Neural Input Research Articles

Radiologist-Level Performance by Using Deep Learning for Segmentation of Breast Cancers on MRI Scans.

PurposeTo develop a deep network architecture that would achieve fully automated radiologist-level segmentation of cancers at breast MRI.Materials and MethodsIn this retrospective study, 38 229 examinations (composed of 64 063 individual breast scans from 14 475 patients) were performed in female patients (age range, 12–94 years; mean age, 52 years ± 10 [standard deviation]) who presented between 2002 and 2014 at a single clinical site. A total of 2555 breast cancers were selected that had been segmented on two-dimensional (2D) images by radiologists, as well as 60 108 benign breasts that served as examples of noncancerous tissue; all these were used for model training. For testing, an additional 250 breast cancers were segmented independently on 2D images by four radiologists. Authors selected among several three-dimensional (3D) deep convolutional neural network architectures, input modalities, and harmonization methods. The outcome measure was the Dice score for 2D segmentation, which was compared between the network and radiologists by using the Wilcoxon signed rank test and the two one-sided test procedure.ResultsThe highest-performing network on the training set was a 3D U-Net with dynamic contrast-enhanced MRI as input and with intensity normalized for each examination. In the test set, the median Dice score of this network was 0.77 (interquartile range, 0.26). The performance of the network was equivalent to that of the radiologists (two one-sided test procedures with radiologist performance of 0.69–0.84 as equivalence bounds, P < .001 for both; n = 250).ConclusionWhen trained on a sufficiently large dataset, the developed 3D U-Net performed as well as fellowship-trained radiologists in detailed 2D segmentation of breast cancers at routine clinical MRI.Keywords: MRI, Breast, Segmentation, Supervised Learning, Convolutional Neural Network (CNN), Deep Learning Algorithms, Machine Learning AlgorithmsPublished under a CC BY 4.0 license. Supplemental material is available for this article.

Synaptic Gain Abnormalities in Schizophrenia and the Potential Relevance for Cognition

SEE CORRESPONDING ARTICLE ON PAGE 202 Computational Modeling of Electroencephalography and Functional Magnetic Resonance Imaging Paradigms Indicates a Consistent Loss of Pyramidal Cell Synaptic Gain in SchizophreniaBiological PsychiatryVol. 91Issue 2PreviewDiminished synaptic gain—the sensitivity of postsynaptic responses to neural inputs—may be a fundamental synaptic pathology in schizophrenia. Evidence for this is indirect, however. Furthermore, it is unclear whether pyramidal cells or interneurons (or both) are affected, or how these deficits relate to symptoms. Full-Text PDF Open Access

First break picking with deep learning – evaluation of network architectures

ABSTRACTIn recent years, various convolutional neural network architectures have been proposed for first break picking. In this paper, we compare the standard auto‐encoder and U‐net architectures as well as versions enhanced with ResNet style skip connections. The U‐net appears to have become the standard network for segmentation, judging from the number of published articles. Still, there is some variety in neural network architectural choices. In this paper, we assess the impact of neural network depth, width and input data size, as well as some small modifications for deep networks offered by the ResNet. In general, results improve as the networks get deeper, but with diminishing returns. The more complex the data, the more benefit the deeper networks bring. We use complete shot gathers, albeit rescaled for efficiency, to train the neural networks. For shot gathers with a simple piecewise linear moveout, this approach yields results with good accuracy when gathers are resampled to 128 × 128 samples. For shot gathers with more complex first break moveout, using our approach it is advised to stay close to the original dimension of each gather for best accuracy, at the expense of increased training times. A good trade‐off between network depth, image size and training times is to use a nine‐stage U‐net with 256 sample images. Despite the advantages in other applications, the basic U‐net outperforms a U‐net with ResNet features. We show that changing the input data dimensions for trained networks does not work, despite the fact the fully convolutional networks are independent of image size. The U‐net based first break picking is not sensitive to picking errors, as in many cases the neural network predictions are better than the training data where the training data have random mispicks. This suggests a practical application; namely, to train or re‐train a pre‐trained network on a single data set after conventional first break picking with the objective of improving conventionally picked first breaks.

