Where is the light? Bayesian perceptual priors for lighting direction.

Perception of shaded three-dimensional figures is inherently ambiguous, but this ambiguity can be resolved if the brain assumes that figures are lit from a specific direction. Under the Bayesian framework, the visual system assigns a weighting to each possible direction, and these weightings define a prior probability distribution for light-source direction. Here, we describe a non-parametric maximum-likelihood estimation method for finding the prior distribution for lighting direction. Our results suggest that each observer has a distinct prior distribution, with non-zero values in all directions, but with a peak which indicates observers are biased to expect light to come from above left. The implications of these results for estimating general perceptual priors are discussed.

Free-lunch learning: modeling spontaneous recovery of memory.

After a language has been learned and then forgotten, relearning some words appears to facilitate spontaneous recovery of other words. More generally, relearning partially forgotten associations induces recovery of other associations in humans, an effect we call free-lunch learning (FLL). Using neural network models, we prove that FLL is a necessary consequence of storing associations as distributed representations. Specifically, we prove that (1) FLL becomes increasingly likely as the number of synapses (connection weights) increases, suggesting that FLL contributes to memory in neurophysiological systems, and (2) the magnitude of FLL is greatest if inactive synapses are removed, suggesting a computational role for synaptic pruning in physiological systems. We also demonstrate that FLL is different from generalization effects conventionally associated with neural network models. As FLL is a generic property of distributed representations, it may constitute an important factor in human memory.

Derivation of a clinical decision rule to guide the interhospital transfer of patients with blunt traumatic brain injury.

OBJECTIVE: To derive a clinical decision rule for people with traumatic brain injury (TBI) that enables early identification of patients requiring specialised trauma care. METHODS: We collected data from 1999 through 2003 on a retrospective cohort of consecutive people aged 18-65 years with a serious head injury (AIS > or =3), transported directly from the scene of injury, and evaluated in the ED. Information on 22 demographical, physiological, radiographic, and lab variables was collected. Resource based "high therapeutic intensity" measures occurring within 72 hours of ED arrival (the outcome measure) were identified a priori and included: neurosurgical intervention, exploratory laparotomy, intensive care interventions, or death. We used classification and regression tree analysis to derive and cross validate the decision rule. RESULTS: 504 consecutive trauma patients were identified as having a serious head injury: 246 (49%) required at least one of the HTI measures. Five ED variables (GCS, respiratory rate, age, temperature, and pulse rate) identified subjects requiring at least one of the HTI measures with 94% sensitivity (95% CI 91 to 97%) and 63% specificity (95% CI 57 to 69%) in the derivation sample, and 90% sensitivity and 55% specificity using cross validation. CONCLUSIONS: This decision rule identified among a cohort of head injured patients evaluated in the ED the majority of those who urgently required specialised trauma care. The rule will require prospective validation in injured people presenting to non-tertiary care hospitals before implementation can be recommended.

Spatiotemporal independent component analysis of event-related fMRI data using skewed probability density functions.

We introduce two independent component analysis (ICA) methods, spatiotemporal ICA (stICA) and skew-ICA, and demonstrate the utility of these methods in analyzing synthetic and event-related fMRI data. First, stICA simultaneously maximizes statistical independence over both time and space. This contrasts with conventional ICA methods, which maximize independence either over time only or over space only; these methods often yield physically improbable solutions. Second, skew-ICA is based on the assumption that images have skewed probability density functions (pdfs), an assumption consistent with spatially localized regions of activity. In contrast, conventional ICA is based on the physiologically unrealistic assumption that images have symmetric pdfs. We combine stICA and skew-ICA, to form skew-stICA, and use it to analyze synthetic data and data from an event-related, left-right visual hemifield fMRI experiment. Results obtained with skew-stICA are superior to those of principal component analysis, spatial ICA (sICA), temporal ICA, stICA, and skew-sICA. We argue that skew-stICA works because it is based on physically realistic assumptions and that the potential of ICA can only be realized if such prior knowledge is incorporated into ICA methods.

