Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Networks”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 577 records · Page 32Linked to original sources
PubMed2006

Reserve networks based on richness hotspots and representation vary with scale.

While the importance of spatial scale in ecology is well established, few studies have investigated the impact of data grain on conservation planning outcomes. In this study, we compared species richness hotspot and representation networks developed at five grain sizes. We used species distribution maps for mammals and birds developed by the Arizona and New Mexico Gap Analysis Programs (GAP) to produce 1-km2, 100-kmn2, 625-km2, 2500-km2, and 10,000-km2 grid cell resolution distribution maps. We used these distribution maps to generate species richness and hotspot (95th quantile) maps for each taxon in each state. Species composition information at each grain size was used to develop two types of representation networks using the reserve selection software MARXAN. Reserve selection analyses were restricted to Arizona birds due to considerable computation requirements. We used MARXAN to create best reserve networks based on the minimum area required to represent each species at least once and equal area networks based on irreplaceability values. We also measured the median area of each species' distribution included in hotspot (mammals and birds of Arizona and New Mexico) and irreplaceability (Arizona birds) networks across all species. Mean area overlap between richness hotspot reserves identified at the five grain sizes was 29% (grand mean for four within-taxon/state comparisons), mean overlap for irreplaceability reserve networks was 32%, and mean overlap for best reserve networks was 53%. Hotspots for mammals and birds showed low overlap with a mean of 30%. Comparison of hotspots and irreplaceability networks showed very low overlap with a mean of 13%. For hotspots, median species distribution area protected within reserves declined monotonically from a high of 11% for 1-km2 networks down to 6% for 10,000-km2 networks. Irreplaceability networks showed a similar, but more variable, pattern of decline. This work clearly shows that map resolution has a profound effect on conservation planning outcomes and that hotspot and representation outcomes may be strikingly dissimilar. Thus, conservation planning is scale dependent, such that reserves developed using coarse-grained data do not subsume fine-grained reserves. Moreover, preserving both full species representation and species rich areas may require combined reserve design strategies.

Animals↗
PubMedDate not supplied

How networks reshape organizations--for results.

Recently a new term-networks-has entered the vocabulary of corporate renewal. Yet there remains much confusion over just what networks are and how they operate. Ram Charan, a leading international consultant, has spent four years observing and participating in the creation of networks at ten companies in North America and Europe. These companies--which include Conrail, Dun & Bradstreet Europe, Du Pont, and Royal Bank of Canada-are clear about why they are creating networks, what networks are, and how they operate. A network is recognized group of managers (seldom more than 100, often fewer than 25) assembled by the CEO. Membership criteria are simple but subtle: What select group of managers, by virtue of its business skills, personal motivations and drive, and control of resources is uniquely positioned to shape and deliver on the strategy? Networks begin to matter when they change behavior-the frequency, intensity, and honesty of the dialogue among managers on priority tasks. The process of building a network starts at the top. Senior managers work as change agents to build a new "social architecture." Once the network is in place, they play three additional roles: 1. Define with clarity the business outputs they expect of the network and the time frame in which they expect it to deliver. 2. Guarantee the visibility and free flow of information to all members of the network who need it. 3. Develop new criteria for performance evaluation that emphasize horizontal collaboration and leadership.

Commerce↗
PubMed1999

Neuronal networks in the genetically epilepsy-prone rat.

