SEARCH · PubMed Health

Results for “Predictive Learning Models”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

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

At least 1,639 records · Page 91Linked to original sources

PubMed2002

Clinical determinants of survival in patients with 5-fluorouracil-based treatment for metastatic colorectal cancer: results of a multivariate analysis of 3825 patients.

BACKGROUND: Patients with metastatic colorectal cancer are usually offered systemic chemotherapy as palliative treatment. A multivariate analysis was performed in order to identify predictors and their constellation that allow a valid prediction of the outcome in patients treated with 5-fluorouracil (5-FU)-based therapy. PATIENTS AND METHODS: A total of 3825 patients treated with 5-FU within 19 prospective randomised and three phase II trials were separated into learning (n = 2549) and validation (n = 1276) samples. Data were analysed by tree analysis using the recursive partition and amalgamation method (RECPAM). A predictor could only enter the RECPAM analysis if the number of patients with missing values was < 33.3% within a node, and the minimal node size was set to 50 patients. Twenty-three potential predictors were grouped into subsets of laboratory variables (11 parameters), tumour-related variables (seven parameters) and clinical variables (five parameters). In the first step, tree analysis was performed separately for each predictor subset. The selected prognostic parameters of the resulting partial models (the 'winners') were entered into the general model. The classification rule from the data of the learning set was applied to the independent validation set. RESULTS: Winners of the subgroup analysis for laboratory variables were: platelets > or = 400 x 10(9)/l, alkaline phosphatase > or = 300 U/l, white blood cell (WBC) count > or = 10 x 10(9)/l and haemoglobin < 11 x 10(9)/l, and all predicted a worse outcome. Negative predictors within the subgroup of tumour parameters were: number of tumour sites more than one or more than two, presence of liver metastases or peritoneal carcinomatosis, which predicted a worse outcome. Furthermore, presence of lung metastases, a primary rectal cancer and presence of lymph node metastases all predicted a better outcome in the multivariate setting. Among the clinical parameters only performance status of ECOG 0 or 1 predicted better outcome. In the final regression tree, three risk groups could be identified: low risk group (n = 1111) with a median survival of 15 months for patients with ECOG 0/1 and only one tumour site; intermediate risk group (n = 904) with a median survival of 10.7 months for patients with ECOG 0/1 and more than one tumour site and alkaline phosphatase < 300 U/l or patients with ECOG > 1, WBC count < 10 x 10(9)/l and only one tumour site; high risk group (n = 534) with a median survival of 6.1 months for patients with ECOG 0/1 and more than one tumour site and alkaline phosphatase of > or = 300 U/l or patients with ECOG > 1 and more than one tumour site or WBC count > 10 x 10(9)/l. The median survival times for the good, intermediate and high risk groups in the validation sample were 14.7, 10.5 and 6.4 months, respectively. CONCLUSIONS: Patients can be divided into at least three risk groups depending on the four baseline clinical parameters: performance status, WBC count, alkaline phosphatase and number of metastatic sites. Any molecular or biological marker should be validated against these clinical parameters and decisions for more or less intensive treatments may be studied separately in these three risk groups. Also, clinical trials should be stratified according to the three risk groups.

Alkaline Phosphatase↗

PubMed2003

Animal models of focal and global cerebral ischemia.

The use of appropriate animal models is essential to predict the value and effect of therapeutic approaches in human subjects. Focal (stroke) and global (cardiac arrest) cerebral ischemia represents diseases that are common in the human population. Stroke and cardiac arrest, which are major causes of death and disability, affect millions of individuals around the world and are responsible for the leading health care costs of all diseases. Understanding the mechanisms of injury and neuroprotection in these diseases is critical if we are ever to learn new target sites to treat ischemia. There are many animal models available to investigate injury mechanisms and neuroprotective strategies. This review summarizes many (but not all) small and large animal models of focal and global cerebral ischemia and discusses their advantages and disadvantages.

Animals↗

PubMed2005

Absolute pitch: music and beyond.