Using artificial intelligence to predict the intrauterine insemination success rate among infertile couples

BackgroundTo evaluate the use of artificial intelligence (AI) in predicting the success rate of intrauterine insemination (IUI) treatment among infertile couples and also to determine the importance of each of the parameters affecting IUI success. This study was a retrospective cohort study in which information from 380 infertile couples undergoing IUI treatment (190 cases resulting in positive pregnancy test and 190 cases of failed IUI) including underlying factors, female factors, sperm parameters at the beginning of the treatment cycle, and fertility results were collected from 2013 to 2019 and evaluated to determine the effectiveness of AI in predicting IUI success.ResultsWe used the most important factors influencing the success of IUI as a neural network input. With the help of a three-layer neural network, the accuracy of the AI to predict the success rate of IUI was 71.92% and the sensitivity and specificity were 76.19% and 66.67%, respectively. The effect of each of the predictive factors was obtained by calculating the ROC curve and determining the cut-off point.ConclusionsThe morphology, total motility, and progressive motility of the sperm were found to be the most important predictive factors for IUI success. In this study, we concluded that by predicting IUI success rate, artificial intelligence can help clinicians choose individualized treatment for infertile couples and to shorten the time to pregnancy.

Differential Effects of Neurectomy and Botox-induced Muscle Paralysis on Bone Phenotype and Titanium Implant Osseointegration.

Metabolic bone is highly innervated by both sensory and sympathetic nerves. In addition to skeletal development, neural regulation participates in local bone remodeling, which is important for successful osseointegration of titanium implants. Neurectomy is a model used to investigate the lack of neural function on bone homeostasis, but the relative impacts of direct denervation to bone or denervation-induced muscle paralysis are less well defined. To investigate this difference, we used two nerve intervention models, sciatic and femoral neurectomy (SFN) v. botox-induced muscle paralysis (BTX) and assessed the resulting femoral bone phenotype and Ti implant osseointegration. Male Sprague Dawley rats (19) were randomly divided into three groups: implant control (n=5), SFN (n=7), and BTX (n=7). Ti implants (microrough/hydrophilic [modSLA], Institut Straumann AG) were placed in the distal metaphysis of each femur on day 24 post-SFN or BTX. Bone and muscle were examined on day 28 after implant insertion. Both nerve intervention models impaired osseointegration. MicroCT and histology indicated that both models had reduced trabecular bone formation. Only BTX reduced cortical bone formation and increased cortical bone porosity. BTX resulted in more bone loss characterized by the least trabecular and cortical bone, as well as osseointegration. Osteoblasts isolated from the tibia exhibited a model-specific phenotype when they were grown on Ti substrates in vitro. Neurectomy caused more severe muscle atrophy than botox injection. These results indicate that neural regulation directly modulates bone formation and osseointegration. Muscle paralysis modulated the effects of loss of neural inputs into bone, supporting the hypothesis that mechanical loading of bone is a factor in achieving successful osseointegration. The different effects of botox and neurectomy on bone phenotype indicated that the sensory and sympathetic nerves had a role in the osseointegration process.

Tunicate swarm algorithm-trained multi-layered perceptron for data centre energy demand forecasting and relative percentage contribution analysis of input parameters

