Full 3D Data Research Articles

Comparison of 2D and 3D convolutional neural networks in hyperspectral image analysis of fruits applied to orange bruise detection.

Recent advances in hyperspectral imaging (HSI) have demonstrated its ability to detect defects in fruit that may not be visible in RGB images. HSIs can be considered 3D images containing two spatial dimensions and one spectral dimension. Therefore, the first question that arises is how to process this type of information, either using 2D or 3D models. In this study, HSI in the 550-900nm spectral range was used to detect bruising in oranges. Sixty samples of Thompson oranges were subjected to a mechanical bruising process, and HSIs were taken at different time intervals: before bruising, and 8 and 16h after bruising. The samples were then classified using two convolutional neural network (CNN) models, a shallow 7-layer network (CNN-7) and a deep 18-layer network (CNN-18). In addition, two different input processing approaches are used: using 2D information from each band, and using the full 3D data from each HSI. The 3D models were the most accurate, with 94% correct classification for 3D-CNN-18, compared to 90% for 3D-CNN-7, and less than 83% for the 2D models. Our study suggests that 3D HSI may be a more effective technique for detecting fruit bruising, allowing the development of a fast, accurate, and nondestructive method for fruit sorting. PRACTICAL APPLICATION: Orange bruises can reduce the market value of food, which is why the food processing industry needs to carry out quality inspections. An effective way to perform this inspection is by using hyperspectral images that can be processed with 2D or 3D models, either with deep or shallow neural networks. The results of the comparison performed in this work can be useful for the development of more accurate and efficient bruise detection methods for fruit inspection.

Identifying a response parameter in a model of brain tumour evolution under therapy

Abstract A nonlinear conjugate gradient method is derived for the inverse problem of identifying a treatment parameter in a nonlinear model of reaction–diffusion type corresponding to the evolution of brain tumours under therapy. The treatment parameter is reconstructed from additional information about the tumour taken at a fixed instance of time. Well-posedness of the direct problems used in the iterative method is outlined as well as uniqueness of a solution to the inverse problem. Moreover, the parameter identification is recasted as the minimization of a Tikhonov type functional and the existence of a minimizer to this functional is shown. Finite-difference discretization of the space and time derivatives are employed for the numerical implementation. Numerical simulations on full 3D brain data are included showing that information about a spacewise-dependent treatment parameter can be recovered in a stable way.

Invertible Modeling of Bidirectional Relationships in Neuroimaging With Normalizing Flows: Application to Brain Aging.

Many machine learning tasks in neuroimaging aim at modeling complex relationships between a brain's morphology as seen in structural MR images and clinical scores and variables of interest. A frequently modeled process is healthy brain aging for which many image-based brain age estimation or age-conditioned brain morphology template generation approaches exist. While age estimation is a regression task, template generation is related to generative modeling. Both tasks can be seen as inverse directions of the same relationship between brain morphology and age. However, this view is rarely exploited and most existing approaches train separate models for each direction. In this paper, we propose a novel bidirectional approach that unifies score regression and generative morphology modeling and we use it to build a bidirectional brain aging model. We achieve this by defining an invertible normalizing flow architecture that learns a probability distribution of 3D brain morphology conditioned on age. The use of full 3D brain data is achieved by deriving a manifold-constrained formulation that models morphology variations within a low-dimensional subspace of diffeomorphic transformations. This modeling idea is evaluated on a database of MR scans of more than 5000 subjects. The evaluation results show that our bidirectional brain aging model (1) accurately estimates brain age, (2) is able to visually explain its decisions through attribution maps and counterfactuals, (3) generates realistic age-specific brain morphology templates, (4) supports the analysis of morphological variations, and (5) can be utilized for subject-specific brain aging simulation.

Research on a Design and Manufacturing Integration Mode Based on a Full 3D Model

In view of issues such as low reuse rate of three-dimensional design model in process and production during product research and design, insufficient ability of 3D process design and optimization application, and urgent needs for a closed-loop design-manufacture-inspection integrated control system based on 3D model, this paper is centered around the digital definition and application technologies based on 3D models and facing full process information of product design-process design-production-inspection, combining the product development process, and studies the design and manufacturing integration framework based on a full 3D model, software architecture and data integration management architecture. This paper too selects products with multi-subject features as verification objects and verifies the feasibility and superiority of the integrated design-production-inspection development mode based on a full 3D model.

