Multi Scale Research Articles

Multi scale feature extraction network with machine learning algorithms for water body extraction from remote sensing images

ABSTRACT Water Body Extraction (WBE) is a challenging task in remote sensing, owing to the complexity of recognizing surface body objects with rich texture, spatial, spectral, temporal, and radiometric features. The use of spectral indices has shown to be successful in separating surface water from its surroundings at the cost of knowledge of appropriate threshold values. In the absence of knowledge on threshold values, extracting the water from remote sensing data is challenging, which is addressed by several Machine Learning (ML) and Deep Learning (DL) algorithms. However, the effectiveness of both ML and DL classifications is witnessed from visual features to semantic categories at the cost of distinct recognition between the water body and non-water body features. In this paper, a novel Multi Scale Feature Extraction Network (MSFEN) for extracting the pixel-level features from medium resolution remote sensing images is proposed and used traditional ML classifiers to extract the surface water bodies using pixel-level features extracted by MSFEN. The proposed framework is trained and tested on Linear Imaging Self Scanning Sensor – III (LISS-III) multispectral satellite images over major water reservoirs in Tamilnadu, Karnataka, Madhya Pradesh, and Odisha. Experimental results indicate that the proposed model MSFEN+SVM provides accurate extraction results by outperforming the existing state-of-the-art models (Fully Convolutional Network (FCN), Unet, SegNet, Multi Scale Convolutional Neural Network (MSCNN), Deepwater Map, Pyramid Scene Parsing Network (PSPNet), Improved PSPNet, Multi-scale Water Extraction Convolutional Neural Network (MWEN) and Multi-Scale Lake Water Extraction Network (MSLWENet) in terms of performance metrics considered.

CutPaste-Based Anomaly Detection Model using Multi Scale Feature Extraction in Time Series Streaming Data

CutPaste-Based Anomaly Detection Model using Multi Scale Feature Extraction in Time Series Streaming Data

Improving Medical Video Coding Using Multi Scale Quincunx Lattice: From Low Bitrate to High Quality

In recent years, the healthcare sector has seen an increase in the use of medical images and videos. However, storage and transmission of this huge volume of data remain a challenging task, requiring the use of compression techniques. In this paper, the authors propose an algorithm to improve the visual quality of compressed medical video for lower bitrate without modifying the content of information such as edges and textures, this is a unique way for doctors to store and share medical data over the internet. The algorithm has not yet been sufficiently explored in medical video coding. In this study, the performances of the quincunx wavelet transform (QWT) combined with the set partitioning in hierarchical trees (SPIHT) encoder are discussed. The QWTs were chosen due to their limited number of wavelets family and reduced dilatation factor. The high efficiency of the suggested algorithm is checked against the coding standard based on the discrete cosines transform (DCT) or discrete wavelet transform (DWT). The assessment of the quality of the decoded video is based on the use of the peak signal to noise ratio (PSNR), the mean structural similarity (MSSIM) and the visual information fidelity (VIF). The results prove that the QWT+SPIHT provide competing performance where the PSNR reached 33 dB value for lower bitrate (137.408 Kbps) against previous standards.

Automatic detection of three cell types in a microscope image based on deep learning.

With the continuous integration of deep learning and the technique of molecular biology, target detection models must accurately detect the position of each cell in the image and classify it correctly. We present a model for the multi-scale feature fusion of the existing human cell image dataset based on Gaussian mixedly clustering. First, a novel feature extraction network for extracting preliminary features at picture multi scales was presented, which was based on a residual neural network with Instance Normalization and a Mish activation function. Second, the presented model adopts the idea of feature fusion and introduced a new type of feature fusion network to integrate feature graphs on different scales. Furthermore, a Gaussian hybrid clustering algorithm was proposed to cluster the hyperparameters. Based on the experimental results, the average accuracy of the proposed model in the human cell image dataset exceeds 0.96, which improves by 11.9% compared with the existing target detection methods in the same field. Experiments show that the proposed model had been adapted to datasets with uneven sample distribution, providing new ideas for research on medical images.

