Real Anatomical Data Research Articles

Learning-based deformable image registration: effect of statistical mismatch between train and test images.

Convolutional neural networks (CNNs) offer a promising means to achieve fast deformable image registration with accuracy comparable to conventional, physics-based methods. A persistent question with CNN methods, however, is whether they will be able to generalize to data outside of the training set. We investigated this question of mismatch between train and test data with respect to first- and second-order image statistics (e.g., spatial resolution, image noise, and power spectrum). A UNet-based architecture was built and trained on simulated CT images for various conditions of image noise (dose), spatial resolution, and deformation magnitude. Target registration error was measured as a function of the difference in statistical properties between the test and training data. Generally, registration error is minimized when the training data exactly match the statistics of the test data; however, networks trained with data exhibiting a diversity in statistical characteristics generalized well across the range of statistical conditions considered. Furthermore, networks trained on simulated image content with first- and second-order statistics selected to match that of real anatomical data were shown to provide reasonable registration performance on real anatomical content, offering potential new means for data augmentation. Characterizing the behavior of a CNN in the presence of statistical mismatch is an important step in understanding how these networks behave when deployed on new, unobserved data. Such characterization can inform decisions on whether retraining is necessary and can guide the data collection and/or augmentation process for training.

Impact of correction factors in human brain lesion-behavior inference.

Statistical voxel-based lesion-behavior mapping (VLBM) in neurological patients with brain lesions is frequently used to examine the relationship between structure and function of the healthy human brain. Only recently, two simulation studies noted reduced anatomical validity of this method, observing the results of VLBM to be systematically misplaced by about 16 mm. However, both simulation studies differed from VLBM analyses of real data in that they lacked the proper use of two correction factors: lesion size and "sufficient lesion affection." In simulation experiments on a sample of 274 real stroke patients, we found that the use of these two correction factors reduced misplacement markedly compared to uncorrected VLBM. Apparently, the misplacement is due to physiological effects of brain lesion anatomy. Voxel-wise topographies of collateral damage in the real data were generated and used to compute a metric for the inter-voxel relation of brain damage. "Anatomical bias" vectors that were solely calculated from these inter-voxel relations in the patients' real anatomical data, successfully predicted the VLBM misplacement. The latter has the potential to help in the development of new VLBM methods that provide even higher anatomical validity than currently available by the proper use of correction factors. Hum Brain Mapp 38:1692-1701, 2017. © 2017 Wiley Periodicals, Inc.

MCNPX internal dosimetry studies based on the NORMAN-05 voxel model

Anthropomorphic computational models coupled with radiation transport codes are valuable tools in radiation protection dosimetry. In particular, they are very reliable for the estimate of the energy absorbed by different organs due to an incorporated radionuclide. MIRD-based stylised analytical models are widely accepted as standards but the recent generation of voxel phantoms, developed on real anatomical data derived from tomographic images, can represent a valid alternative for radiation protection and dosimetry purposes. Specific absorbed fraction evaluation and patient-specific dose estimate in nuclear medicine and radiotherapy could be considered as the optimal area for their implementation and use. On the other hand, the accuracy of organ and body structure representation guarantees an improved dose evaluation system also for radiation protection purposes in the workplace in case of accidental internal contamination. In the present work the voxel model NORMAN-05, a modified version of NORMAN (HPA, UK) model, has been employed with the Monte Carlo code MCNPX. Some preliminary investigations were carried out to evaluate the absorbed fractions for a series of source-target organ couples in case of gamma emitters and the organ absorbed doses in case of 90Sr incorporation. The paper summarises the main preliminary outcomes of such studies.

Deriving physical connectivity from neuronal morphology.

A model is presented that allows prediction of the probability for the formation of appositions between the axons and dendrites of any two neurons based only on their morphological statistics and relative separation. Statistics of axonal and dendritic morphologies of single neurons are obtained from 3D reconstructions of biocytin-filled cells, and a statistical representation of the same cell type is obtained by averaging across neurons according to the model. A simple mathematical formulation is applied to the axonal and dendritic statistical representations to yield the probability for close appositions. The model is validated by a mathematical proof and by comparison of predicted appositions made by layer 5 pyramidal neurons in the rat somatosensory cortex with real anatomical data. The model could be useful for studying microcircuit connectivity and for designing artificial neural networks.

Automatic Modeling of Knee-Joint Motion For The Virtual Reality Dynamic Anatomy (VRDA) Tool

This paper presents a method and algorithms for automatic modeling of anatomical joint motion. The method relies on collision detection to achieve stable positions and orientations of the knee joint by evaluating the relative motion of the tibia with respect to the femur (for example, flexion-extension). The stable positions then become the basis for a look-up table employed in the animation of the joint. The strength of this method lies in its robustness to animate any normal anatomical joint. It is also expandable to other anatomical joints given a set of kinematic constraints for the joint type as well as a high-resolution, static, 3-D model of the joint. The demonstration could be patient specific if a person's real anatomical data could be obtained from a medical imaging modality such as computed tomography or magnetic resonance imaging. Otherwise, the demonstration requires the scaling of a generic joint based on patient characteristics. Compared with current teaching strategies, this Virtual Reality Dynamic Anatomy (VRDA) tool aims to greatly enhance students' understanding of 3-D human anatomy and joint motions. A preliminary demonstration of the optical superimposition of a generic knee joint on a leg model is shown.