"Perfect pitch," known in the scientific literature as "absolute pitch" (AP), is a rare phenomenon that has fascinated musicians and scientists alike for over a century. There has been a great deal of conflict in the literature between advocates of the two main theories on the etiology of AP: some believe that AP is learned early in life through intensive musical training, whereas others believe AP to be largely innate. Both theories are alike, however, in considering AP to be exclusively a musical phenomenon. We propose a paradigm shift by presenting here a new model of AP, one that is predicated on two principles: (1) that AP may be relatively independent of musical experience; and (2) that there are different types of AP, each of which can be ascribed to discrete neurobiological mechanisms. We also review data from a diverse series of experiments that were designed to test explicitly both the predictions of our model and a series of historical myths about AP. In each case, the data strongly support our model. We conclude with a general discussion on the nature of AP, the relevance of these findings for other areas of research, and future directions of study.

Acoustic Stimulation↗

PubMed1993

Measuring the effect of attention on simple visual search.

Set-size in visual search may be due to 1 or more of 3 factors: sensory processes such as lateral masking between stimuli, attentional processes limiting the perception of individual stimuli, or attentional processes affecting the decision rules for combining information from multiple stimuli. These possibilities were evaluated in tasks such as searching for a longer line among shorter lines. To evaluate sensory contributions, display set-size effects were compared with cuing conditions that held sensory phenomena constant. Similar effects for the display and cue manipulations suggested that sensory processes contributed little under the conditions of this experiment. To evaluate the contribution of decision processes, the set-size effects were modeled with signal detection theory. In these models, a decision effect alone was sufficient to predict the set-size effects without any attentional limitation due to perception.

Adult↗

PubMed1989

The composition of imaginal transformations.

Three experiments examine the performance of composite imaginal spatial transformations using tasks in which people perform pairs of distinct imaginal transformations either in close succession or simultaneously. The principal aim of the investigation is to determine, for both modes of composition, whether imaginal transformations combine interactively or noninteractively to determine the temporal course of imaging performance. Specifically, does the time to imagine a given composite transformation reflect the independent contributions of its component transformations, or do component transformations have different effects within different composites? Results obtained under a number of distinct task conditions are most consistent with the proposal of noninteractive composition. However, the temporal effects of particular imaginal transformations are neither consistent nor readily predictable across different task conditions. Implications of these results for analogue-spatial models, task demand, and procedural accounts of imaging are discussed.

Adolescent↗

PubMed2002

Perirhinal cortex resolves feature ambiguity in complex visual discriminations.

The present experiment tested predictions of a 'perceptual-mnemonic/feature conjunction' (PMFC) model of perirhinal cortex function. The model predicts that lesions of perirhinal cortex should disrupt complex visual discriminations with a high degree of 'feature ambiguity', a property of visual discrimination problems that can emerge when features of an object are rewarded when they are part of one object, but not when part of another. As feature ambiguity is thought to be the critical factor, such effects should be independent of the number of objects to be discriminated. This was tested directly, by assessing performance of control monkeys and monkeys with aspiration lesions of perirhinal cortex on a series of concurrent discriminations in which the number of object pairs was held constant, but the degree of feature ambiguity was varied systematically. Monkeys were tested in three conditions: Maximum Feature Ambiguity, in which all features were explicitly ambiguous (AB+, CD+, BC-, AD-; the biconditional problem); Minimum Feature Ambiguity, in which no features were explicitly ambiguous (AB+, CD+, EF-, GH-); and Intermediate Feature Ambiguity, in which half the features were explicitly ambiguous (AB+, CD+, CE-, AF-). The pattern of results closely matched that predicted by simulations using a connectionist network: monkeys with perirhinal cortex lesions were unimpaired in the Minimum Feature Ambiguity condition, mildly impaired in the Intermediate Feature Ambiguity condition and severely impaired in the Maximum Feature Ambiguity condition. These results confirm the predictions of the PMFC model, and force a reconsideration of prevailing views regarding perirhinal cortex function.

Animals↗

PubMed2003

Similarity and discrimination in human Pavlovian conditioning.

We report three Pavlovian eyelid conditioning experiments with humans, designed to experimentally decide between elemental and configural learning theories. We used two different designs originally proposed by Redhead and Pearce (1995). In Experiments 1 and 2, three stimulus elements, A, B, and C, were presented in all possible combinations. All patterns were reinforced except for pattern ABC (A/B/C+, AB/AC/BC+, ABC-). According to elemental learning theories, response proportions on A/B/C+ trials should be smaller than on AB/AC/BC+ trials, whereas configural learning theory makes the opposite prediction. The results confirmed neither prediction. In Experiment 3, the A/B/C+, AB/AC/BC+, and ABC- trials were interspersed by D/E/F-, DE/DF/EF-, DEF+ trials. Again, neither prediction was confirmed. We suggest a modification of configural learning theory as a possible explanation of our results.