A new approach to Quantitative Diagenesis Modeling (QDM) by integrated facies, well logs and seismic attributes from the cenomanian-early Turonian Sarvak Formation, Abadan Plain, SW Iran

Diagenesis in carbonate successions has a significant impact on reservoir characteristics. The Quantitively Diagenetic Modeling (QDM) approach, introduced in this study, was conducted based on the micro-to macro-scale data integration within the Cenomanian-early Turonian Sarvak carbonate rocks deposited in the Abadan plain. The adopted approach is based on an integrated analysis of petrographical data interpretation, well logs evaluation and seismic attribute analysis. The sedimentological and diagenetic studies were investigated based on the petrographic descriptions coupled with the analysis of well logs and seismic data. Seven facie associations corresponding to five depositional settings were identified. Among these, the main diagenetic processes including cementation and dissolution were distinguished as the primary diagenetic process affecting the Sarvak reservoir properties. The primary petrographical diagenetic process were classified based on the analysis of well log response and seismic amplitude refraction using the Artificial Neural Network (ANN) approach. In this context, the diagenetic zonations were concluded, reflecting geological features (facies and depositional setting), the reservoir characteristics (porosity, permeability), and seismic reflection (acoustic impedance and seismic reflectivity). The identified classes include mud-dominated facies with low microporosity/cemented micropores (DG-0), microporous cemented grainstone/packstone (DG-1), mud dominated facies with vugs and microfractures (DG-2), grain dominated facies with vuggy pores (DG-3), and dissolved rudist biostrome (DG-4). The proposed diagenetic classes introduced in this study propagated based on the TGS algorithm by using a new workflow to capture the variation of each diagenetic class throughout the studied fields. In this study, a new workflow was proposed based on the quantitative approaches at multiscale to reduce the uncertainty of property modeling within the studied carbonate reservoir considering the diagenetic overprints.The adopted approach implies that using the integrated data at multi scales (well to seismic) has a significant role in understanding the diagenetic pathways and its effects on the reservoir characteristics.

PERMEABILITY ANALYSIS OF FRACTURED-VUGGY CARBONATE RESERVOIRS BASED ON FRACTAL THEORY

Permeability analysis is an essential process in reservoir characterization. As one of the most development potential kind of reservoirs, the coexistence of fractures and vugs brings great challenges to permeability analysis of fractured-vuggy carbonate reservoirs due to its highly heterogeneity on multi scales. In this work, a novel permeability analysis model is developed for fractured-vuggy carbonate reservoirs based on fractal theory, which can predict the permeability of fractured-vuggy porous media within a reliable range. Besides, a Monte Carlo simulation has also been implemented to obtain feasible values of permeability based on experimental sample parameters. The validity of the proposed fractal model is verified by both the experimental data and numerical simulation results. Each parameter in the model has clear physical meaning. The results show that the permeability of a fractured-vuggy carbonate reservoir is related to the fracture fractal dimension, fracture fractal tortuosity dimension, the maximum vug diameter and the maximum fracture aperture.

An experimental study on the effects of introducing carbon nanotube on low velocity impact behavior of carbon/aramid fiber reinforced intra-ply hybrid composites

In this research, the addition effects of Multi Wall Carbon Nanotube (MWCNT) on low velocity impact behavior of Carbon/Aramid Fiber Reinforced (CAFRP) intra-ply hybrid composites were studied. In this regard, three types of pure, 0.1 wt.% and 0.5 wt.% introduced CAFRP intra-ply hybrid composites were fabricated by Vacuum Assisted Resin Infusion Molding (VARIM) method. The profile energy method was used to determine the penetration and perforation threshold. The main purpose of this research was to improve the impact behavior of CAFRP by addition of MWCNT to the matrix of these composites. The low velocity impact test was also carried out in 30, 50, 80 and 120 J energy values. The effect of addition MWCNTs in the matrix of composite was investigated by plotting force, absorbed energy and deflection vs. time and the key parameters of impact test such as maximum contact force, absorbed energy, deflection and duration time were compared for pure and multi scale CAFRP composited. According to the results, the addition of MWCNT improved the low velocity impact behavior of CAFRP intra-ply hybrid composites by not only increasing the toughness of epoxy matrix but also upgrade the adhesion of carbon fibers to the epoxy matrix. It was also found that fiber breakage, fiber pull out, matrix cracking, delamination and fiber matrix debonding were the most important failure modes in these types of tested composites.