It is now possible to develop a dynamic neuronal network model for generalized convulsive seizures because of in vivo data recently obtained in a naturally occurring epilepsy model--the genetically epilepsy-prone rats (GEPR-9s). GEPR-9s exhibit audiogenic seizures (AGS) that consist of a sequence of discrete behavioral phases (i.e., wild running, clonus-tonus, and post-ictal depression). The neuronal firing changes in most nuclei implicated in the network during each phase of AGS in behaving GEPR-9s have been examined. The inferior colliculus is critical in AGS initiation, because extensive firing increases in inferior colliculus are observed preceding seizure initiation. The deep layers of superior colliculus (DLSC) are crucial to wild running, based on the emergence of tonic firing of DLSC neurons just preceding this phase. The pontine reticular nucleus (PRF) and periaqueductal gray (PAG) are critical to the clonic-tonic phase, because tonic firing patterns appear in these neurons just prior to this phase. During post-ictal depression all areas except the PRF are quiescent. These temporal relationships suggest that each nucleus plays a specific hierarchic role in each discrete convulsive behavior. Generalized tonic-clonic seizure behavior observed in human epilepsy, in GEPR-9s, and in other seizure models is likely to involve similar neuronal network components. The neurotransmitter mechanisms subserving the abnormal neuronal responses in the GEPR-9 neuronal network involve an increased availability of glutamate and a decrease in the effectiveness of gamma-aminobutyric acid (GABA) in many brain regions. Focal modification of the effects of GABA, glutamate, norepinephrine, or serotonin also modulates the nuclei of the network differentially. Together, these data reveal the anatomic, neurotransmitter, and neurophysiologic mechanisms of the neuronal network hierarchy in GEPR-9s, which is currently the most completely developed of any generalized convulsive model. Differential effects of anticonvulsants on the AGS phases and concomitant differential modifications of neuronal firing are observed on neurons in these network nuclei. With nearly complete identification of the network nuclei, the differential effects of these anticonvulsant drugs on different aspects of neuronal firing in different brain sites indicate that this experimental approach can likely identify the most sensitive therapeutic target for these agents. This concept is potentially vital to developing the most selective treatment of different convulsive behaviors occurring in human epilepsy. The neuronal network for AGS does not require brain structures rostral to the midbrain for seizure expression. However, the forebrain is recruited into an expanded seizure network through AGS repetition ("kindling"), resulting in prolonged AGS, post-tonic clonus, and epileptiform electrographic cortical abnormalities. AGS kindling produces network expansion into medial geniculate body (MGB) and amygdala and involves neuronal firing increases in MGB.

Animals↗
PubMed2000

Can women's network activities lead to improved health?

The aim of our study was to focus on women's networks in the Swedish county of Jämtland, and to analyse the relationship between network activities and the perceived health among the network participants, as well as participation in the community among its residents. Questionnaires were distributed to all 68 women's networks and 60 responded. The results show that the network activities correlate with an improved perception of health, as well as with increased participation in the community among the residents. The more support, influence, self-reliance and trust in the future experienced by the networks through their work, the better the health among the participants in the network. The greater the self-reliance, trust in the future, amount of network-related unpaid work and new jobs, the better the participation among the community residents. Despite its limitations, our study suggests that women's networks could have an important role in health promotion within the framework of the new public health. The network model, with its bottom-up strategy, could be useful in public health, provided that networking is not used to fill a gap when there are cuts in the public sector.

Attitude to Health↗
PubMed1992

Cytochemical and immunocytochemical characterization of a fibrillar network (GP2) in pancreatic juice: possible role as a sieve in the pancreatic ductal system.