PurposeEnergy management is critical to data centres (DCs) majorly because they are high energy-consuming facilities and demand for their services continue to rise due to rapidly increasing global demand for cloud services and other technological services. This projected sectoral growth is expected to translate into increased energy demand from the sector, which is already considered a major energy consumer unless innovative steps are used to drive effective energy management systems. The purpose of this study is to provide insights into the expected energy demand of the DC and the impact each measured parameter has on the building's energy demand profile. This serves as a basis for the design of an effective energy management system.Design/methodology/approachThis study proposes novel tunicate swarm algorithm (TSA) for training an artificial neural network model used for predicting the energy demand of a DC. The objective is to find the optimal weights and biases of the model while avoiding commonly faced challenges when using the backpropagation algorithm. The model implementation is based on historical energy consumption data of an anonymous DC operator in Cape Town, South Africa. The data set provided consists of variables such as ambient temperature, ambient relative humidity, chiller output temperature and computer room air conditioning air supply temperature, which serve as inputs to the neural network that is designed to predict the DC’s hourly energy consumption for July 2020. Upon preprocessing of the data set, total sample number for each represented variable was 464. The 80:20 splitting ratio was used to divide the data set into training and testing set respectively, making 452 samples for the training set and 112 samples for the testing set. A weights-based approach has also been used to analyze the relative impact of the model’s input parameters on the DC’s energy demand pattern.FindingsThe performance of the proposed model has been compared with those of neural network models trained using state of the art algorithms such as moth flame optimization, whale optimization algorithm and ant lion optimizer. From analysis, it was found that the proposed TSA outperformed the other methods in training the model based on their mean squared error, root mean squared error, mean absolute error, mean absolute percentage error and prediction accuracy. Analyzing the relative percentage contribution of the model's input parameters based on the weights of the neural network also shows that the ambient temperature of the DC has the highest impact on the building’s energy demand pattern.Research limitations/implicationsThe proposed novel model can be applied to solving other complex engineering problems such as regression and classification. The methodology for optimizing the multi-layered perceptron neural network can also be further applied to other forms of neural networks for improved performance.Practical implicationsBased on the forecasted energy demand of the DC and an understanding of how the input parameters impact the building's energy demand pattern, neural networks can be deployed to optimize the cooling systems of the DC for reduced energy cost.Originality/valueThe use of TSA for optimizing the weights and biases of a neural network is a novel study. The application context of this study which is DCs is quite untapped in the literature, leaving many gaps for further research. The proposed prediction model can be further applied to other regression tasks and classification tasks. Another contribution of this study is the analysis of the neural network's input parameters, which provides insight into the level to which each parameter influences the DC’s energy demand profile.

A Spatial Model of Hepatic Calcium Signaling and Glucose Metabolism Under Autonomic Control Reveals Functional Consequences of Varying Liver Innervation Patterns Across Species.

Rapid breakdown of hepatic glycogen stores into glucose plays an important role during intense physical exercise to maintain systemic euglycemia. Hepatic glycogenolysis is governed by several different liver-intrinsic and systemic factors such as hepatic zonation, circulating catecholamines, hepatocellular calcium signaling, hepatic neuroanatomy, and the central nervous system (CNS). Of the factors regulating hepatic glycogenolysis, the extent of lobular innervation varies significantly between humans and rodents. While rodents display very few autonomic nerve terminals in the liver, nearly every hepatic layer in the human liver receives neural input. In the present study, we developed a multi-scale, multi-organ model of hepatic metabolism incorporating liver zonation, lobular scale calcium signaling, hepatic innervation, and direct and peripheral organ-mediated communication between the liver and the CNS. We evaluated the effect of each of these governing factors on the total hepatic glucose output and zonal glycogenolytic patterns within liver lobules during simulated physical exercise. Our simulations revealed that direct neuronal stimulation of the liver and an increase in circulating catecholamines increases hepatic glucose output mediated by mobilization of intracellular calcium stores and lobular scale calcium waves. Comparing simulated glycogenolysis between human-like and rodent-like hepatic innervation patterns (extensive vs. minimal) suggested that propagation of calcium transients across liver lobules acts as a compensatory mechanism to improve hepatic glucose output in sparsely innervated livers. Interestingly, our simulations suggested that catecholamine-driven glycogenolysis is reduced under portal hypertension. However, increased innervation coupled with strong intercellular communication can improve the total hepatic glucose output under portal hypertension. In summary, our modeling and simulation study reveals a complex interplay of intercellular and multi-organ interactions that can lead to differing calcium dynamics and spatial distributions of glycogenolysis at the lobular scale in the liver.