Hierarchical dispersion entropy and its application in fault diagnosis of rolling bearing

Aiming at the disadvantage that multi-scale entropy only analyzes the low-frequency components of time series, this paper proposes a rolling bearing fault feature extraction method based on hierarchical dispersion entropy (HDE). Firstly, the HDE is constructed based on the hierarchical operators and dispersion entropy. Secondly, the influence of HDE parameters on entropy stability is studied. Finally, the fault diagnosis method based on HDE, least square support vector machine is applied in rolling bearings. The experiment results show that the proposed method can distinguish different types of the rolling bearing faults. It owns the higher fault diagnosis accuracy than the traditional multi scale entropy methods, such as multi-scale sample entropy (MSE), multi-scale dispersion entropy (MDE).

Optical Fiber Vibration Signal Recognition Based on the Fusion of Multi-Scale Features.

Because of the problem of low recognition accuracy in the recognition of intrusion vibration events by the distributed Sagnac type optical fiber sensing system, this paper combines the traditional optical fiber vibration signal recognition idea and the characteristics of automatic feature extraction by a convolutional neural network (CNN) to construct a new endpoint detection algorithm and a method of fusing multiple–scale features CNN to recognize fiber vibration signals. Firstly, a new endpoint detection algorithm combining spectral centroid and energy spectral entropy product is used to detect the vibration part of the original signal, which is used to improve the detection effect of endpoint detection. Then, CNNs of different scales are used to extract the multi–level and multi–scale features of the signal. Aiming at the problem of information loss in the pooling process, a new method of combining differential pooling features is used. Finally, a multi–layer perceptron (MLP) is used to recognize the extracted features. Experiments show that the method has an average recognition accuracy rate of 98.75% for the four types of vibration signals. Compared with traditional EMD and VMD pattern recognition and 1D–CNN methods, the accuracy of the optical fiber vibration signal recognition is higher.

Weighted multi-scale denoising via adaptive multi-channel fusion for compressed ultrafast photography.

Being capable of passively capturing transient scenes occurring in picoseconds and even shorter time with an extremely large sequence depth in a snapshot, compressed ultrafast photography (CUP) has aroused tremendous attention in ultrafast optical imaging. However, the high compression ratio induced by large sequence depth brings the problem of low image quality in image reconstruction, preventing CUP from observing transient scenes with fine spatial information. To overcome these restrictions, we propose an efficient image reconstruction algorithm with multi-scale (MS) weighted denoising based on the plug-and-play (PnP) based alternating direction method of multipliers (ADMM) framework for multi-channel coupled CUP (MC-CUP), named the MCMS-PnP algorithm. By removing non-Gaussian distributed noise using weighted MS denoising during each iteration of the ADMM, and adaptively adjusting the weights via sufficiently exploiting the coupling information among different acquisition channels collected by MC-CUP, a synergistic combination of hardware and algorithm can be realized to significantly improve the quality of image reconstruction. Both simulation and experimental results demonstrate that the proposed adaptive MCMS-PnP algorithm can effectively improve the accuracy and quality of reconstructed images in MC-CUP, and extend the detectable range of CUP to transient scenes with fine structures.

