 degrees acquisition provided a statistically significant improvement over 360 degrees acquisition for FBP reconstruction. However, for OS-EM the differences were small and not statistically significant.