Application of Project Whole Process Cost Control in Construction Project Cost Audit

This paper summarizes the theory of project cost, which paves the way for the basic theoretical system of construction project cost estimation. This paper expounds the function of project cost and the main factors affecting project cost. Secondly, the basic principles of BP neural network method and grey theory are described, which provides technical support for the establishment of construction cost estimation system model. In view of the shortcomings of BP neural network, such as slow convergence speed, easy to fall into local minimum and inaccurate prediction, this paper proposes an improved method to process the data of BP neural network input layer with grey one-time accumulation, and then use grey one-time subtraction to process the output layer. Finally, the optimization model based on grey BP neural network method is established to establish a more accurate knowledge framework system in order to solve the construction cost estimation.

Predictive Analysis and Evaluation Model of Chronic Liver Disease Based on BP Neural Network with Improved Ant Colony Algorithm.

Timely prediction of the mechanism and characteristics of chronic liver disease using next-generation information technology is an effective way to improve the diagnosis rate of chronic liver disease. In this paper, we have proposed a modified backpropagation (BP) neural network with improved ant colony optimization algorithm to process multiple index attribute items describing chronic liver disease and construct a chronic liver disease assessment model. The proposed model is very effective in detecting chronic liver disease on time with acceptable level of accuracy and precision ratio. To verify these claims, the proposed scheme is checked experimentally where 125 groups of 20-dimensional medical test index data items of patients with chronic liver disease were analyzed. Moreover, 13-dimensional index items were preferentially selected as test index attribute items with high sensitivity to chronic liver disease using well-known ROC curves. The 13-dimensional index items were reduced to 5-dimensional comprehensive data items by principal component analysis. The proposed neural network-based model was trained with 115 sets of test indicator sample sets, and the remaining 10 sets of sample sets were used as test samples. Compared with the original 20-dimensional data as the neural network input, the proposed model not only reduces the complexity but also improves the prediction accuracy by 15.07%.

A data-driven approach to predicting the anisotropic mechanical behaviour of voided single crystals

Machine learning techniques are increasingly used to extract important physical information from a broad range of materials and to identify their processing-structure–property relationships. In this work, a neural network framework is coupled with a crystal plasticity finite element based approach to predict the deformation and ductile failure behaviour of porous FCC single crystals when subjected to multi-axial loading conditions. The work relies on 3D unit cells with a centrally located spherical void to represent the microstructure of the porous single crystal material, and on a crystallographic slip-based crystal plasticity constitutive model to describe its deformation behaviour. Stress–strain data generated by unit cell finite element simulations are relied upon to construct the neural network model. Different strategies for the neural network input and output variables and parameters are first explored so as to optimise performance and accuracy. Both proportional and non-proportional loading conditions resulting from a constant and a varying stress triaxiality during deformation, respectively, are considered. The optimum neural network strategy is shown to successfully predict the behaviour of the porous single crystal under both proportional and non-proportional loading, albeit with the void behaviour at high stress triaxialities better described than that at low triaxialities. The results also reveal that the use of tensorial quantities for both stresses and strains as input and output neural network quantities is more suitable as a form of data representation for multiaxial loading conditions than uniaxial equivalent stress and strain quantities. It was also found that the inclusion of prior knowledge as neural network input quantities in the form of reference stress–strain solutions for the void-free single crystal considerably improves the predictive capabilities of the proposed data-driven approach, even when only a very limited number of training cases was used.

Common Neural Input within and across Lower Limb Muscles: A Preliminary Study.