Multi-scale evaluation of global evapotranspiration products derived from remote sensing images: Accuracy and uncertainty

• Five global remote sensing-based evapotranspiration (ET) products are compared. • Their accuracy at point scale is evaluated based on 94 worldwide flux stations. • Accuracy at basin scale is assessed by water balance analysis across 414 basins. • ET uncertainty at pixel scale is quantified with three-cornered hat method. • ET quantity and trend at the global scale is assessed through comparison analysis. Advances in satellite remote sensing (RS) techniques have greatly prompted the development of large-scale evapotranspiration (ET) models, yielding several freely available global ET products. A comprehensive evaluation on these products is necessary for selecting the most suitable ET products and developing large-scale ET models step forward. In this study, the accuracy and uncertainty of five RS-based global ET products (MOD16, SSEBop, GLEAM, AVHRR, and BESS) were evaluated at multi scales. At point scale their accuracy was evaluated through a direct comparison with in-situ observations from 94 worldwide flux towers. Results indicate that accuracy differences of these five products vary with statistical metrics and land cover types. The combination of all statistics illustrates that GLEAM and AVHRR outperform three other products, which is consistent with the findings of most previous continental-scale studies. The accuracy at basin scale was assessed through water balance analysis across 19 big river basins. BESS is found to be the product that is superior to four other ET products at both monthly and multi-annual scale, which may be related with its low random error at point scale. The uncertainty of these five products at pixel scale was assessed by the three-cornered hat (TCH) method and comparison analysis. The TCH outputs reveal that BESS and SSEBop are the two products with the lowest and largest relative uncertainty, respectively. That explains the good performance of BESS at basin scale from another perspective. As for the two MODIS-based ET products (MOD16 and SSEBop), their accuracy at FLUXNET site scale is comparable in monthly ET estimation. However, SSEBop has special advantage in the capture of multiyear ET patterns. The comparison analysis indicates that SSEBop is the product that agrees best with the ensemble mean not only on a global scale but also in four of the six continents. Our results imply that it is difficult or even impossible to figure out the best ET product in all respects. The selection of ET products for scientific research should consider their performance difference in spatial scale as well as the influence of land cover and climate conditions.

Multi-scale rotation-equivariant graph neural networks for unsteady Eulerian fluid dynamics

The simulation of fluid dynamics, typically by numerically solving partial differential equations, is an essential tool in many areas of science and engineering. However, the high computational cost can limit application in practice and may prohibit exploring large parameter spaces. Recent deep-learning approaches have demonstrated the potential to yield surrogate models for the simulation of fluid dynamics. While such models exhibit lower accuracy in comparison, their low runtime makes them appealing for design-space exploration. We introduce two novel graph neural network (GNN) models, multi-scale (MuS)-GNN and rotation-equivariant (RE) MuS-GNN, for extrapolating the time evolution of the fluid flow. In both models, previous states are processed through multiple coarsening of the graph, which enables faster information propagation through the network and improves the capture and forecast of the system state, particularly in problems encompassing phenomena spanning a range of length scales. Additionally, REMuS-GNN is architecturally equivariant to rotations, which allows the network to learn the underlying physics more efficiently, leading to improved accuracy and generalization. We analyze these models using two canonical fluid models: advection and incompressible fluid dynamics. Our results show that the proposed GNN models can generalize from uniform advection fields to high-gradient fields on complex domains. The multi-scale graph architecture allows for inference of incompressible Navier–Stokes solutions, within a range of Reynolds numbers and design parameters, more effectively than a baseline single-scale GNN. Simulations obtained with MuS-GNN and REMuS-GNN are between two and four orders of magnitude faster than the numerical solutions on which they were trained.