Predicting spontaneous recovery of memory.

Long after a new language has been learned and forgotten, relearning a few words seems to trigger the recall of other words. Neural-network models indicate that this form of spontaneous recovery may result from the storage of distributed representations, which are thought to mediate human memory. Here we use a psychomotor learning task to show that a corresponding effect of spontaneous memory recovery occurs in human subjects.

Blind source separation using temporal predictability.

A measure of temporal predictability is defined and used to separate linear mixtures of signals. Given any set of statistically independent source signals, it is conjectured here that a linear mixture of those signals has the following property: the temporal predictability of any signal mixture is less than (or equal to) that of any of its component source signals. It is shown that this property can be used to recover source signals from a set of linear mixtures of those signals by finding an un-mixing matrix that maximizes a measure of temporal predictability for each recovered signal. This matrix is obtained as the solution to a generalized eigenvalue problem; such problems have scaling characteristics of O(N3), where N is the number of signal mixtures. In contrast to independent component analysis, the temporal predictability method requires minimal assumptions regarding the probability density functions of source signals. It is demonstrated that the method can separate signal mixtures in which each mixture is a linear combination of source signals with supergaussian, subgaussian, and gaussian probability density functions and on mixtures of voices and music.

Impact of donor type on outcome of bone marrow transplantation for Wiskott-Aldrich syndrome: collaborative study of the International Bone Marrow Transplant Registry and the National Marrow Donor Program.

Human leukocyte antigen (HLA)-identical sibling bone marrow transplantation is an effective treatment for Wiskott-Aldrich syndrome. However, most children with this disease lack such donors and many patients receive transplants from alternative donors. This study compared outcomes of HLA-identical sibling, other related donor, and unrelated donor transplantation for Wiskott-Aldrich syndrome. The outcome of 170 transplantations for Wiskott-Aldrich syndrome, from 1968 to 1996, reported to the International Bone Marrow Transplant Registry and/or National Marrow Donor Program were assessed. Fifty-five were from HLA-identical sibling donors, 48 from other relatives, and 67 from unrelated donors. Multivariate proportional hazards regression was used to compare outcome by donor type and identify other prognostic factors. Most transplant recipients were younger than 5 years (79%), had a pretransplantation performance score greater than or equal to 90% (63%), received pretransplantation preparative regimens without radiation (82%), and had non-T-cell-depleted grafts (77%). Eighty percent received their transplant after 1986. The 5-year probability of survival (95% confidence interval) for all subjects was 70% (63%-77%). Probabilities differed by donor type: 87% (74%-93%) with HLA-identical sibling donors, 52% (37%-65%) with other related donors, and 71% (58%-80%) with unrelated donors (P =.0006). Multivariate analysis indicated significantly lower survival using related donors other than HLA-identical siblings (P =.0004) or unrelated donors in boys older than 5 years (P =.0001), compared to HLA-identical sibling transplants. Boys receiving an unrelated donor transplant before age 5 had survivals similar to those receiving HLA-identical sibling transplants. The best transplantation outcomes in Wiskott-Aldrich syndrome are achieved with HLA-identical sibling donors. Equivalent survivals are possible with unrelated donors in young children.

When is now? Perception of simultaneity.

We address the following question: Is there a difference (D) between the amount of time for auditory and visual stimuli to be perceived? On each of 1000 trials, observers were presented with a light-sound pair, separated by a stimulus onset asynchrony (SOA) between -250 ms (sound first) and +250 ms. Observers indicated if the light-sound pair came on simultaneously by pressing one of two (yes or no) keys. The SOA most likely to yield affirmative responses was defined as the point of subjective simultaneity (PSS). PSS values were between -21 ms (i.e. sound 21 ms before light) and +150 ms. Evidence is presented that each PSS is observer specific. In a second experiment, each observer was tested using two observer-stimulus distances. The resultant PSS values are highly correlated (r = 0.954, p = 0.003), suggesting that each observer's PSS is stable. PSS values were significantly affected by observer-stimulus distance, suggesting that observers do not take account of changes in distance on the resultant difference in arrival times of light and sound. The difference RTd in simple reaction time to single visual and auditory stimuli was also estimated; no evidence that RTd is observer specific or stable was found. The implications of these findings for the perception of multisensory stimuli are discussed.