Adult↗

PubMed1992

Computational behavior dynamics: an alternative description of Nevin (1969).

A computational processing behavior-dynamic model was instantiated in the form of a computer program that "behaved" on the task developed by Nevin (1969). In this classic discrete-trials experiment, the relative frequency of choosing a response alternative matched the relative frequency of reinforcement for that alternative, the local structure of responding was opposite that predicted by momentary maximizing (i.e., the probability of a changeover decreased with run length), and absolute and relative response rates varied independently. The behavior-dynamic model developed here qualitatively reproduced these three results (but not in quantitative and specific detail) and also generated some interesting, as-yet-untested predictions about performance in Nevin's task. The model was discussed as an example of a stochastic behavior-dynamic alternative to algebraic behavior theory.

Animals↗

PubMed2003

Assessing power PC.

In a recent theoretical paper, Cheng (1997) presented a new causal model, power PC. She argued that power PC was able to account for data in the literature that raised problems for associative models--notably, the Rescorla-Wagner model (Rescorla & Wagner, 1972). The purpose of the present paper is threefold: (1) to show that, overall, the data in the literature, which Cheng relied on to make her case, do not in fact provide support for power PC, (2) to show that, overall, the experiments reported in the literature since the publication of Cheng, designed specifically to evaluate the predictions of power PC, also do not provide support for power PC, and (3) to suggest that Cheng's assessment of associative models was too narrowly defined.

Algorithms↗

PubMed1990

Similarity between Fourier transforms of objects predicts their experimental confusions.

Two sets of stimuli were presented tachistoscopically to 4 subjects. On each trial, a single stimulus was presented, and the subject was required to identify the stimulus by verbal response. An exposure duration was chosen such that the subject's identification performance fell within a range from faultless identification to chance guessing. The object-identification data of each subject obtained for all stimulus exposures were pooled to form an object confusion matrix. A model of visual processing based on two-dimensional spatial frequency content (Fourier transforms) was used to predict confusions among stimulus pairs. The model properties that appear to be the most essential are those that allow it (1) to account for the obvious dependence of the Fourier transform on the choice of an origin point; and (2) choose the point of origin for each object separately, irrespective of other objects of the set. The point of origin of the reference frame, in which Fourier transforms are performed, is chosen so as to minimize the low-frequency phase component for each object. A high correlation (up to .96) between confusion matrices and model interobject distances was attained. The results demonstrate that such a distance measure gives a good prediction of object confusability.

Adult↗

PubMedDate not supplied

Adaptation of a primate operant test battery to the rat: effects of chlorpromazine.

The National Center for Toxicological Research (NCTR) Operant Test Battery (OTB) has been used extensively in rhesus monkeys to characterize the effects of drugs and toxicants on the performance of tasks designed to model several cognitive functions. Recently, the majority of the OTB tasks have been adapted for use in rats. The current study is the first to examine the effects of a prototypic pharmacological agent previously assessed in monkeys on rat OTB performance. The effects of the dopamine antagonist chlorpromazine (0.56-5.6 mg/kg, i.p.) were assessed in rats performing tasks designed to model auditory-visual-position discrimination, learning, time estimation, and appetitive motivation. All four tasks were equally sensitive to the behavioral effects of chlorpromazine. This pattern of sensitivity was very similar to that obtained when chlorpromazine was tested in monkeys performing the OTB. These data thus suggest that operant tasks designed to model cognitive functions in monkeys can also be used in rats, and that the effects of chlorpromazine on the performance of these tasks may be predictive of results obtained with monkeys. Further characterization of the rat OTB using prototypic pharmacological agents will further determine the extent to which drug effects on rat OTB performance can be generalized to primates.

Animals↗

PubMed1998

Haloperidol does not affect motivational processes in an operant runway model of food-seeking behavior.