Motor units (MUs) are the basic unit of motor control. MU synchronization has been evaluated to identify common inputs in neural circuitry during motor coordination. Recent studies have compared common inputs between muscles in the lower limb, but further investigation is needed to compare common inputs to MUs both within a muscle and between MUs of different muscle pairs. The goal of this preliminary study was to characterize levels of common inputs to MUs in three muscle groups: MUs within a muscle, between bilateral homologous pairs, and between agonist/antagonist muscle pairs. To achieve this, surface electromyography (EMG) was recorded during bilateral ankle dorsiflexion and plantarflexion on the right and left tibiales anterior (RTA, LTA) and gastrocnemii (RGA, LGA) muscles. After decomposing EMG into active MU firings, we conducted coherence analyses of composite MU spike trains (CSTs) in each muscle group in both the beta (13-30 Hz) and gamma (30-60 Hz) frequency bands. Our results indicate MUs within a muscle have the greatest levels of common input, with decreasing levels of common input to bilateral and agonist/antagonist muscle pairs, respectively. Additionally, each muscle group exhibited similar levels of common input between the beta and gamma bands. This work may provide a way to unveil mechanisms of functional coordination in the lower limb across motor tasks.

Entrainment of Astrocytic and Neuronal Ca2+ Population Dynamics During Information Processing of Working Memory in Mice.

Astrocytes are increasingly recognized to play an active role in learning and memory, but whether neural inputs can trigger event-specific astrocytic Ca2+ dynamics in real time to participate in working memory remains unclear due to the difficulties in directly monitoring astrocytic Ca2+ dynamics in animals performing tasks. Here, using fiber photometry, we showed that population astrocytic Ca2+ dynamics in the hippocampus were gated by sensory inputs (centered at the turning point of the T-maze) and modified by the reward delivery during the encoding and retrieval phases. Notably, there was a strong inter-locked and antagonistic relationship between the astrocytic and neuronal Ca2+ dynamics with a 3-s phase difference. Furthermore, there was a robust synchronization of astrocytic Ca2+ at the population level among the hippocampus, medial prefrontal cortex, and striatum. The inter-locked, bidirectional communication between astrocytes and neurons at the population level may contribute to the modulation of information processing in working memory.

Self-Organization in Vocal Mechanics and Physiology

Self-Organization in Vocal Mechanics and Physiology Ingo R. Titze (bio) VOICE RESEARCH AND TECHNOLOGY Self-organization is now an accepted concept in biological systems. Complex systems with many interactive components do not need microcontrol of each component to produce a well coordinated output of the system. If given enough individual freedom, the components will interact in such a way as to produce an organization for optimum efficiency. Contrary to classical thinking, regularity can be produced with much variability in the system, or even irregularity in the input. Heartbeat is a good example. The neural inputs to the heart muscles are not as regular as the heartbeat itself. The muscles self-organize for their collective output. In the old way of thinking, noise-in always meant noise-out. In the new way of thinking, some added noise can actually help in self-organization. It gives the system more choices for possible coordinated activity. Feedback and ample exchange of energy between components is a requirement. Feedback can be positive or negative. Positive feedback reinforces the output of a component by taking a portion of the output and adding it to the input in phase. Negative feedback subtracts some of the output from the input. With the two forms of feedback, many forms of interaction are possible between subsystems. Consider some examples in vocalization. The subsystems can be the left vocal fold, the right vocal fold, the subglottal resonator, and the supraglottal resonator. Each has its own natural resonance properties. They all can be thought of as oscillators. Energy transfer is facilitated by glottal aerodynamics and acoustic wave propagation in the airways. Without energy transfer and feedback between the subsystems, the components all would likely have their own frequency, amplitude, and phase in their oscillatory patterns. The output would not be periodic (harmonic) unless each is controlled with precision. With interaction, however, they can synchronize their motions and the forces that keep them in motion. No separate control is needed. However, synchronization is more likely if the subsystems have great flexibility, a wide range of operation. So, what is the take-home message for singers? To obtain regularity and predictability in our vocal output, we need to resist the tendency to restrict the range of operation of some of the components. The theory suggests that we not limit the range of tongue or lip movements, or the range of lung pressure, or the choice of vowels, or the choice of larynx position. Allowing for maximal self-organization would speak against such restrictions. Focusing on the desired output and its repeatability pays the higher dividends. If the system [End Page 233] knows what to produce, it will use its internal variability to produce it. We should not constrain the interaction between the system components. This is especially the case if there is a deficiency or abnormality in one or more of the components. The rest of the system has to compensate, which means that the range of operation and mobility of some components has to increase. In summary, the message is: Let the vocal system organize itself. Give it the right output targets and it will find a way to get there with lots of alternative solutions. [Dr. Titze's bio is republished here because his updated version regrettably was omitted in the last issue.] [End Page 234] Ingo R. Titze Ingo R. Titze, PhD, is Emeritus Distinguished Professor of Speech Science and Voice at the University of Iowa and Executive Director of the National Center for Voice and Speech at the University of Utah. His formal education is in physics and electrical engineering, but he has devoted much of his studies to vocal music and speech. Dr. Titze has published more than 500 articles in scientific and educational journals, coedited two books titled Vocal Fold Physiology, and now has four books in print: Principles of Voice Production, The Myoelastic Aerodynamic Theory of Phonotion, Vocology: The Science and Practice of Voice Habilitation, and Fascinations with the Human Voice. He has lectured throughout the world and has appeared on such educational television series as Innovation, Quantum, and Beyond 2000. He is a recipient of the William and Harriott Gould Award for laryngeal physiology...