The secretory product of the exocrine pancreas contains sedimentable and non-sedimentable materials. Electron microscopy of the pellet obtained after ultracentrifugation reveals two major components: microvesicles (pancreasomes) and a fibrillar network of small mesh size. Negative staining of an unfixed pellet demonstrated that these structures are not fixation artifacts. Cytochemical analysis showed that pancreasomes are reactive to osmication and uranyl acetate staining, whereas the fibrillar network was unreactive thereby indicating that the latter does not contain lipids; however, lead citrate staining reveals the network. Alcian blue, known to bind sulfate groups of mucosubstances, reacted strongly with the fibrillar network. The pellet was also characterized by immunocytochemistry with specific antibodies to amylase and glycoprotein 2 (GP2). Both antibodies were located only on the fibrillar network. Washing of the pellet with 100 mM KCl-250 mM NaBr had little effect on GP2 content, but reduced considerably alpha-amylase associated with the reticular matrix. It appeared that GP2 was the major component of the scaffolding that gives rise to the fibrillar network and that other proteins such as alpha-amylase could reversibly bind to it. When double-labeling immunocytochemistry was carried out on the unwashed pellet, labeling of the first antigen reduced the labeling of the second. Removal of amylase by washing the pellet increased the GP2 signal. These results indicate that amylase is bound on the GP2 network. Although the function of the GP2 network is still not clearly defined several possibilities could be envisaged at the level of the pancreatic duct system: 1) The network could drain off any aggregates or precipitates forming in small ducts. 2) The small mesh of the network would present a physical barrier to infecting bacteria that could enter into the duct system from the intestine, especially in conditions of low flow rates. 3) The network may exert a mechanical pressure on the membranes bordering the acinar lumen and small ducts thereby preventing their collapse in basal conditions.

Animals↗
PubMedDate not supplied

Better sales networks.

Anyone in sales will tell you that social networks are critical. The more contacts you have, the more leads you'll generate, and, ultimately, the more sales you'll make. But that's a vast oversimplification. Different configurations of networks produce different results, and the salesperson who develops a nuanced understanding of social networks will outshine competitors. The salesperson's job changes over the course of the selling process. Different abilities are required in each stage of the sale: identifying prospects, gaining buy-in from potential customers, creating solutions, and closing the deal. Success in the first stage, for instance, depends on the salesperson acquiring precise and timely information about opportunities from contacts in the marketplace. Closing the deal requires the salesperson to mobilize contacts from prior sales to act as references. Managers often view sales networks only in terms of direct contacts. But someone who knows lots of people doesn't necessarily have an effective network because networks often pay off most handsomely through indirect contacts. Moreover, the density of the connections in a network is important. Do a salesperson's contacts know all the same people, or are their associates widely dispersed? Sparse networks are better, for example, at generating unique information. Managers can use three levers--sales force structure, compensation, and skills development--to encourage salespeople to adopt a network-based view and make the best possible use of social webs. For example, the sales force can be restructured to decouple lead generation from other tasks because some people are very good at building diverse ties but not so good at maintaining other kinds of networks. Companies that take steps of this kind to help their sales teams build better networks will reap tremendous advantages.

Commerce↗
PubMed1990

[Three-dimensional study of the microvascular network in the excretory end portion and the excretory ducts system of dog parotid gland].

The purpose of this study was to undertake a three-dimensional analysis of the microvascular network of the glandular body and the excretory ducts system of the dog parotid gland using microcorrosive resin casting technique and examined under scanning electron microscope. Results were as follows: 1. Contributed to the direct injection of the resin through the papilla of parotid, not only the excretory ducts but also the striated ducts, intercalated ducts and the end portion were completely filled with resin, thus enabling the fabrication of the intact casts. 2. One of the advantages of the present method was the ability to distinguish the each portion of the excretory ducts system based on the resin casts surface appearance of the different structure of the wall of the lumen. 3. Within the glandular body arteries, veins and excretory ducts vessels were found running their course through the connective tissue, finally emerging through the hilus of lobules into the lobules. However, several arterioles and venules, besides of running through the hilus, directly terminated at the capillary network in lobules. 4. No arteriovenous anastomoses (AVAs) were found in the lobules. However, AVAs were observed between the arterioles and venules distributed in the connective tissue around the lobules. 5. A large number of venous valves were observed in the veins immediately emerging from the hilus of lobules and also in the veins within the connective tissue of the gland. 6. Observing the vascular network of excretory ducts system, there was no proper vascular network belong to the intercalated ducts, however, on the striated ducts, one layer of the vascular network composed of dense capillaries was identified. Confluencing with each other, the diameter of excretory duct increased and its vascular network consist of two layers of vascular network of which inner layer was a dense capillary network similar to the striated ducts, the outer layer was a loose network composed of arterioles and venules. In the main excretory duct, these two layers of the vascular network were composed completely. 7. Some AVAs were observed in the arteriole and venule network around the main excretory duct.