Urbanization’s environmental imprint: A review

The urban population interact with their environment and change it through the consumption of resources. The changed condition may impact the wellbeing and life nature of the urban population. The uncontrollable growth of urbanization has been reported to be dangerous for mental health and sanity of many citizens. Urban areas whether small or megacities each generate an ecological foot print.The present day urbanization, consequently requires a sustainable development pattern and changes in present day styles of boom to cause them to be more equitable and more resource and energy efficientAs of now, urban administration is to a great extent concentrated on single issues, for example, water, transportation, or waste. A multi scale administration framework that expressly inscribes interconnected asset chains and interconnected spots is essential as a way to transition closer to sustainable urbanizationwhich require, prevention of urban pollution, decreasing production capacity and supporting recycling, while discouraging non-profit development and fading poor and rich differences. . Therefore, the present study was conducted to clarify the effect of urbanization on the environmental quality, resource use and sustainability. The study conducted throughout the globe indicated sustainable urbanization can be achieved by following concept of development wherein natural resources are restored and not replaced by technology.

RPPG-FuseNet: Non-contact heart rate estimation from facial video via RGB/MSR signal fusion

Remote photoplethysmography (rPPG) is an effective way of measuring Heart Rate (HR) from facial video of a person. Several methods have been proposed but measurement in realistic situation i.e, uncontrolled motion and illumination, is still a challenging task. It is observed that the performance of state-of-the-art methods often gets degraded in weekly illuminated environment.In this paper, a simple fusion based approach is proposed to reduce the performance degradation due to non-uniform illumination. An algorithm is proposed to fuse two complimentary color spaces i.e, Red, Green and Blue (RGB) and Multi Scale Retinex (MSR). The methodology includes salient spatiotemporal map generation from fusion of RGB and MSR. CNN is employed to map spatiotemporal features to corresponding HR value.To prove the efficacy of the proposed method, comparison is done with the state-of-the art methods with and without deep learning approach. Achieved experimental results proves that the proposed method yields comparable result for both inter and intra database testing on publicly available databases.

Polymer Electrolyte Fuel Cell Degradation Investigations Using X-Ray Computed Tomography

On-going research is aimed at commercializing low temperature fuel cell technology systems as zero emission alternatives for automotive applications in order to reduce greenhouse gas emissions and air pollution. These systems use polymer electrolyte fuel cells (PEFCs) to generate electricity via an electrochemical process using hydrogen and ambient air (oxygen) to produce water. Advantages of PEFCs include quick start-up time, low operating temperature, low weight, high efficiency, and relatively simple design. An important area of current research is the identification of factors which affect fuel cell performance degradation during operation and ultimately, its durability. A technique that has yielded new insights in the investigation and identification of failure modes in fuel cells is lab-based X-ray computed tomography (XCT), which is most advantageous due to its on-demand availability. The Fuel Cell Research Laboratory (FCReL) at Simon Fraser University currently operates Canada’s only facility for multi-length scale XCT, comprising of two state-of-the-art laboratory-based XCT scanners from Carl Zeiss X-ray Microscopy (Zeiss Xradia 520 Versa and 810 Ultra) with complementary resolution and field of view capabilities. This unique combination offers unprecedented access to investigations at multi length scales; with an ability to probe fuel cell components at the micro as well as the nano scale. Figure 1 illustrates the fuel cell holder and its orientation with respect to the X-ray beam as well as an exploded view of the miniature fuel cell design.An overview of recent fuel cell degradation investigations at FCReL using the XCT technique will be shown. The XCT based workflow facilitates determination and quantification of material structure and properties changes resulting from degradation stresses associated with operational parameters such as temperature, relative humidity, and voltage. The non-destructive nature of lab-based XCT visualization coupled with the ability to scan the same fuel cell multiple times without inducing damage [1] has enabled detailed studies of fuel cell degradation evolution in four dimensions (3D space, 1D time) [2 - 4]. The new knowledge gained from this procedure has led to root cause identification with respect to membrane and catalyst layer crack initiation and propagation [5 - 7], sealing issues [8,9], and subsequent mitigation toward enhanced fuel cell durability. Acknowledgement This research was supported by the Natural Sciences and Engineering Research Council of Canada, Canada Foundation for Innovation, British Columbia Knowledge Development Fund, Western Economic Diversification Canada, Canada Research Chairs, and Ballard Power Systems.