Flexible filtering by neural inputs supports motion computation across states and stimuli

Sensory systems flexibly adapt their processing properties across a wide range of environmental and behavioral conditions. Such variable processing complicates attempts to extract a mechanistic understanding of sensory computations. This is evident in the highly constrained, canonical Drosophila motion detection circuit, where the core computation underlying direction selectivity is still debated despite extensive studies. Here we measured the filtering properties of neural inputs to the OFF motion-detecting T5 cell in Drosophila. We report state- and stimulus-dependent changes in the shape of these signals, which become more biphasic under specific conditions. Summing these inputs within the framework of a connectomic-constrained model of the circuit demonstrates that these shapes are sufficient to explain T5 responses to various motion stimuli. Thus, our stimulus- and state-dependent measurements reconcile motion computation with the anatomy of the circuit. These findings provide a clear example of how a basic circuit supports flexible sensory computation.

A DSC approach to adaptive dynamic region‐based tracking control for strict‐feedback non‐linear systems

As an extension of the conventional set-point control problem, the dynamic region-based tracking control scheme with obstacle avoidance is proposed for a class of uncertain strict-feedback non-linear systems. A novel adaptive tracking controller is designed by a fusion of artificial potential field, recursive backstepping approach, neural networks, dynamic surface control technique, calculus method, and Lyapunov stability theory. In the proposed control scheme, the objective region cannot be required to have a regular shape or a fixed size for the passibility of the system in constrained space. The region tracking error is transformed into a new virtual error variable for recursively designing a dynamic surface controller, and the dimension of neural network inputs can be greatly reduced, especially for high-order systems. The Lyapunov theorem is used to confirm the stability and uniform boundedness of all closed-loop signals. Simulation results are provided to demonstrate the effectiveness of the proposed controller.

Behavior of motor units during submaximal isometric contractions in chronically strength-trained individuals.