Animals↗
PubMed1994

Network analysis of cortical visual pathways mapped with PET.

Brain metabolic mapping techniques, such as positron emission tomography (PET), can provide information about the functional interactions within entire neural systems. With the large quantity of data that can accumulate from a mapping study, a network analysis, which makes sense of the complex interactions among neural elements, is necessary. A network analysis was performed on data obtained from a PET study that examined both the changes in regional cerebral blood flow (rCBF) and interregional correlations among human cortical areas during performance of an object vision (face matching) and spatial vision (dot-location matching) task. Brain areas for the network were selected based on regions showing significant rCBF or interregional correlations between tasks. Anterior temporal and frontal lobe regions were added to the network using a principal components analysis. Interactions among selected regions were quantified with structural equation modeling. In the structural equation models, connections between brain areas were based on known neuroanatomy and the interregional correlations were used to calculate path coefficients representing the magnitude of the influence of each directional path. The combination of the anatomical network and interregional correlations created a functional network for each task. The functional network for the right hemisphere showed that in the object vision task, dominant path influences were among occipitotemporal areas, while in the spatial vision task, occipitoparietal interactions were stronger. The network for the spatial vision task also had a strong feedback path from area 46 to occipital cortex, an effect that was absent in the object vision task. There were strong interactions between dorsal and ventral pathways in both networks. Functional networks for the left hemisphere did not differ between tasks. Networks for the interhemispheric interactions showed that the dominant pathway in the right hemisphere also had stronger effects on homologous left hemisphere areas and are consistent with a hypothesis that intrahemispheric interactions were greater in the right hemisphere in both tasks, and that these influences were transmitted callosally to the left hemisphere.

Adult↗
PubMed2004

Preliminary studies on the in silico evolution of biochemical networks.

Due to the variety and importance of roles performed by signalling networks, understanding their function and evolution is of great interest. Signalling networks allow organisms to process and react to changes in their internal and external environment. Current estimates suggest that two to three percent of all genomes code for proteins involved in signalling networks. The study of signalling networks is hindered by the complexities of the networks and difficulties in ascribing function to form. For example, a very complex dense network might comprise eighty or more densely connected proteins. In the majority of cases there is very little understanding of how these networks process signals. Unlike in electronics, where there is a broad practical and theoretical understanding of how to construct devices that can process almost any kind of signal, in biological signalling networks there is no equivalent theory. Part of the problem stems from the fact that in most cases it is unknown what particular signal processing circuits would look like in a biological form. This paper describes the evolutionary methods used to generate networks with particular signal- and computational-processing capabilities. The techniques involved are described, and the approach is illustrated by evolving computational circuits such as multiplication, radicals and logarithmic functions. The experiments also illustrate the evolution of modularity within biochemical reaction networks.

Biochemistry↗
PubMed1995

Quantitative structure-pharmacokinetic relationships (QSPR) of beta blockers derived using neural networks.

This study demonstrates the application of neural networks to predict the pharmacokinetic properties of beta-adrenoreceptor antagonists in humans. A congeneric series of 10 beta-blockers, whose critical pharmacokinetic parameters are well established, was selected for the study. An appropriate neural network system was constructed and tested for its ability to predict the pharmacokinetic parameters from the octanol/water partition coefficient (shake flask method), the pKa, or the fraction bound to plasma proteins. Neural networks successfully trained and the predicted pharmacokinetic values agreed well with the experimental values (average difference = 8%). The neural network-predicted values showed better agreement with the experimental values than those predicted by multiple regression techniques (average difference = 47%). Because the neural networks had a large number of connections, two tests were conducted to determine if the networks were memorizing rather than generalizing. The "leave-one-out" method verified the generalization of the networks by demonstrating that any of the compounds could be deleted from the training set and its value correctly predicted by the new network (average error = 19%). The second test involved the prediction of pharmacokinetic properties of compounds never seen by the network, and reasonable results were obtained for three out of four compounds tested. The results indicate neural networks can be a powerful tool in exploration of quantitative structure-pharmacokinetic relationships.