The present experiment examined the effects of dopamine receptor antagonism on subjects' motivation to seek food. Rats were trained to discriminate between 2 olfactory cues predicting either the presence (S+) or absence (S-) of food reinforcement in the goal box of a straight-arm runway. Rats learned to traverse the alley quickly when presented with the S+ and much more slowly when presented with the S-. Haloperidol pretreatment was unable to alter this pattern of behavior (i.e., rats still ran quickly when presented with the scent that predicted food availability). Thus, it seems that the same dopamine antagonist treatments that have been shown to disrupt food reinforcement do not prevent the food-seeking behavior produced by presentation of food-predictive cues.

Analysis of Variance↗

PubMed2006

A comparison and evaluation of the predictions of relational and conjunctive accounts of hippocampal function.

Relational and conjunctive memory theory each postulate that the hippocampus participates in the formation of long-term memory representations comprised of associations between multiple elements. The goals of the current work were to clarify and contrast these theories by outlining the nature of the representations that are spared vs. impaired following hippocampal damage according to each theoretical perspective. Relational theory predicts that hippocampal lesions will impair performance on tasks that require the formation of new long-term representations in which distinct elements must be regarded in relation to all other elements. Representations that remain intact despite hippocampal damage include separate representations of distinct individual elements or multiple stimuli fused into a static "blend" such as several elements viewed from one vantage point. Additionally, the relational account predicts that rapid incidental online processing of the relations can be achieved through structures other than the hippocampus, but this information will not be stored. In contrast, conjunctive theory predicts that hippocampal damage will impair the rapid formation of unitary representations that contain features of elements and their relative relationships bound in an inflexible manner. Deficits in the rapid formation of these conjunctive representations result in impaired performance on tasks that require rapid incidental stimulus binding. However, intact formation of conjunctive representations can occur over multiple trials in the service of problem solving. Using these theoretical frameworks, recent findings from the human and nonhuman animal literature are reexamined in order to determine whether one theory better accounts for current findings. We discuss empirical studies that serve as "critical experiments" in addressing the relational vs. conjunctive debate, and find that the predictions of relational theory are supported by existing findings over those from the conjunctive account.

Animals↗

PubMed2001

Retrospective and prospective coding for predicted reward in the sensory thalamus.

Reward is important for shaping goal-directed behaviour. After stimulus-reward associative learning, an organism can assess the motivational value of the incoming stimuli on the basis of past experience (retrospective processing), and predict forthcoming rewarding events (prospective processing). The traditional role of the sensory thalamus is to relay current sensory information to cortex. Here we find that non-primary thalamic neurons respond to reward-related events in two ways. The early, phasic responses occurred shortly after the onset of the stimuli and depended on the sensory modality. Their magnitudes resisted extinction and correlated with the learning experience. The late responses gradually increased during the cue and delay periods, and peaked just before delivery of the reward. These responses were independent of sensory modality and were modulated by the value and timing of the reward. These observations provide new evidence that single thalamic neurons can code for the acquired significance of sensory stimuli in the early responses (retrospective coding) and predict upcoming reward value in the late responses (prospective coding).

Animals↗

PubMed2005

A review on the integration of artificial intelligence into coastal modeling.

With the development of computing technology, mechanistic models are often employed to simulate processes in coastal environments. However, these predictive tools are inevitably highly specialized, involving certain assumptions and/or limitations, and can be manipulated only by experienced engineers who have a thorough understanding of the underlying theories. This results in significant constraints on their manipulation as well as large gaps in understanding and expectations between the developers and practitioners of a model. The recent advancements in artificial intelligence (AI) technologies are making it possible to integrate machine learning capabilities into numerical modeling systems in order to bridge the gaps and lessen the demands on human experts. The objective of this paper is to review the state-of-the-art in the integration of different AI technologies into coastal modeling. The algorithms and methods studied include knowledge-based systems, genetic algorithms, artificial neural networks, and fuzzy inference systems. More focus is given to knowledge-based systems, which have apparent advantages over the others in allowing more transparent transfers of knowledge in the use of models and in furnishing the intelligent manipulation of calibration parameters. Of course, the other AI methods also have their individual contributions towards accurate and reliable predictions of coastal processes. The integrated model might be very powerful, since the advantages of each technique can be combined.

Artificial Intelligence↗

PubMed2001

Computer-automated dementia screening using a touch-tone telephone.