Neural and morphological adaptations combine to underpin the enhanced muscle strength following prolonged exposure to strength training, although their relative importance remains unclear. We investigated the contribution of motor unit (MU) behavior and muscle size to submaximal force production in chronically strength-trained athletes (ST) versus untrained controls (UT). Sixteen ST (age: 22.9 ± 3.5 yr; training experience: 5.9 ± 3.5 yr) and 14 UT (age: 20.4 ± 2.3 yr) performed maximal voluntary isometric force (MViF) and ramp contractions (at 15%, 35%, 50%, and 70% MViF) with elbow flexors, whilst high-density surface electromyography (HDsEMG) was recorded from the biceps brachii (BB). Recruitment thresholds (RTs) and discharge rates (DRs) of MUs identified from the submaximal contractions were assessed. The neural drive-to-muscle gain was estimated from the relation between changes in force (ΔFORCE, i.e. muscle output) relative to changes in MU DR (ΔDR, i.e. neural input). BB maximum anatomical cross-sectional area (ACSAMAX) was also assessed by MRI. MViF (+64.8% vs. UT, P < 0.001) and BB ACSAMAX (+71.9%, P < 0.001) were higher in ST. Absolute MU RT was higher in ST (+62.6%, P < 0.001), but occurred at similar normalized forces. MU DR did not differ between groups at the same normalized forces. The absolute slope of the ΔFORCE - ΔDR relationship was higher in ST (+66.9%, P = 0.002), whereas it did not differ for normalized values. We observed similar MU behavior between ST athletes and UT controls. The greater absolute force-generating capacity of ST for the same neural input demonstrates that morphological, rather than neural, factors are the predominant mechanism for their enhanced force generation during submaximal efforts.NEW & NOTEWORTHY In this study, we observed that recruitment strategies and discharge characteristics of large populations of motor units identified from biceps brachii of strength-trained athletes were similar to those observed in untrained individuals during submaximal force tasks. We also found that for the same neural input, strength-trained athletes are able to produce greater absolute muscle forces (i.e., neural drive-to-muscle gain). This demonstrates that morphological factors are the predominant mechanism for the enhanced force generation during submaximal efforts.

Prediction of flood using optimized neural network with Gray wolf algorithm (Maroon River case study)

Floods, as one of the most frequent natural hazards, cause irreparable damage to infrastructure and agriculture, and housing every year. In order to avoid financial and human losses, the importance of flood forecasting seems inevitable. Considering that floods are caused by many natural and anthropogenic factors and also have limitations such as lack of complete information. In this study, artificial neural networks have been used as an efficient method for flood prediction. The neural network inputs include the Dubai River and the Eshel River, this data was collected over 8 Years from the Maroon River. The network used is a multilayer perceptron, also the neural network weights are optimized by the Gray wolf algorithm and the results are compared with other common methods. Analysis of the output results shows that the neural network with the Gray Wolf algorithm has better results than neural network and Genetic algorithms and the error of this method is 0.53%, which indicates high accuracy and precision for flood prediction compared to other evolutionary algorithms. This method is used to obtain the best amount of data for testing and training. As the results, the best rate is 80% for training and 20% for testing. Obtained results show the neural network error squares with 80% of the training data and 20% of the test data.

Toward Robotic Weed Control: Detection of Nutsedge Weed in Bermudagrass Turf Using Inaccurate and Insufficient Training Data

To enable robotic weed control, we develop algorithms to detect nutsedge weed from bermudagrass turf. Due to the similarity between the weed and the background turf, manual data labeling is expensive and error-prone. Consequently, directly applying deep learning methods for object detection cannot generate satisfactory results. Building on an instance detection approach (i.e. Mask R-CNN), we combine synthetic data with raw data to train the network. We propose an algorithm to generate high fidelity synthetic data, adopting different levels of annotations to reduce labeling cost. Moreover, we construct a nutsedge skeleton-based probabilistic map (NSPM) as the neural network input to reduce the reliance on pixel-wise precise labeling. We also modify loss function from cross entropy to Kullback-Leibler divergence which accommodates uncertainty in the labeling process. We implement the proposed algorithm and compare it with both Faster R-CNN and Mask R-CNN. The results show that our design can effectively overcome the impact of imprecise and insufficient training sample issues and significantly outperform the Faster R-CNN counterpart with a false negative rate of only 0.4%. In particular, our approach also reduces labeling time by 95% while achieving better performance if comparing with the original Mask R-CNN approach.

Novel intrinsic neurogenic and myogenic mechanisms underlying the formation of faecal pellets along the large intestine of guinea-pigs.