Positron Emission Particle Tracking for Liquid‐Solid Mixing in Stirred Tanks

AbstractMixing in stirred tanks is vital to a wide range of industrial processes, typically requiring solids to be fully suspended and evenly distributed. The quality of suspension and mixing is typically measured visually. Such measurements are thus subject to human interpretation and can only feasibly be conducted in transparent systems – an uncommon condition in industrial systems. Visual measurements of homogeneity must also infer full 3D distributions of particles solely from the observation of a limited section thereof. This work details methods by which a system's homogeneity and state of suspension may be measured and quantified with no human interpretation, which can extract information from opaque systems and consider full 3D data acquired throughout the interior of a given system.

3D-FLARE: A Touchless Full-3D Fingerprint Recognition System Based on Laser Sensing

“The friction ridge pattern is a 3D structure which, in its natural state, is not deformed by contact with a surface”. Building upon this rather trivial observation, the present work constitutes a first solid step towards a paradigm shift in fingerprint recognition from its very foundations. We explore and evaluate the feasibility to move from current technology operating on 2D images of elastically deformed impressions of the ridge pattern, to a new generation of systems based on full-3D models of the natural non-deformed ridge pattern itself. There are already a small number of previous studies that have already started scratching the surface of 3D fingerprint recognition and that should not go overlooked. However, the vast majority of these few successful approaches published so far, are based on the reconstruction of fingerprints from multiple 2D images acquired with different lighting conditions (photometric stereo 3D reconstruction) or acquired from different angles (stereo vision 3D reconstruction). Such reconstruction methods lead in general to 2D fingerprints wrapped over the overall volume of the finger. These volumetric fingerprints have shown some promising performance, but still miss the real depth information of the ridge pattern, which, in the best case scenario, is coarsely estimated during the error-prone reconstruction process. In the present work we take one step further, directly acquiring for the first time in a consistent and repeatable manner, full-3D fingerprint models stored as point-clouds, where each point is defined by its $[x,y,z]$ coordinates. This way, the 3D data is directly measured by the sensor, with no post-processing reconstruction stage required. The complete recognition system developed represents as well an alternative to traditional technology based on minutiae detection. It shows that image-based processing algorithms and descriptors can be successfully applied to the new full-3D data, reaching very competitive results and confirming the high distinctiveness of the models.

Plantar pressure and shear measurement using surface stress-sensitive film

This paper describes the performance of a system for dynamic measurements of pressure and shear forces over the entire plantar surface of the foot. During human locomotion, mechanical stress at the plantar surface is 3D: pressure acting vertically and shear stress acting longitudinally and laterally. The system described here offers gait, foot deformity, diabetic ulceration, and foot-ankle kinetics researchers a means to obtain the full 3D data that is missing in most current research efforts. Surface Stress Sensitive Film™ (S3F) provides the sensing element for this measurement device. S3F is an incompressible elastic film that deforms in response to applied forces. Upon removal of the forces, the film returns to its original shape. The dynamic, 3D deformations are captured with a digital camera. The measured deformations can be analyzed directly, since they are proportional to the applied forces, or used as inputs to a finite element model of the film to reconstruct the pressure and shear force values. System performance is evaluated using both a standard form and human foot contact. The fundamental approach involves comparing the integrated normal and shear loads from the S3F-based sensor to the measurements provided by a force plate. The results showed high linearity (<3% nonlinearity), low uncertainty (<4%), low cross talk (<3%), high spatial resolution (1.6 mm2), no loss of fidelity at walking speeds with a 45 Hz sampling rate, and good repeatability for human foot contact measurements.