Determinants of object recognition.

We demonstrate that performance on an object recognition task can be explained in terms of observer-specific perceptual profiles. These profiles are derived from a battery of tests, including the effects of stereo, texture, outline (occluding contour), and motion cues on amplitude judgements of curved surfaces. Using a task in which observers learned to recognise 'amoeboid' objects, a multivariate regression analysis revealed that three psychometric variables derived from the test battery account for 74% of the variance in learning rate. These variables are choice reaction time, and the relative dependence of amplitude judgements on motion and outline cues. The implications of these findings for the existence of observer-specific perceptual profiles, and their relation to the fundamental psychophysical competences associated with object recognition are discussed.

Factor analysis of three standardized tests of memory in a clinical population.

OBJECTIVES: The aim of this study was to determine the factor structure of three standardized memory tests: Wechsler Memory Scale-Revised (WMS-R), Warrington Recognition Memory Test (WRMT), Doors and People Test (D&P). We investigated whether these different standardized tests of memory are consistent in their evaluation of memory function, and the extent to which these tests discriminate between different memory functions (e.g. recall/recognition and verbal/non-verbal memory). DESIGN: Fifty patients with selective memory impairment were tested on the WMS-R, WRMT and D&P. METHODS: Age-scaled scores from selective measures of these tests (WMS-R-verbal, WMS-R-visual, WMS-R-delay, WRMT-words, WRMT-faces, D&P-people, D&P-doors, D&P-shapes, D&P-names) were used as input to a factor analysis. RESULTS: Maximum likelihood factor analysis yielded a three-factor solution consistent with a theoretically motivated fractionation of memory function into recall and recognition components. Recognition performance, but not recall performance, showed dissociation into visual and verbal components. CONCLUSIONS: The WMS-R, WRMT and D&P are highly consistent in their assessment of memory function. The results of the factor analysis are consistent with a theoretically motivated fractionation of recall and recognition memory. They are also partially consistent with a dissociation between visual and verbal memory function.

Long-term survival and late deaths after allogeneic bone marrow transplantation. Late Effects Working Committee of the International Bone Marrow Transplant Registry.

BACKGROUND AND METHODS: It is uncertain whether mortality rates among patients who have undergone bone marrow transplantation return to the level of the mortality rates of the general population. We analyzed the characteristics of 6691 patients listed in the International Bone Marrow Transplant Registry. All the patients were free of their original disease two years after allogeneic bone marrow transplantation. Mortality rates in this cohort were compared with those of an age-, sex-, and nationality-matched general population. Cox proportional-hazards regression was used to identify risk factors for death more than two years after transplantation (late death). RESULTS: Among patients who were free of disease two years after transplantation, the probability of living for five more years was 89 percent (95 percent confidence interval, 88 to 90 percent). Among patients who underwent transplantation for aplastic anemia, the risk of death by the sixth year after transplantation did not differ significantly from that of a normal population. Mortality remained significantly higher than normal throughout the study among patients who underwent transplantation for acute lymphoblastic leukemia or chronic myelogenous leukemia and through the ninth year among those who underwent transplantation for acute myelogenous leukemia. Recurrent leukemia was the chief cause of death among patients who received a transplant for leukemia, whereas chronic graft-versus-host disease was the chief cause among those who received a transplant for aplastic anemia. Advanced, long-standing disease before transplantation and active chronic graft-versus-host disease were important risk factors for late death. CONCLUSIONS: In patients who receive an allogeneic bone marrow transplant as treatment for acute myelogenous or lymphoblastic leukemia, chronic myelogenous leukemia, or aplastic anemia and who are free of their original disease two years later, the disease is probably cured. However, for many years after transplantation, the mortality among these patients is higher than that in a normal population.