Acebutolol↗
PubMed1997

Frustrated chaos in biological networks.

The behaviour of three biologically inspired networks is investigated: Immune Idiotypic Network (IIN), Hopfield Network (HN) and Coupled Map Lattice (CML), mainly the effect on the dynamics induced by connecting the network in a frustrated way. Frustration occurs when the global structure is such that local connectivity patterns responsible for stable behaviour are intertwined, leading to mutually competing attractors and chaotic itinerancy among brief appearance of these attractors. Frustration destabilizes the network and provokes an unpredictable "wavering" among the stable dynamic regimes which characterize the same network when it is interconnected in a non-frustrated way. As the main contribution of this paper, an immune idiotypic network in which the prevailing behaviour is oscillatory is studied in detail. It is shown how connecting an elementary three-clone network in a frustrating way transforms the oscillatory regime into a chaotic one. This chaotic regime is further analysed and several interesting aspects are discussed such as the variable homogeneity, the intrinsic chaotic itinerancy among brief oscillatory regimes and the strong unpredictability. In addition, dynamical regimes obtained by frustrating the connectivity of HN and CML are presented and the similarities as well as the differences with the IIN dynamics are emphasized. Common to all these networks is the description of the frustrated chaos as a succession of attempts to relax the network into one of the oscillatory regimes given by a weaker and non-frustrated connectivity, an impossible achievement making the dynamics rambling over brief but repelling orbits.

Animals↗
PubMed1991

Application of an artificial neural network in radiographic diagnosis.

The description of 44 cases of bone tumors was used by an artificial neural network to rank the likelihood of 55 possible pathologic diagnoses. The performance of the artificial neural network was compared with the performance of experienced (3 or more years of radiology training) residents and inexperienced (less than 1 year of radiology training) residents. The artificial neural network was trained using descriptions of 110 radiographs of bone tumors with known diagnoses. The descriptions of a separate set of 44 cases were used to test the neural network. The neural network ranked 55 possible pathologic diagnoses on a scale from 1 to 55. Experienced and inexperienced residents also ranked the possible diagnoses in the same 44 cases. Inexperienced residents had a significantly lower mean proportion of diagnoses ranked first or second than did the neural network. Experienced residents had a significantly higher proportion of correct diagnoses ranked first than did the network. Otherwise, a significant difference between the performance of the network and experienced or inexperienced residents was not identified. These results demonstrate that artificial neural networks can be trained to classify bone tumors. Whether neural network performance in classification of bone tumors can be made accurate enough to assist radiologists in clinical practice remains an open question. These preliminary results indicate that further investigation of this technology for interpretation assistance is warranted.

Bone Neoplasms↗
PubMed2006

[Main issues and contributions of the evaluation of health care networks].

Health care networks correspond to a form of horizontal and flexible organisation, which consists in dealing with complex problems of health and long-term care. Networks rest entirely on the mechanisms of coordination between the various actors of care, whose keystone is the medical information flow. The challenge of the health networks evaluation is to prove the added value of a network organization, instead of a conventional disease-centred health care. The evaluation of health care networks is legally required but not sustained by specialists. Evaluation must ensure the existence of a real management; measure the interest of health care professionals and the impact of the network on professional practices. Moreover, evaluation accompanies the development of the network and offers the advisability to the decision maker of obtaining a tool for control of management, i.e. an information tool and a decision-making help, with the aim of a continuous improvement of the quality of care. However the absence of formation of the medical coordinators to management, makes networks management problematic. Thus, health care networks cannot make the saving in a good administrative and management base. In the same way health care networks must obtain an information system adapted to their type of coordination, allowing the evaluation required by the financier.