Murine Mouse Models of Metastatic Colorectal Cancer in Lungs: Foci Counts in Three Dimensions

In standard histological slide preparations, only a few representative samples are taken, which introduces sampling bias and necessarily eschews information about the full tissue sample. However, it is possible to obtain a full 3D data set of an entire tissue sample using the Knife Edge Scanning Microscope (KESM), an automated simultaneous sectioning and imaging microscope. In this study, we performed manual counts of metastases of colorectal cancer in mouse model lungs stained with India ink and hematoxylin and eosin, which were then sectioned and imaged in their entirety. We were able to get more accurate count of tumor foci than the current standard of surface nodule count can allow; multiple tumor foci were buried hundreds of microns deep into the parenchyma and hence not grossly visible from the surface. Volumetric analysis of whole samples also revealed unusual tumor morphology and permitted the visualization of unique architectural features where multiple tumor foci had grown together. Following the manual count of tumor foci, we developed a semi‐automated algorithm that could increase the efficiency of screening for foci, and create the opportunity for digital 3D reconstruction and morphology interrogation. Without access to a full dataset of the lungs, it is impossible to determine true foci number or morphology.This abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.

3D-specific absorption rate estimation from high-intensity focused ultrasound sonications using the Green's function heat kernel.

To evaluate a numerical inverse Green's function method for deriving specific absorption rates (SARs) from high-intensity focused ultrasound (HIFU) sonications using tissue parameters (thermal conductivity, specific heat capacity, and mass density) and three-dimensional (3D) magnetic resonance imaging (MRI) temperature measurements. SAR estimates were evaluated using simulations and MR temperature measurements from HIFU sonications. For simulations, a "true" SAR was calculated using the hybrid angular spectrum method for ultrasound simulations. This "true" SAR was plugged into a Pennes bioheat transfer equation(PBTE) solver to provide simulated temperature maps, which were then used to calculate the SAR estimate using the presented method. Zero mean Gaussian noise, corresponding to temperature precisions between 0.1 and 2.0°C, was added to the temperature maps to simulate a variety of invivo situations. Experimental MR temperature maps from HIFU sonications in a gelatin phantom monitored with a 3D segmented echo planar imaging MRI pulse sequence were also used. To determine the accuracy of the simulated and phantom data, we reconstructed temperature maps by plugging in the estimated SAR to the PBTE solver. In both simulations and phantom experiments, the presented method was compared to two previously published methods of determining SAR, a linear and an analytical method. The presented numerical method utilized the full 3D data simultaneously, while the two previously published methods work on a slice-by-slice basis. In the absence of noise, SAR distribution estimates obtained from the simulated heating profiles match closely (within 10%) to the initial true SAR distribution. The resulting temperature distributions also match closely to the corresponding initial temperature distributions (<0.2°C RMSE). In the presence of temperature measurement noise, the SAR distributions have noise amplified by the inverse convolution process, while the resulting temperature distributions still match closely to the initial "true" temperature distributions. In general, temperature RMSE was observed to be approximately 20-30% higher than the level of the added noise. By contrast, the previously published linear method is less sensitive to noise, but significantly underpredicts the SAR. The analytic method is also less sensitive to noise and matches SAR in the central plane, but greatly underpredicts in the longitudinal direction. Similar observations are made from the phantom studies. The described numerical inverse Green's function method is very fast - at least two orders of magnitude faster than the compared methods. The presented numerical inverse Green's function method is computationally fast and generates temperature maps with high accuracy. This is true despite generally overestimating the true SAR and amplifying the input noise.

P1246Age-, body size-, and sex-specific reference values for tricuspid valve apparatus parameters by real-time three-dimensional transthoracic echocardiographyP1247Surgical therapy in infective endocarditis: predisposing and prognostic factorsP1248Evalutation of MV annulus geometry modifications before and after mitraclip implantation with 3D transesophageal echocardiographyP1249Prognostic echocardiographic parameters for successful transcatheter edge-to-edge mitral valve repairP1250Influence of body surface area in diagnosis and outcome patient-prosthesis mismatch after TAVI.

and Purpose-Tricuspid valve (TV) apparatus parameters can vary significantly with demographic and anthropometric factors and could be useful for clinical decision making. Our aim was to (1) establish the reference values for TV apparatus parameters using transthoracic three-dimensional (3D) echocardiography; (2) investigate the influence of age, sex, and body size on TV anatomy.