Modelling covariate adjusted mortality relative to a standard population.

A study of long term survival of 1487 patients given an allogeneic bone marrow transplant for acute myelogenous leukaemia and 729 patients given a transplant for severe aplastic anaemia was conducted by the International Bone Marrow Transplant Registry. One aim of this study is to determine if the mortality rates of these patients return after some period of time to the same mortality rate as in the general population. To examine this question a model for the relative mortality of a bone marrow transplant patient relative to a matched individual in the general population is presented. This model allows for different relative mortality rates depending on the risk factors the patient may have. We discuss an estimation procedure for this model and construct a test that the mortality rate in the transplanted population is the same as in the reference population over a given time interval.

Object recognition: view-specificity and motion-specificity.

This paper describes an experiment to distinguish between two theories of human visual object recognition. According to the view-specificity hypothesis, object recognition is based on particular learned views, whereas the motion-specificity hypothesis states that object recognition depends on particular directed view-sequences. Both hypotheses imply a degree of view-bias (i.e. recognition of a given object is associated with a small number of views). Whereas the view-specificity hypothesis attributes this view-bias to a preference for particular views, the motion-specificity hypothesis attributes view-bias to a preference for particular directed view-sequences. Results presented here suggest that recognition of 3D rotating objects involves significant view-bias. This view-bias appears to be associated with an underlying bias for particular directed view-sequences, and not for particular views.

Temporal constraints on visual learning: a computational model.

Given a constant stream of perceptual stimuli, how can the underlying invariances associated with a given input be learned? One approach consists of using generic truths about the spatiotemporal structure of the physical world as constraints on the types of quantities learned. The learning methodology employed here embodies one such truth: that perceptually salient properties (such as stereo disparity) tend to vary smoothly over time. Unfortunately, the units of an artificial neural network tend to encode superficial image properties, such as individual grey-level pixel values, which vary rapidly over time. However, if the states of units are constrained to vary slowly, then the network is forced to learn a smoothly varying function of the training data. We implemented this temporal-smoothness constraint in a backpropagation network which learned stereo disparity from random-dot stereograms. Temporal smoothness was formalized with the use of regularization theory by modifying the standard cost function minimised during training of a network. Temporal smoothness was found to be similar to other techniques for improving generalisation, such as early stopping and weight decay. However, in contrast to these, the theoretical underpinnings of temporal smoothing are intimately related to fundamental characteristics of the physical world. Results are discussed in terms of regularization theory and the physically realistic assumptions upon which temporal smoothing is based.

Object recognition using spatiotemporal signatures.

The sequence of images generated by motion between observer and object specifies a spatiotemporal signature for that object. Evidence is presented that such spatiotemporal signatures are used in object recognition. Subjects learned novel, three-dimensional, rotating objects from image sequences in a continuous recognition task. During learning, the temporal order of images of a given object was constant. During testing, the order of images in each sequence was reversed, relative to its order during learning. This image sequence reversal produced significant reaction time increases and recognition rate decreases. Results are interpreted in terms of object-specific spatiotemporal signatures.

Learning perceptually salient visual parameters using spatiotemporal smoothness constraints.

A model is presented for unsupervised learning of low level vision tasks, such as the extraction of surface depth. A key assumption is that perceptually salient visual parameters (e.g., surface depth) vary smoothly over time. This assumption is used to derive a learning rule that maximizes the long-term variance of each unit's outputs, whilst simultaneously minimizing its short-term variance. The length of the half-life associated with each of these variances is not critical to the success of the algorithm. The learning rule involves a linear combination of anti-Hebbian and Hebbian weight changes, over short and long time scales, respectively. This maximizes the information throughput with respect to low-frequency parameters implicit in the input sequence. The model is used to learn stereo disparity from temporal sequences of random-dot and gray-level stereograms containing synthetically generated subpixel disparities. The presence of temporal discontinuities in disparity does not prevent learning or generalization to previously unseen image sequences. The implications of this class of unsupervised methods for learning in perceptual systems are discussed.