BACKGROUND: This study investigated the sensitivity and specificity of a computer-automated telephone system to evaluate cognitive impairment in elderly callers to identify signs of early dementia. METHODS: The Clinical Dementia Rating Scale was used to assess 155 subjects aged 56 to 93 years (n = 74, 27, 42, and 12, with a Clinical Dementia Rating Scale score of 0, 0.5, 1, and 2, respectively). These subjects performed a battery of tests administered by an interactive voice response system using standard Touch-Tone telephones. Seventy-four collateral informants also completed an interactive voice response version of the Symptoms of Dementia Screener. RESULTS: Sixteen cognitively impaired subjects were unable to complete the telephone call. Performances on 6 of 8 tasks were significantly influenced by Clinical Dementia Rating Scale status. The mean (SD) call length was 12 minutes 27 seconds (2 minutes 32 seconds). A subsample (n = 116) was analyzed using machine-learning methods, producing a scoring algorithm that combined performances across 4 tasks. Results indicated a potential sensitivity of 82.0% and specificity of 85.5%. The scoring model generalized to a validation subsample (n = 39), producing 85.0% sensitivity and 78.9% specificity. The kappa agreement between predicted and actual group membership was 0.64 (P<.001). Of the 16 subjects unable to complete the call, 11 provided sufficient information to permit us to classify them as impaired. Standard scoring of the interactive voice response-administered Symptoms of Dementia Screener (completed by informants) produced a screening sensitivity of 63.5% and 100% specificity. A lower criterion found a 90.4% sensitivity, without lowering specificity. CONCLUSIONS: Computer-automated telephone screening for early dementia using either informant or direct assessment is feasible. Such systems could provide wide-scale, cost-effective screening, education, and referral services to patients and caregivers.

Aged↗

PubMed2001

The determination of cardiac surgical risk using artificial neural networks.

BACKGROUND: In the study presented here, an artificial neural network was used to "learn" the relationship between 11 risk factors and patient surgical outcome (survival or death). The network was then used to predict the surgical outcome of other patients. METHODS: Eleven risk factors were presented as inputs to an artificial neural network (ANN). The ANN model was developed by training and testing on 1875 patients. The results of the ANN were compared with the results from two versions of the VA surgical risk model (Denver model). Cutoffs were determined for each model in order to compare their results. RESULTS: The ANN model gave the best results when compared with the new and old Denver models. The ANN model had the lowest overall percentage of error. It predicted living patients with an error of 14% and death with an error of approximately 31.0%. The old Denver model predicted living patients with an error of 15% and deaths with an error of 31%. The new Denver model predicts living patients with an error around 18% and the deaths with an error of 31%. CONCLUSIONS: The combined predictions of the ANN model were slightly more accurate than either the new or the old Denver models. The ANN model was created from 1875 patients in about 1 month, while both of the Denver models were developed over a 3 year time period and used more than 12,000 patients. Additionally, the ANN model is easily modified, allowing instant addition or deletion of parameters to suit the users needs.

Cardiac Surgical Procedures↗

PubMed2004

Internal models for bi-manual tasks.

Co-ordinated bi-manual actions form the basis for many everyday motor skills. In this review, the internal model approach to the problem of bi-manual co-ordination is presented. Bi-manual coordinative tasks are often regarded as a hallmark of complex action. They are often associated with object manipulation, whether the holding of a single object between the two hands or holding an object in each hand. However, the task of movement and control is deceptively difficult even when we execute an action with a single hand without holding an object. The simplest voluntary action requires the problems of co-ordination, timing and interaction between neural, muscular and skeletal structures to be overcome. When we are making a movement whilst holding an object, a further requirement is that an internal model is able to predict the dynamics of the object that is being held as well as the dynamics of the motor system. There has been extensive work examining the formation of internal models when acting in novel environments. The majority of studies examine uni-lateral learning of a task generally to the participant's dominant hand. However, many everyday motor tasks are bi-manual, and the existing findings regarding the learning of internal models in uni-manual tasks and their subsequent generalization highlights the complexities that must underlie the formation of bi-manual tasks. Our ability to perform bi-manual tasks raises interesting questions about how internal models are specified for co-ordinative actions, and also for how the motor system learns to represent the properties of objects.

Attention↗