Marginal Space Deep Learning: Efficient Architecture for Volumetric Image Parsing.

Robust and fast solutions for anatomical object detection and segmentation support the entire clinical workflow from diagnosis, patient stratification, therapy planning, intervention and follow-up. Current state-of-the-art techniques for parsing volumetric medical image data are typically based on machine learning methods that exploit large annotated image databases. Two main challenges need to be addressed, these are the efficiency in scanning high-dimensional parametric spaces and the need for representative image features which require significant efforts of manual engineering. We propose a pipeline for object detection and segmentation in the context of volumetric image parsing, solving a two-step learning problem: anatomical pose estimation and boundary delineation. For this task we introduce Marginal Space Deep Learning (MSDL), a novel framework exploiting both the strengths of efficient object parametrization in hierarchical marginal spaces and the automated feature design of Deep Learning (DL) network architectures. In the 3D context, the application of deep learning systems is limited by the very high complexity of the parametrization. More specifically 9 parameters are necessary to describe a restricted affine transformation in 3D, resulting in a prohibitive amount of billions of scanning hypotheses. The mechanism of marginal space learning provides excellent run-time performance by learning classifiers in clustered, high-probability regions in spaces of gradually increasing dimensionality. To further increase computational efficiency and robustness, in our system we learn sparse adaptive data sampling patterns that automatically capture the structure of the input. Given the object localization, we propose a DL-based active shape model to estimate the non-rigid object boundary. Experimental results are presented on the aortic valve in ultrasound using an extensive dataset of 2891 volumes from 869 patients, showing significant improvements of up to 45.2% over the state-of-the-art. To our knowledge, this is the first successful demonstration of the DL potential to detection and segmentation in full 3D data with parametrized representations.

Three-dimensional valency mapping in ceria nanocrystals.

Using electron tomography combined with electron energy loss spectroscopy (EELS), we are able to map the valency of the Ce ions in CeO2-x nanocrystals in three dimensions. Our results show a clear facet-dependent reduction shell at the surface of ceria nanoparticles; {111} surface facets show a low surface reduction, whereas at {001} surface facets, the cerium ions are more likely to be reduced over a larger surface shell. Our generic tomographic technique allows a full 3D data cube to be reconstructed, containing an EELS spectrum in each voxel. This possibility enables a three-dimensional investigation of a plethora of material-specific physical properties such as valency, chemical composition, oxygen coordination, or bond lengths, triggering the synthesis of nanomaterials with improved properties.

Texture feature analysis for computer-aided diagnosis on pulmonary nodules.

Differentiation of malignant and benign pulmonary nodules is of paramount clinical importance. Texture features of pulmonary nodules in CT images reflect a powerful character of the malignancy in addition to the geometry-related measures. This study first compared three well-known types of two-dimensional (2D) texture features (Haralick, Gabor, and local binary patterns or local binary pattern features) on CADx of lung nodules using the largest public database founded by Lung Image Database Consortium and Image Database Resource Initiative and then investigated extension from 2D to three-dimensional (3D) space. Quantitative comparison measures were made by the well-established support vector machine (SVM) classifier, the area under the receiver operating characteristic curves (AUC) and the p values from hypothesis t tests. While the three feature types showed about 90% differentiation rate, the Haralick features achieved the highest AUC value of 92.70% at an adequate image slice thickness, where a thinner or thicker thickness will deteriorate the performance due to excessive image noise or loss of axial details. Gain was observed when calculating 2D features on all image slices as compared to the single largest slice. The 3D extension revealed potential gain when an optimal number of directions can be found. All the observations from this systematic investigation study on the three feature types can lead to the conclusions that the Haralick feature type is a better choice, the use of the full 3D data is beneficial, and an adequate tradeoff between image thickness and noise is desired for an optimal CADx performance. These conclusions provide a guideline for further research on lung nodule differentiation using CT imaging.