Community Networks↗
PubMed2005

Artificial neural network models for prediction of acute coronary syndromes using clinical data from the time of presentation.

STUDY OBJECTIVE: Clinical and ECG data from presentation are highly discriminatory for diagnosis of acute coronary syndromes, whereas definitive diagnosis from serial ECG and cardiac marker protein measurements is usually not available for several hours. Artificial neural networks are computer programs adept at pattern recognition tasks and have been used to analyze data from chest pain patients with a view to developing diagnostic algorithms that might improve triage practices in the emergency department. The aim of this study is to develop and optimize artificial neural network models for diagnosis of acute coronary syndrome, to test these models on data collected prospectively from different centers, and to establish whether the performance of these models was superior to that of models derived using a standard statistical technique, logistic regression. METHODS: The study used data from 3,147 patients presenting to 3 hospitals with acute chest pain. Data from hospital 1 were used to train the models, which were then tested on independent data from the other 2 hospitals. From 40 potential factors, variables were selected according to the logarithm of their likelihood ratios to produce models using 8, 13, 20, and 40 factors. Identical data were used for logistic regression and artificial neural network models. Calibration and performance were assessed, the latter using receiver operating characteristic (ROC) curve analysis. RESULTS: Although the performance of artificial neural network models generally increased with increasing numbers of factors, this was insignificant. The 13-factor model was therefore used for the rest of the study owing to its marginally improved calibration over the smallest model. Area under the ROC curve (with standard error) was 0.97 (0.006). The overall sensitivity and specificity of this model for acute coronary syndrome diagnosis using the training data was 0.93. ROC curves for logistic regression and artificial neural network models applied to data from the 3 hospitals were identical. For the 13-factor artificial neural network model tested on data from hospitals 2 and 3, area under the ROC curves (standard error) were 0.93 (0.006) and 0.95 (0.009), respectively. Investigation of the performance of the artificial neural network models throughout the range of predicted probabilities showed that they were well calibrated. CONCLUSION: This study confirms that artificial neural networks can offer a useful approach for developing diagnostic algorithms for chest pain patients; however, the exceptional performance and simplicity of the logistic model militates in favor of logistic regression for the present task. Our artificial neural network models were well calibrated and performed well on unseen data from different centers. These issues have not been addressed in previous studies. However, and unlike in previous studies, we did not find the performance of artificial neural network models to be significantly different from that of suitably optimized logistic regression models.

Adolescent↗
PubMed2006

Topological relationships between brain and social networks.

Brains are complex networks. Previously, we revealed that specific connected structures are either significantly abundant or rare in cortical networks. However, it remains unknown whether systems from other disciplines have similar architectures to brains. By applying network-theoretical methods, here we show topological similarities between brain and social networks. We found that the statistical relevance of specific tied structures differs between social "friendship" and "disliking" networks, suggesting relation-type-specific topology of social networks. Surprisingly, overrepresented connected structures in brain networks are more similar to those in the friendship networks than to those in other networks. We found that balanced and imbalanced reciprocal connections between nodes are significantly abundant and rare, respectively, whereas these results are unpredictable by simply counting mutual connections. We interpret these results as evidence of positive selection of balanced mutuality between nodes. These results also imply the existence of underlying common principles behind the organization of brain and social networks.

Animals↗
PubMed1997

Artificial neural networks applied to outcome prediction for colorectal cancer patients in separate institutions.

