scholarly journals Three Dimensional Root CT Segmentation using Multi-Resolution Encoder-Decoder Networks

2019 ◽  
Author(s):  
Mohammadreza Soltaninejad ◽  
Craig J. Sturrock ◽  
Marcus Griffiths ◽  
Tony P. Pridmore ◽  
Michael P. Pound
Keyword(s):  
Receptive Field ◽  
Contextual Information ◽  
Three Dimensional ◽  
User Interaction ◽  
Root System Architecture ◽  
Complex Problem ◽  
Field Size ◽  
Learning Approach ◽  
Analysis Tool ◽  
Ct Segmentation

AbstractWe address the complex problem of reliably segmenting root structure from soil in X-ray Computed Tomography (CT) images. We utilise a deep learning approach, and propose a state-of-the-art multi-resolution architecture based on encoder-decoders. While previous work in encoder-decoders implies the use of multiple resolutions simply by downsampling and upsampling images, we make this process explicit, with branches of the network tasked separately with obtaining local high-resolution segmentation, and wider low-resolution contextual information. The complete network is a memory efficient implementation that is still able to resolve small root detail in large volumetric images. We evaluate our approach by comparing against a number of different encoder-decoder based architectures from the literature, as well as a popular existing image analysis tool designed for root CT segmentation. We show qualitatively and quantitatively that a multi-resolution approach offers substantial accuracy improvements over a both a small receptive field size in a deep network, or a larger receptive field in a shallower network. We obtain a Dice score of 0.59 compared with 0.41 for the closest competing method. We then further improve performance using an incremental learning approach, in which failures in the original network are used to generate harder negative training examples. Results of this process raise the precision of the network, and improve the Dice score to 0.66. Our proposed method requires no user interaction, is fully automatic, and identifies large and fine root material throughout the whole volume. The 3D segmented output of our method is well-connected, allowing the recovery of structured representations of root system architecture, and so may be successfully utilised in root phenotyping.

scholarly journals Root CT Segmentation Using Incremental Learning Methodology on Improved Multiple Resolution Images

2021 ◽  
Vol 3 (4) ◽  
pp. 347-356
Author(s):  
K. Geetha
Keyword(s):  
Receptive Field ◽  
Incremental Learning ◽  
User Interaction ◽  
Learning Approach ◽  
Analysis Tool ◽  
Entire Volume ◽  
Efficient System ◽  
Volumetric Images ◽  
X Ray Computed ◽  
Ct Segmentation

The real-time issue of reliability segmenting root structure while using X-Ray Computed Tomography (CT) images is addressed in this work. A deep learning approach is proposed using a novel framework, involving decoders and encoders. The encoders-decoders framework is useful to improve multiple resolution by means of upsampling and downsampling images. The methodology of the work is enhanced by incorporating network branches with individual tasks using low-resolution context information and high-resolution segmentation. In large volumetric images, it is possible to resolve small root details by implementing a memory efficient system, resulting in the formation of a complete network. The proposed work, recent image analysis tool developed for root CT segmented is compared with several other previously existing methodology and it is found that this methodology is more efficient. Quantitatively and qualitatively, it is found that a multiresolution approach provides high accuracy in a shallower network with a large receptive field or deep network in a small receptive field. An incremental learning approach is also embedded to enhance the performance of the system. Moreover, it is also capable of detecting fine and large root materials in the entire volume. The proposed work is fully automated and doesn’t require user interaction.

Dendritic impulse collisions and shifting sites of action potential initiation contract and extend the receptive field of an amacrine cell

2007 ◽  
Vol 24 (4) ◽  
pp. 619-634 ◽  
Author(s):  
AUDREY S. ROYER ◽  
ROBERT F. MILLER
Keyword(s):  
Receptive Field ◽  
Amacrine Cell ◽  
Three Dimensional ◽  
Dendritic Tree ◽  
Amacrine Cells ◽  
Mechanistic Explanation ◽  
Field Size ◽  
Dimensional Structure ◽  
Na Channels ◽  
Sites Of Action