Two-pulse biexponential-weighted 23Na imaging

A new method is proposed for acquiring 3D biexponential-weighted sodium images with two instead of three RF pulses to allow for shorter repetition time at high magnetic fields (B0⩾7T) and reduced SAR.The second pulse converts single- into triple-quantum coherences in regions containing sodium ions which are restricted in mobility. Since only single-quantum coherences can be detected, an image acquired after the second pulse is intrinsically single-quantum-filtered and can be used to generate a biexponential-weighted sodium image by a weighted subtraction with the spin-density-weighted image acquired between the pulses.The proposed sequence generates biexponential-weighted sodium images of in vivo human brain with 140% higher SNR than triple-quantum-filtered sodium images and 4% higher SNR than a biexponential-weighted sequence with three RF pulses at equal acquisition time and with 1/3 lower SAR. As SAR is reduced, accordingly repetition time can be spared to obtain even higher SNR-time efficiency. In comparison to a difference image generated from two images of a double-readout sequence, the proposed two-pulse sequence yields about 14% higher SNR.Our new two-pulse biexponential-weighted sequence allows for acquisition of full 3D data sets of the human brain in vivo with a nominal resolution of (5mm)3 in about 10min.

Design and Implementation of a Custom Built Optical Projection Tomography System

Optical projection tomography (OPT) is an imaging modality that has, in the last decade, answered numerous biological questions owing to its ability to view gene expression in 3 dimensions (3D) at high resolution for samples up to several cm3. This has increased demand for a cabinet OPT system, especially for mouse embryo phenotyping, for which OPT was primarily designed for. The Medical Research Council (MRC) Technology group (UK) released a commercial OPT system, constructed by Skyscan, called the Bioptonics OPT 3001 scanner that was installed in a limited number of locations. The Bioptonics system has been discontinued and currently there is no commercial OPT system available. Therefore, a few research institutions have built their own OPT system, choosing parts and a design specific to their biological applications. Some of these custom built OPT systems are preferred over the commercial Bioptonics system, as they provide improved performance based on stable translation and rotation stages and up to date CCD cameras coupled with objective lenses of high numerical aperture, increasing the resolution of the images. Here, we present a detailed description of a custom built OPT system that is robust and easy to build and install. Included is a hardware parts list, instructions for assembly, a description of the acquisition software and a free download site, and methods for calibration. The described OPT system can acquire a full 3D data set in 10 minutes at 6.7 micron isotropic resolution. The presented guide will hopefully increase adoption of OPT throughout the research community, for the OPT system described can be implemented by personnel with minimal expertise in optics or engineering who have access to a machine shop.

DIRECT ORTHOPHOTO GENERATION FROM COLOR POINT CLOUDS OF COMPLEX SCENES

Abstract. During the last decade, laser scanning and automated photogrammetric techniques, proved that it is possible to acquire dense colored point cloud datasets. The density of such datasets sometimes equals or exceeds the density of orthophotos. The standard orthophoto generation process using such dense and full 3D data sets, is tedious as it requires manual corrections over the 3D mesh model, as well as mosaicking and gap filling from adjacent photographs. This paper suggests a fast automated method to produce orthophotos directly from color point clouds, without the need to create a mesh surface, nor to mosaic or fill gaps. Photos from radio controlled helicopter are being processed with traditional photogrammetric process and modern computer vision techniques, against the proposed method. Reference data from land surveying are being used to compare orthophotos from different sources.