BULLDOZER: AN AUTOMATIC SELF-DRIVEN LARGE SCALE DTM EXTRACTION METHOD FROM DIGITAL SURFACE MODEL

Abstract. This study presents a Digital Terrain Model extraction method called Bulldozer. The only required input of Bulldozer is a Digital Surface Model generated from any sensors (usually optical or LIDAR) with any kind of software. After reviewing both the initial DrapCloth algorithm (Zhang et al., 2016) and its multi scale implementation (Leotta et al., 2019), some issues have been highlighted when extracting DTM from stereo satellite images such as the lost of ground adhesion under rising terrain areas, the appearance of sinks due to correlation issues when computing the DSM and finally the lack of scalability when processing large input data. Bulldozer has been developed to tackle all these issues and proposes a full automatic scalable pipeline composed of a pre-processing step to clean noisy DSMs by detecting and smoothing disturbed areas, a DTM extraction step based on a modified DrapCloth algorithm to stick to the ground under rising terrain and a post-processing step to smooth sharp sinks. The scalability has been solved using a tiling strategy and the definition of a stability margin that ensures identical results to those obtained if the whole DSM would have been processed at once in memory. As a result, Bulldozer outperforms its concurrent with respect to runtime execution while providing high quality DTMs over various types of landscapes.

An accurate estimation of preterm infants’ limb pose from depth images using deep neural networks with densely connected atrous spatial convolutions

Continuing monitoring preterm infants’ spontaneous motility during and after the hospitalization is of primary importance to early recognizing preterm-birth neurological complications. This work proposes a convolutional neural network (CNN)-based pipeline to estimate preterm infants’ limb pose from depth images acquired in the neonatal intensive care unit (NICU) during the actual clinical practice. The limb-pose estimation relies on two consecutive CNNs with dense-structured pathways of atrous spatial convolutions. The first CNN (DeA detection) has the key role of roughly detecting joint and joint-connection position while the second (DeA regression) refines the previously found predictions to trace the limb-skeleton. The architectural design of the CNNs allows smooth information flow while gathering multi scale knowledge thus enriching CNNs’ ability when predicting challenging frames (e.g., frames with self- or external- occlusions). The proposed pipeline was validated on a custom-dataset of 27000 depth frames acquired in the NICU. When applied to joint-detection, it achieved a median recall of 0.894 while the limb-pose estimation achieved a median root mean square distance error equal to 10.789 pixel. A study of efficiency on the DeA detection CNN was also conducted to derive an effective and efficient architecture for addressing the problem. We derived a CNN that achieved the same recall of the DeA detection with 2M fewer parameters. Continuing to improve the research in the field of preterm infants’ contactless monitoring will pave the way for the development of increasingly cutting-edge decision support systems embedded with deep learning for the assessment of infants’ spontaneous motility.

Design and simulation of reverse osmosis process in a hybrid forward osmosis-reverse osmosis system

Reverse Osmosis (RO) is one of the widely adopted methods for water desalination to mitigate shortage of freshwater for domestic, irrigation and industrial consumption. Recently RO is also used as a draw solution regeneration step for most hybrid Forward Osmosis-Reverse Osmosis (FO-RO) systems used for water desalination. Accurate model is required which is more flexible and reliable, to design and operate RO treatment plants for producing freshwater from various water resources. However, most models do not represent multi element field scale RO systems accurately. Plant engineers widely use the available commercial programs for the design of multi element RO systems where the computational code is not provided, which makes the program inflexible. This work presents a computational tool developed on the python platform for the prediction and analysis of the RO process employed in a hybrid FO-RO system. A steady state model based on the solution-diffusion and film theory is established by employing concentration polarization, pressure drop and average solution properties in the computational calculations of the RO process. The model is used to design and estimate the performance parameters of an entire RO system having multiple membrane elements installed in a pressure vessel. A comparison between the outputs obtained by the model and membrane manufacturers software shows good agreement.