We evaluated the contributions of somatic and dendritic impulses to the receptive field dimensions of amacrine cells in the amphibian retina. For this analysis, we used the NEURON simulation program with a multicompartmental, multichannel model of an On-Off amacrine cell with a three-dimensional structure obtained through computer tracing techniques. Simulated synaptic inputs were evenly spaced along the dendritic branches and organized into eight annuli of increasing radius. The first set of simulations activated each ring progressively to simulate an area summation experiment, while a second approach activated each annulus individually. Both sets of simulations were done with and without the presence of Na channels in the dendrites and soma. Unexpectedly, the receptive field dimensions observed in the area summation simulations was often smaller than that predicted from the summation of the annular simulations. Collisions of action potentials moving in opposite directions in the dendrites largely accounted for this contraction in receptive field size for the area summation studies. The presence of dendritic Na channels increased the size of the receptive field beyond that achieved in their absence and allowed the physiological size of the receptive field to approximate the physical dimensions of the dendritic tree. This receptive field augmentation was the result of impulse generating ability in the dendrites which enhanced the signal observed at the soma. These simulations provide a plausible mechanistic explanation for physiological recordings from amacrine cells that show similar phenomena.

scholarly journals Graph Attention Feature Fusion Network for ALS Point Cloud Classification

Sensors ◽  
2021 ◽  
Vol 21 (18) ◽  
pp. 6193
Author(s):  
Jie Yang ◽  
Xinchang Zhang ◽  
Yun Huang
Keyword(s):  
Receptive Field ◽  
Point Cloud ◽  
Laser Scanning ◽  
Feature Fusion ◽  
Contextual Information ◽  
Point Clouds ◽  
Classification Performance ◽  
Field Size ◽  
Semantic Features ◽  
Sampling Cost

Classification is a fundamental task for airborne laser scanning (ALS) point cloud processing and applications. This task is challenging due to outdoor scenes with high complexity and point clouds with irregular distribution. Many existing methods based on deep learning techniques have drawbacks, such as complex pre/post-processing steps, an expensive sampling cost, and a limited receptive field size. In this paper, we propose a graph attention feature fusion network (GAFFNet) that can achieve a satisfactory classification performance by capturing wider contextual information of the ALS point cloud. Based on the graph attention mechanism, we first design a neighborhood feature fusion unit and an extended neighborhood feature fusion block, which effectively increases the receptive field for each point. On this basis, we further design a neural network based on encoder–decoder architecture to obtain the semantic features of point clouds at different levels, allowing us to achieve a more accurate classification. We evaluate the performance of our method on a publicly available ALS point cloud dataset provided by the International Society for Photogrammetry and Remote Sensing (ISPRS). The experimental results show that our method can effectively distinguish nine types of ground objects. We achieve more satisfactory results on different evaluation metrics when compared with the results obtained via other approaches.

scholarly journals Visual performance in behaving cats after prenatal unilateral enucleation

1993 ◽  
Vol 90 (23) ◽  
pp. 11142-11146 ◽  
Author(s):  
S Bisti ◽  
C Trimarchi
Keyword(s):  
Visual Acuity ◽  
Visual Field ◽  
Receptive Field ◽  
Receptive Fields ◽  
Direction Selectivity ◽  
Visual Performance ◽  
Field Size ◽  
Electrophysiological Recordings ◽  
Severe Impairment ◽  
Orientation Columns

Prenatal unilateral enucleation in mammals causes an extensive anatomical reorganization of visual pathways. The remaining eye innervates the entire extent of visual subcortical and cortical areas. Electrophysiological recordings have shown that the retino-geniculate connections are retinotopically organized and geniculate neurones have normal receptive field properties. In area 17 all neurons respond to stimulation of the remaining eye and retinotopy, orientation columns, and direction selectivity are maintained. The only detectable change is a reduction in receptive field size. Are these changes reflected in the visual behavior? We studied visual performance in cats unilaterally enucleated 3 weeks before birth (gestational age at enucleation, 39-42 days). We tested behaviorally the development of visual acuity and, in the adult, the extension of the visual field and the contrast sensitivity. We found no difference between prenatal monocularly enucleated cats and controls in their ability to orient to targets in different positions of the visual field or in their visual acuity (at any age). The major difference between enucleated and control animals was in contrast sensitivity:prenatal enucleated cats present a loss in sensitivity for gratings of low spatial frequency (below 0.5 cycle per degree) as well as a slight increase in sensitivity at middle frequencies. We conclude that prenatal unilateral enucleation causes a selective change in the spatial performance of the remaining eye. We suggest that this change is the result of a reduction in the number of neurones with large receptive fields, possibly due to a severe impairment of the Y system.