Building 3D GIS data models using open source software

Today, many GIS applications would require data sets containing full 3D data; application fields like urban and town planning, civil protection, cadastres and environmental studies could use full 3D data, both to carry out spatial analyses and inquiries as well as for only visualization. Data model can be accessed by desktop application as well as from Internet. This second option is becoming today the most requested. But in order to share 3D data model through Internet it is necessary to configure a suitable system. A 3D model can be shared on Internet in different ways. It is possible to incorporate a model inside an executable or 3D PDF file or build something like a multimedia application (i.e. flash based). However a better option is to put a model inside a geospatial database and build suitable web services to access it using Internet browsers. This second option can be performed using different solutions for web servers, development platforms and programming languages. Fundamental for this option is the construction of the model that should be structured, according a conceptual model to be replicable and interoperable. Hence, the model should be stored in a geospatial database in order to be accessed by an application running under a web server, so allowing access to the model. All this is the object of the research—Management and use of distributed 3D data by open source WebGIS software—that is part of the Italian PRIN 2007 research project: Interoperability and cooperative management of geographic, dynamic, multi-dimensional and distributed data with Free and Open Source GIS (Principal investigator Paolo Zatelli), aimed to build urban and suburban 3D models with CAD software or solid modellers, share them with a 3D Web Feature Service Open Geospatial Consortium web feature service like, and them using Internet browsers all done using only free and open source software. The research work, here illustrated, which was funded by previous research project, deals with the construction of a suitable model and the storage of the model inside a geospatial database. The work is based on the use of free and open source software; Blender and PostgreSQL with PostGIS have been used, respectively, to draw objects of the model and save their data with the help of some Python scripts; the server side part of the system has been realized using Tomcat with servlets and applets all written in Java. Buildings have been modelled according to GIANT3D model (Geographical Interoperable Advanced Numerical Topological 3-Dimensional Model) previously developed in the research "PRIN 2004", regarding "Evolved structure of numerical cartography for GIS and WebGIS". Python scripts, activated by Blender, allow saving data into a spatial database implemented through PostgreSQL and PostGIS that could be a remote database somewhere on the net; all geometrical and topological information, implemented in the 3D model, is so transferred in PostGIS. This information can be retrieved by Blender using other Python scripts, so Blender fully interacts with 3D data allocated in PostGIS. These data can be also accessed by many other clients, both directly using a database client and using other protocols (like HTTP on Internet). The second part of the research deals with the creation of a 3D Web Feature Service and an applet for internet browsers, in order to allow clients to visualize, explore and inquiry 3D model, retrieving data from database.

P2-10-09: Detecting Breast Cancer with Dynamic Diffuse Optical Tomographic Imaging.

Abstract Background: Over the last decade diffuse optical tomography (DOT) has emerged as a novel medical imaging modality. Near-infrared light is used to non-invasively probe biological tissue, and three-dimensional (3D) maps of blood-dependent parameters can be obtained. Several studies are underway to show the clinical utility of DOT for imaging brain disease, joint disease, and breast cancer. Here we present the largest clinical study to date that uses hemodynamic effects caused by a simple breath hold to identify breast tumors using DOT. Methods: We have designed and built a DOT breast imaging system that can acquire full 3D data sets within a fraction of a second. The system affords the ability to study fast hemodynamic effects in both breasts simultaneously. Using non-compressive imaging heads we measured the oxygenated and deoxygenated hemoglobin levels in the breasts of 15 patients whose mammogram showed a mass &amp;gt;1 cm (4 benign, 11 malignant) and 3 healthy controls. Data was acquired over the course of a 30-second breath hold and 30 seconds thereafter. In addition, a baseline measurement of 30 seconds prior to the breath hold was obtained. A multi-wavelength image reconstruction algorithm was used to create 3D maps of hemoglobin-dependent parameters (Δ[HbO2] and Δ[Hb]) in the breast every 0.58 seconds, over the course of the 90-second experiment. An image analysis algorithm identified regions of peak percentage change in [Hb] and [HbO2] in the breast and then computed the average hemoglobin levels in those regions. Results: We observed an increase in the hemoglobin levels in all breasts during the breath hold. Upon resuming breathing, these hemoglobin levels returned to baseline. Tumor bearing breasts showed a statistically significant slower return to baseline than healthy breasts. In particular, we found that tumors can be detected by a substantially larger Δ[Hb] value as compared to normal tissue in images acquired 15 seconds following the end of the breath hold. In 10 of 11 patients the malignant tumor was identified using this technique, suggesting sensitivity over 90%. No regions of increased Δ[Hb] were seen in the healthy breasts, or in the breasts with benign masses. The peak percentage change in [Hb] at the 15 second post-breath hold time point was 10.0 ± 6.0% (n=11) in the malignant tumors compared to 1.4 ± 0.5% (n=3) (p=0.001) in healthy patients and 4.8 ± 1.9% in benign masses (n=4)(p=0.03). Discussion: A breath hold impedes venous return to the heart, which causes pooling of blood in the breast. This is observed as an increase in [Hb] and [HbO2] using DOT. Tumor vasculature is known to be more disorganized, tortuous, and leakier than normal vasculature. Therefore, once the breath hold is released and blood is allowed to drain from the breast, blood accumulated in the tumor during the breath hold will drain more slowly than blood in healthy tissue. This study has shown that DOT allows us to visualize these hemodynamic effects and use them to detect tumors with a simple breath hold and compression-free imaging head. Future studies need to explore the detection limits and general clinical utility of this technique for screening, differentiating malignant from benign masses, and treatment monitoring. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-10-09.