BACKGROUND: Artificial neural networks are computer programs that can be used to discover complex relations within data sets. They permit the recognition of patterns in complex biological data sets that cannot be detected with conventional linear statistical analysis. One such complex problem is the prediction of outcome for individual patients treated for colorectal cancer. Predictions of outcome in such patients have traditionally been based on population statistics. However, these predictions have little meaning for the individual patient. We report the training of neural networks to predict outcome for individual patients from one institution and their predictive performance on data from a different institution in another region. METHODS: 5-year follow-up data from 334 patients treated for colorectal cancer were used to train and validate six neural networks designed for the prediction of death within 9, 12, 15, 18, 21, and 24 months. The previously trained 12-month neural network was then applied to 2-year follow-up data from patients from a second institution; outcome was concealed. No further training of the neural network was undertaken. The network's predictions were compared with those of two consultant colorectal surgeons supplied with the same data. FINDINGS: All six neural networks were able to achieve overall accuracy greater than 80% for the prediction of death for individual patients at institution 1 within 9, 12, 15, 18, 21, and 24 months. The mean sensitivity and specificity were 60% and 88%. When the neural network trained to predict death within 12 months was applied to data from the second institution, overall accuracy of 90% (95% CI 84-96) was achieved, compared with the overall accuracy of the colorectal surgeons of 79% (71-87) and 75% (66-84). INTERPRETATION: The neural networks were able to predict outcome for individual patients with colorectal cancer much more accurately than the currently available clinicopathological methods. Once trained on data from one institution, the neural networks were able to predict outcome for patients from an unrelated institution.

Bias↗
PubMed1998

Neural network computer program to determine photorefractive keratectomy nomograms.

PURPOSE: To evaluate a commercially available neural network program for calculation of photorefractive keratectomy treatment nomograms. SETTING: University referral refractive surgery clinic. METHODS: PRK/LASIK Brain, a commercial neural network computer program, was trained using the demographics, preoperative clinical data, surgical parameters, and 1 year postoperative clinical data of 44 patients treated with a Summit Technology excimer laser using a 5.0 mm optical zone. The neural-network derived nomogram was compared with the standard treatment nomogram for each patient. The relative contribution of age, sex, keratometry, and intraocular pressure (IOP) to the predicted nomograms was also assessed. RESULTS: Nomograms produced by the neural network were qualitatively similar to the standard nomogram. The sequence of data entry during training affected the network's predictions. Entry ordered by outcome (as opposed to entry by chronological order) yielded a nomogram that was more consistent with the standard nomogram. However, both outcome- and chronologically ordered network-derived nomograms diverged from the standard nomogram in individual patients, including a subset for whom use of the standard nomogram yielded desired refractive results (within 0.25 diopter of emmetropia). Further analysis of the neural-network-derived nomograms revealed marked sensitivity to sex, age, keratometry readings, and IOP. CONCLUSIONS: Neural networks offer a potential means of individualizing treatment nomograms, to account for patient demographics, preoperative examination, surgeon style, and equipment bias. However, a data set of 44 patients was not sufficient to train the PRK/LASIK Brain network to accurately predict treatment parameters in individual cases in the training set. A larger training set or a different learning algorithm may be required to improve the neural network's performance.

Adult↗
PubMed2000

Modular neural networks for non-linearity recovering by the Haar approximation.

The paper deals with the design of a composite neural system for recovering non-linear characteristics from random input-output measurement data. It is assumed that non-linearity output measurements are corrupted by an additive zero-mean white random noise and that the input excitation is an i.i.d. random sequence with an arbitrary (and unknown) probability density function. A class of modular networks is developed. The class is based on the Haar approximation of functions with piecewise constant functions on a refinable grid and consists of the networks composed of perceptron-like modules connected in parallel. The networks provide a local mean value estimators of functions. The relationship between complexity and accuracy of modular networks is analysed. It is shown that under mild conditions on the non-linearities and input probability density functions the networks yield pointwise consistent estimates of non-linear characteristics, provided that complexity of the networks grows appropriately with the number of training data. Efficiency of the networks is examined and the asymptotic rate of convergence of the network estimates is established. Specifically, local ability of the networks to recover non-linear characteristics in dependence on local smoothness of the underlying non-linear function and the input probability density is discussed. Optimum complexity selection rules, guaranteeing the best performance of the networks, are given. Illustrative simulation examples are provided.

Computer Simulation↗