STUDY ON THE OPTIMIZATION FOR THE SIZE OF FUEL CELL FLOW FIELD UNDER INADEQUATE AIR SUPPLY CONDITION

Química Nova ◽  
2021 ◽  
Author(s):  
Shi Lei ◽  
Zheng Minggang
Keyword(s):  
Fuel Cell ◽  
Flow Field ◽  
Proton Exchange Membrane ◽  
Three Dimensional ◽  
Proton Exchange ◽  
Field Size ◽  
Air Supply ◽  
Supply Condition ◽  
Length Width ◽  
Exchange Membrane

In this paper, the influence of the optimization for flow field size on the proton exchange membrane fuel cell (PEMFC) performance under the inadequate air supply of cathode was studied based on the three-dimensional, steady-state, and constant temperature PEMFC monomer model. Additionally, the effect of the optimization for hybrid factors, including length, width, depth and width-depth, on the PEMFC performance was also investigated. The results showed that the optimization of the flow field size can improve the performance of the PEMFC and ensure that it is close to the level under the normal gas supply.

An Automated CFD Design and Analysis Tool for Inlet-Bleed Systems

1998 ◽  
Author(s):  
Tom I-P. Shih ◽  
Yu-Liang Lin ◽  
Andrew J. Flores ◽  
Mark A. Stephens ◽  
Mark J. Rimlinger ◽  
...  
Keyword(s):  
Boundary Layer ◽  
Three Dimensional ◽  
Incident Shock ◽  
Navier Stokes ◽  
Oblique Shock Wave ◽  
Design Parameters ◽  
Analysis Tool ◽  
Cfd Analysis ◽  
Circular Holes ◽  
Embedded Grids

Abstract A pre-processor was developed to assist CFD experts and non-experts in performing steady, three-dimensional Navier-Stokes analysis of a class of inlet-bleed problems involving oblique shock-wave/ boundary-layer interactions on a flat plate with bleed into a plenum through rows of circular holes. With this pre-processor, once geometry (e.g., hole dimensions and arrangement) and flow conditions (e.g., Mach number, boundary-layer thickness, incident shock location) are inputted, it will automatically generate every file needed to perform a CFD analysis from the grid system to initial and boundary conditions. This is accomplished by accessing a knowledge base established by experts who understand both CFD and the class of problems being analyzed. For experts in CFD, this tool greatly reduces the amount of time and effort needed to setup a problem for CFD analysis. It also provides experts with knobs to make changes to the setup if desired. For non-experts in CFD, this tool enables reliable and correct usage of CFD. A typical session on a workstation from data input to the generation of all files needed to perform a CFD analysis involves less than ten minutes. This pre-processor, referred to as AUTOMAT-V2, is an improved version of a code called AUTOMAT. Improvements made include: (1) multi-block structured grids can be patched in addition to being overlapped; (2) embedded grids can be introduced near bleed holes to reduce the number of grid points/cells needed by a factor of up to four; (3) grid systems generated allow up to three levels of multigrid; (4) CFL3D is supported in addition to OVERFLOW, two well-known and highly regarded Navier-Stokes solvers developed at NASA’s Langley and Ames Research Centers; (5) all files needed to run RONNIE for patched grids and MAGGIE for overlapped grids are also generated; and (6) more design parameters can be investigated including the study of micro bleed and effects of flow/hole misalignments.