Constitutive Modeling of Coronary Arterial Media—Comparison of Three Model Classes

Accurate modeling of arterial elasticity is imperative for predicting pulsatile blood flow and transport to the periphery, and for evaluating the mechanical microenvironment of the vessel wall. The goal of the present study is to compare a recently developed structural model of porcine left anterior descending artery media to two commonly used typical representatives of phenomenological and structure-motivated invariant-based models, in terms of the number of model parameters, model descriptive and predictive powers, and requisite different test protocols for reliable parameter estimation. The three models were compared against 3D data of radial inflation, axial extension, and twist tests. Also checked are the models predictive capabilities to response data not used for estimation, including both tests outside the range of estimation database, as well as protocols of a different nature. The results show that the descriptive estimation error (model fit to estimation database), measured by the sum of squared residuals (SSE) between full 3D data and model predictions, was about twice as low for the structural (4.58%) model compared to the other two (9.71 and 8.99% for the phenomenological and structure-motivated models, respectively). Similar SSE ratios were obtained for the predictive capabilities. Prediction SSE at high stretch based on estimation of two low stretches yielded an SSE value of 2.81% for the structural model, and 10.54% and 7.87% for the phenomenological and structure-motivated models, respectively. For the prediction of twist from inflation-extension data, SSE values for the torsional stiffness was 1.76% for the structural model and 39.62 and 2.77% for the phenomenological and structure-motivated models. The required number of model parameters for the structural model is four, whereas the phenomenological model requires six to nine and the structure-motivated has four parameters. These results suggest that modeling based on the tissue structural features improves model reliability in describing given data and in predicting the tissue general response.

The HUDSEN Atlas: a three‐dimensional (3D) spatial framework for studying gene expression in the developing human brain

We are developing a three-dimensional (3D) atlas of the human embryonic brain using anatomical landmarks and gene expression data to define major subdivisions through 12 stages of development [Carnegie Stages (CS) 12-23; approximately 26-56 days post conception (dpc)]. Virtual 3D anatomical models are generated from intact specimens using optical projection tomography (OPT). Using MAPAINT software, selected gene expression data, gathered using standard methods of in situ hybridization and immunohistochemistry, are mapped to a representative 3D model for each chosen Carnegie stage. In these models, anatomical domains, defined on the basis of morphological landmarks and comparative knowledge of expression patterns in vertebrates, are linked to a developmental neuroanatomic ontology. Human gene expression patterns for genes with characteristic expression in different vertebrates (e.g. PAX6, GAD65 and OLIG2) are being used to confirm and/or refine the human anatomical domain boundaries. We have also developed interpolation software that digitally generates a full domain from partial data. Currently, the 3D models and a preliminary set of anatomical domains and ontology are available on the atlas pages along with gene expression data from approximately 100 genes in the HUDSEN Human Spatial Gene Expression Database (http://www.hudsen.org). The aim is that full 3D data will be generated from expression data used to define a more detailed set of anatomical domains linked to a more advanced anatomy ontology and all of these will be available online, contributing to the long-term goal of the atlas, which is to help maximize the effective use and dissemination of data wherever it is generated.