High depolarization temperature and superior piezoelectric performance with complex structural evolution in BiFeO3[sbnd]PbTiO3–(Sr0.7Bi0.2□0.1)TiO3 systems

Bismuth ferrite–based ferroelectric ceramics are considered strong competitors in high–temperature piezoelectric applications that benefits from their high depolarization temperature (Td), but problems of large conductivity and low piezoelectric coefficient (d33) should be tackled. BiFeO3PbTiO3–x(Sr0.7Bi0.2□0.1)TiO3 (BF–PT–xSBT) ternary system are designed in this work that successfully resolves this tough paradox. Rietveld refinements show that this system exhibits multiphase coexistence with complex structural transition among rhombohedral, tetragonal and pseudocubic phase. Interestingly, a high Td ∼320 °C and d33 ∼ 236 pC/N are synergistically optimized in x=0.20 composition near pseudo–phase boundary, which are indicated by multi–scale techniques (phase structure, dielectric analysis etc.). What's more, dual electrostrain peaks appear at separate phase boundaries for x=0.20 and 0.66 compositions with distinct piezo/ferroelectric properties. Rayleigh law and piezoelectric force microscopy analysis clarify that the optimal electrostrain performance for both critical compositions are dominated by intrinsic (x=0.20) and extrinsic (x=0.66) contributions, respectively. This work not only provides a new BF–PT–xSBT system with a high Td and superior d33 that are promising candidates in high–temperature actuator and sensor applications, but also presents a possibility of multiphase–coexistence established exotic macroscopic performances.

Improved resolution of D-bar images of ventilation using a Schur complement property and an anatomical atlas.

BackgroundElectrical impedance tomography (EIT) is a nonionizing imaging technique for real‐time imaging of ventilation of patients with respiratory distress. Cross‐sectional dynamic images are formed by reconstructing the conductivity distribution from measured voltage data arising from applied alternating currents on electrodes placed circumferentially around the chest. Since the conductivity of lung tissue depends on air content, blood flow, and the presence of pathology, the dynamic images provide regional information about ventilation, pulsatile perfusion, and abnormalities. However, due to the ill‐posedness of the inverse conductivity problem, EIT images have a coarse spatial resolution. One method of improving the resolution is to include prior information in the reconstruction.PurposeIn this work, we propose a technique in which a statistical prior built from an anatomical atlas is used to postprocess EIT reconstructions of human chest data. The effectiveness of the method is demonstrated on data from two patients with cystic fibrosis.MethodsA direct reconstruction algorithm known as the D‐bar method was used to compute a two‐dimensional reconstruction of the conductivity distribution in the plane of the electrodes. Reconstructions using one step in an iterative (regularized) Newton's method were also computed for comparison. An anatomical atlas consisting of 1589 synthetic EIT images computed from X‐ray computed tomography (CT) scans of 74 adult male subjects was computed for use as a statistical prior. The resolution of the D‐bar images was then improved by maximizing the conditional probability density function of an image that is consistent with the a priori information and the statistical model. A new method to evaluate the accuracy of the EIT images using CT scans of the imaged patient as ground truth is presented. The novel approach is tested on data from two patients with cystic fibrosis.Results and ConclusionsThe D‐bar images resulted in better structural similarity index measures (SSIM) and multiscale (MS) SSIM measures for both subjects using the mask or amplitude evaluation approach than the one‐step (regularized) Newton's method. Further improvement was achieved using the Schur complement (SC) approach, with MS‐SSIM values of 0.718 and 0.682 using SC evaluated with the mask and amplitude approach, respectively, for Patient 1, and MS‐SSIM values of 0.726 and 0.692 using SC evaluated with the mask and amplitude approach, respectively, for Patient 2. The results from applying an anatomical atlas and statistical prior to EIT data from two patients with cystic fibrosis suggest that the spatial resolution of the EIT image can be improved to reveal pathology that may be difficult to discern in the original EIT image. The novel metric of evaluation is consistent with the appearance of improved spatial resolution and provides a new way to evaluate the accuracy of EIT reconstructions when a CT scan is available.