Learning stereo disparity using temporal smoothness constraints: a computational model.

An unsupervised learning algorithm is presented for learning stereo disparity. A key assumption is that surface depth varies smoothly over time. This assumption is consistent with a learning rule which maximizes the long-term variance of each unit's outputs, whilst simultaneously minimizing its short-term variance. The learning rule involves a linear combination of anti-Hebbian and Hebbian weight changes, over short and long time scales, respectively. The model is demonstrated on a hyperacuity task: estimating sub-pixel stereo disparity from a temporal sequence of stereograms. The algorithm generalizes, without additional learning, to previously unseen image sequences.

A canonical microfunction for learning perceptual invariances.

An unsupervised method is presented which permits a set of model neurons, or a microcircuit, to learn low-level vision tasks, such as the extraction of surface depth. Each micro-circuit implements a simple, generic strategy which is based on a key assumption: perceptually salient visual invariances, such as surface depth, vary smoothly over time. In the process of learning to extract smoothly varying invariances, each microcircuit maximises a microfunction. This is achieved by means of a learning rule which maximises the long-term variance of the state of a model neuron and simultaneously minimises its short-term variance. The learning rule involves a linear combination of anti-Hebbian and Hebbian weight changes, over short and long time scales, respectively. The method is demonstrated on a hyperacuity task: estimating subpixel stereo disparity from a temporal sequence of random-dot stereograms. After learning, the micro-circuit generalises, without additional learning, to previously unseen image sequences. It is proposed that the approach adopted here may be used to define a canonical microfunction, which can be used to learn many perceptually salient invariances.

Financial effect of clinical decisions: case study of a dialysis center.

The purpose of this study was to specify the financial effect of clinical decisions in a dialysis center. A consecutive sample of 14,343 outpatient hemodialysis treatments (OHD), 16,111 continuous ambulatory peritoneal dialysis (CAPD), and 4,513 chronic cycler-assisted peritoneal dialysis (CCPD) days of treatment was analyzed. An activity-based cost calculation method was applied to the analysis of alternative treatments (service bundles). The weekly cost of OHD was higher ($338 versus $241/$242), and the contribution margin (reimbursement minus total cost) of CAPD/CCPD was much greater ($.48 versus $148/$147 per patient week). Clinical decision-making had an influence on less than 6.8% of OHD and 45.4%/46.6% of CAPD/CCPD related expenses. In comparison to activity-based cost calculation, conventional methods overestimated the overhead expense of CAPD by 3-48%. This study documented that most cost control opportunities reside in the usual process of care and less can be influenced by a direct interference with the patient-physician contacts. Paying for 1 week of renal replacement (capitation) could simplify the process of reimbursement and cost tracking.

Triage in an established trauma system.

OBJECTIVE: The goal of this study was to determine patient and injury characteristics that predict undertriage and overtriage. DESIGN: This study was a retrospective analysis of admissions for acute injury. MATERIALS AND METHODS: All admissions for acute injuries in a 2 1/2-year period were included (N = 26,025). ICD-9 clinical modification codes were converted to Injury Severity Scores. MAIN RESULTS: Seventy-nine percent of severely injured patients were admitted to level I trauma centers. Severely injured patients admitted to other hospitals (undertriage) were more likely elderly (odds ratio = 5.44) and less likely had multisystem injuries (odds ratio = 0.55). One-fourth of patients with minor injuries were admitted to level I trauma centers (overtriage). Overtriaged patients were more likely intoxicated, obese, or had an injury to the head or face. CONCLUSIONS: In a developed trauma system, severely injured elderly trauma patients (especially females) are at risk for undertriage. The characteristics of patients at risk for overtriage reflect the difficulties of prospective out-of-hospital triage.