Receptive Field Size and Response Latency Are Correlated Within the Cat Visual Thalamus

2005 ◽  
Vol 93 (6) ◽  
pp. 3537-3547 ◽  
Author(s):  
Chong Weng ◽  
Chun-I Yeh ◽  
Carl R. Stoelzel ◽  
Jose-Manuel Alonso
Keyword(s):  
Receptive Field ◽  
Spatial Frequency ◽  
Response Latency ◽  
Time Course ◽  
Frequency Tuning ◽  
Receptive Fields ◽  
Cortical Cell ◽  
Field Size ◽  
Receptive Field Size ◽  
Spatial Frequency Tuning

Each point in visual space is encoded at the level of the thalamus by a group of neighboring cells with overlapping receptive fields. Here we show that the receptive fields of these cells differ in size and response latency but not at random. We have found that in the cat lateral geniculate nucleus (LGN) the receptive field size and response latency of neighboring neurons are significantly correlated: the larger the receptive field, the faster the response to visual stimuli. This correlation is widespread in LGN. It is found in groups of cells belonging to the same type (e.g., Y cells), and of different types (i.e., X and Y), within a specific layer or across different layers. These results indicate that the inputs from the multiple geniculate afferents that converge onto a cortical cell (approximately 30) are likely to arrive in a sequence determined by the receptive field size of the geniculate afferents. Recent studies have shown that the peak of the spatial frequency tuning of a cortical cell shifts toward higher frequencies as the response progresses in time. Our results are consistent with the idea that these shifts in spatial frequency tuning arise from differences in the response time course of the thalamic inputs.

Changes in the response properties of neurons in the ventroposterior lateral thalamic nucleus of the raccoon after peripheral deafferentation

1996 ◽  
Vol 75 (6) ◽  
pp. 2441-2450 ◽  
Author(s):  
D. D. Rasmusson
Keyword(s):  
Receptive Field ◽  
Thalamic Nucleus ◽  
Receptive Fields ◽  
Field Size ◽  
Mechanical Stimuli ◽  
Sensory Pathways ◽  
Average Latency ◽  
Somatotopic Map ◽  
Almost All ◽  
Stimulation Of

1. Single neurons in the ventroposterior lateral thalamic nucleus were studied in 10 anesthetized raccoons, 4 of which had undergone amputation of the fourth digit 4-5 mo before recording. Neurons with receptive fields on the glabrous skin of a forepaw digit were examined in response to electrical stimulation of the “on-focus” digit that contained the neuron's receptive field and stimulation of an adjacent, “off-focus” digit. 2. In normal raccoons all neurons responded to on-focus stimulation with an excitation at a short latency (mean 13 ms), whereas only 63% of the neurons responded to off-focus digit stimulation. The off-focus responses had a longer latency (mean 27.2 ms) and a higher threshold than the on-focus responses (800 and 452 microA, respectively). Only 3 of 32 neurons tested with off-focus stimulation had both a latency and a threshold within the range of on-focus values. Inhibition following the excitation was seen in the majority of neurons with both types of stimulation. 3. In the raccoons with digit removal, the region of the thalamus that had lost its major peripheral input (the “deafferented” region) was distinguished from the normal third and fifth digit regions on the basis of the sequence of neuronal receptive fields within a penetration and receptive field size as described previously. 4. Almost all of the neurons in the deafferented region (91%) were excited by stimulation of one or both adjacent digits. The average latency for these responses was shorter (15.3 ms) and the threshold was lower than was the case with off-focus stimulation in control animals. These values were not significantly different from the responses to on-focus stimulation in the animals with digit amputation. 5. These results confirm that reorganization of sensory pathways can be observed at the thalamic level. In addition to the changes in the somatotopic map that have been shown previously with the use of mechanical stimuli, the present paper demonstrates an improvement in several quantitative measures of single-unit responses. Many of these changes suggest that this reorganization could be explained by an increased effectiveness of preexisting, weak connections from the off-focus digits; however, the increase in the proportion of neurons responding to stimulation of adjacent digits may indicate that sprouting of new connections also occurs.

Sign in / Sign up

Export Citation Format

Share Document