Soft faecal material is transformed into discrete, pellet-shaped faeces at the colonic flexure. Here, analysis of water content in natural faecal material revealed a decline from cecum to rectum without significant changes at the flexure. Thus, pellet formation is not explained by changes in viscosity alone. We then used video imaging of colonic wall movements with electromyography in isolated preparations containing guinea-pig proximal colon, colonic flexure and distal colon. To investigate the pellet formation process, the colonic segments were infused with artificial contents (Krebs solution and 4-6% methylcellulose) to simulate physiological faecal content flow. Remarkably, pellet formation took place in vitro, without extrinsic neural inputs. Infusion evoked slowly propagating neurogenic contractions, the proximal colon migrating motor complexes (∼0.6cpm), which initiated pellet formation at the flexure. Lesion of the flexure, but not the proximal colon, disrupted the formation of normal individual pellets. In addition, a distinct myogenic mechanism was identified, whereby slow phasic contractions (∼1.9cpm) initiated at the flexure and propagated short distances retrogradely into the proximal colon and antegradely into the distal colon. There were no detectable changes in the density or distribution of pacemaker-type interstitial cells of Cajal across the flexure. The findings provide new insights into how solid faecal content is generated, suggesting the major mechanisms underlying faecal pellet formation involve the unique interaction at the colonic flexure between antegrade proximal colon migrating motor complexes, organized by enteric neurons, and retrograde myogenic slow phasic contractions. Additional, as yet unidentified extrinsic and/or humoral influences appear to contribute to processing of faecal content in vivo. KEY POINTS: In herbivores, including guinea-pigs, clearly defined faecal pellets are formed at a distinct location along the large intestine (colonic flexure). The mechanism underlying the formation of these faecal pellets at this region has remained unknown. We reveal a progressive and gradual reduction in water content of faecal content along the bowel. Hence, the distinct transition from amorphous to pellet shaped faecal content could not be explained by a dramatic increase in water reabsorption from a specific site. We discovered patterns of anterograde neurogenic and retrograde myogenic motor activity that facilitate the formation of faecal pellets. The formation of 'pellet-like' boluses at the colonic flexure involves interaction of an antegrade migrating motor complex in the proximal colon and retrograde myogenic slow phasic contractions that emerge from the colonic flexure. The findings uncover intrinsic mechanisms responsible for the formation of discrete faecal scybala in the large intestine of a vertebrate.

Efferent and Afferent Connections of Neuropeptide Y Neurons in the Nucleus Accumbens of Mice

Neuropeptide Y (NPY) is a neural peptide distributed widely in the brain and has various functions in each region. We previously reported that NPY neurons in the nucleus accumbens (NAc) are involved in the regulation of anxiety behavior. Anterograde and retrograde tracing studies suggest that neurons in the NAc project to several areas, such as the lateral hypothalamus (LH) and ventral pallidum (VP), and receive afferent projections from the cortex, thalamus, and amygdala. However, the neural connections between accumbal NPY neurons and other brain areas in mice remain unclear. In this study, we sought to clarify these anatomical connections of NPY neurons in the NAc by investigating their neural outputs and inputs. To selectively map NPY neuronal efferents from the NAc, we injected Cre-dependent adeno-associated viruses (AAVs) into the NAc of NPY-Cre mice. This revealed that NAc NPY neurons exclusively projected to the LH. We confirmed this by injecting cholera toxin b subunit (CTb), a retrograde tracer, into the LH and found that approximately 7–10% of NPY neurons in the NAc were double-labeled for mCherry and CTb. Moreover, retrograde tracing using recombinant rabies virus (rRABV) also identified NAc NPY projections to the LH. Finally, we investigated monosynaptic input to the NPY neurons in the NAc using rRABV. We found that NPY neurons in the NAc received direct synaptic connections from the midline thalamic nuclei and posterior basomedial amygdala. These findings provide new insight into the neural networks of accumbal NPY neurons and should assist in elucidating their functional roles.