Generating neural circuits that implement probabilistic reasoning.

We extend the hypothesis that neuronal populations represent and process analog variables in terms of probability density functions (PDFs). Aided by an intermediate representation of the probability density based on orthogonal functions spanning an underlying low-dimensional function space, it is shown how neural circuits may be generated from Bayesian belief networks. The ideas and the formalism of this PDF approach are illustrated and tested with several elementary examples, and in particular through a problem in which model-driven top-down information flow influences the processing of bottom-up sensory input.

Neural representation of probabilistic information.

It has been proposed that populations of neurons process information in terms of probability density functions (PDFs) of analog variables. Such analog variables range, for example, from target luminance and depth on the sensory interface to eye position and joint angles on the motor output side. The requirement that analog variables must be processed leads inevitably to a probabilistic description, while the limited precision and lifetime of the neuronal processing units lead naturally to a population representation of information. We show how a time-dependent probability density rho(x; t) over variable x, residing in a specified function space of dimension D, may be decoded from the neuronal activities in a population as a linear combination of certain decoding functions phi(i)(x), with coefficients given by the N firing rates a(i)(t) (generally with D << N). We show how the neuronal encoding process may be described by projecting a set of complementary encoding functions phi;(i)(x) on the probability density rho(x; t), and passing the result through a rectifying nonlinear activation function. We show how both encoders phi;(i)(x) and decoders phi(i)(x) may be determined by minimizing cost functions that quantify the inaccuracy of the representation. Expressing a given computation in terms of manipulation and transformation of probabilities, we show how this representation leads to a neural circuit that can carry out the required computation within a consistent Bayesian framework, with the synaptic weights being explicitly generated in terms of encoders, decoders, conditional probabilities, and priors.

A general framework for neurobiological modeling: an application to the vestibular system.

The otolith organs in the vestibular system are excellent detectors of linear accelerations. However, any measurement of linear acceleration is ambiguous between a tilt in a gravitational field and an inertial acceleration. Angelaki et al. have put forward a general hypothesis about how inertial accelerations can be computed based on vestibular signals (J. Neurosci. 19 (1999) 316). We have constructed a realistic, detailed model of the relevant systems to test this hypothesis. The model produces useful predictions about what kinds of neurons should be found in the vestibular nucleus if such a computation is actually performed in the vestibular system. The model is constructed using general principles of neurobiological simulation (J. Neurophys. 84 (2000) 2113).

An integrated software suite for surface-based analyses of cerebral cortex.

The authors describe and illustrate an integrated trio of software programs for carrying out surface-based analyses of cerebral cortex. The first component of this trio, SureFit (Surface Reconstruction by Filtering and Intensity Transformations), is used primarily for cortical segmentation, volume visualization, surface generation, and the mapping of functional neuroimaging data onto surfaces. The second component, Caret (Computerized Anatomical Reconstruction and Editing Tool Kit), provides a wide range of surface visualization and analysis options as well as capabilities for surface flattening, surface-based deformation, and other surface manipulations. The third component, SuMS (Surface Management System), is a database and associated user interface for surface-related data. It provides for efficient insertion, searching, and extraction of surface and volume data from the database.

Image enhancement by nonlinear extrapolation in frequency space.

A technique for enhancing the perceptual sharpness of an image is described. The enhancement algorithm augments the frequency content of the image using shape-invariant properties of edges across scale by using a nonlinearity that generates phase coherent higher harmonics. The procedure utilizes the Laplacian transform and the Laplacian pyramid image representation. Results are presented depicting the power-spectra augmentation and the visual enhancement of several images. Simplicity of computations and ease of implementation allow for real-time applications such as high-definition television (HDTV).

Computation of optical flow using basis functions.

The issues governing the computation of optical flow in image sequences are addressed. The trade-off between accuracy versus computation cost is shown to be dependent on the redundancy of the image representation. This dependency is highlighted by reformulating Horn's (1986) algorithm, making explicit use of the approximations to the continuous basis functions underlying the discrete representation. The computation cost of estimating optical flow, for a fixed error tolerance, is shown to be a minimum for images resampled at twice the Nyquist rate. The issues of derivative calculation and multiresolution representation are also briefly discussed in terms of basis functions and information encoding. A multiresolution basis function formulation of Horn's algorithm is shown to lead to large improvements in dealing with high frequencies and large displacements.

Computerized mappings of the cerebral cortex: a multiresolution flattening method and a surface-based coordinate system.

We present a new method for generating two-dimensional maps of the cerebral cortex. Our computerized, two-stage flattening method takes as its input any well-defined representation of a surface within the three-dimensional cortex. The first stage rapidly converts this surface to a topologically correct two-dimensional map, without regard for the amount of distortion introduced. The second stage reduces distortions using a multiresolution strategy that makes gross shape changes on a coarsely sampled map and further shape refinements on progressively finer resolution maps. We demonstrate the utility of this approach by creating flat maps of the entire cerebral cortex in the macaque monkey and by displaying various types of experimental data on such maps. We also introduce a surface-based coordinate system that has advantages over conventional stereotaxic coordinates and is relevant to studies of cortical organization in humans as well as non-human primates. Together, these methods provide an improved basis for quantitative studies of individual variability in cortical organization.

The effect of oophorectomy and hormone replacement on neurohypophyseal hormone secretion in women.

OBJECTIVE: There is evidence to suggest that the release of neurohypophyseal hormones may be influenced by the circulating concentrations of gonadal steroids. We therefore monitored this relationship in women undergoing prophylactic bilateral oophorectomy at the time of hysterectomy with subsequent hormone replacement therapy and compared it with that in women undergoing hysterectomy with conservation of ovaries. DESIGN: Patients were randomly allocated to receive either transdermal oestradiol patches, 0.05 mg/day, or subdermal implants containing either 50 mg oestradiol or 50 mg oestradiol with 100 mg testosterone. Blood samples for determination of plasma hormone concentrations, electrolytes and osmolality were obtained immediately before and after surgery and then at two-monthly intervals for 8 months and finally at 12 months. MEASUREMENTS: Free oestradiol, vasopressin and oxytocin were measured by radioimmunoassay. RESULTS: Vasopressin concentrations were found to fall after surgery in oophorectomized women, but not in those with ovaries. There were no changes in fluid balance to account for the reduced plasma vasopressin concentrations. During treatment oestradiol appeared to enhance and testosterone to suppress vasopressin release. Oophorectomy had no significant effect on plasma oxytocin concentrations, but in the groups receiving oestradiol implants concentrations fell significantly at 8 and 12 months compared with the value at 6 days. CONCLUSION: The observed changes in plasma vasopressin concentrations were consistent with the observations in experimental animals and provide evidence that vasopressin release in the human is influenced by gonadal steroids.

A multiscale dynamic routing circuit for forming size- and position-invariant object representations.

We describe a neural model for forming size- and position-invariant representations of visual objects. The model is based on a previously proposed dynamic routing circuit that remaps selected portions of an input array into an object-centered reference frame. Here, we show how a multiscale representation may be incorporated at the input stage of the model, and we describe the control architecture and dynamics for a hierarchical, multistage routing circuit. Specific neurobiological substrates and mechanisms for the model are proposed, and a number of testable predictions are described.

In vitro Ca deposition by rat matrix vesicles: is the membrane association of alkaline phosphatase essential for matrix vesicle-mediated calcium deposition?

1. Phosphatidylinositol phospholipase C (PI-PLC) treatment of rachitic rat matrix vesicles (MVs) released about 80% of membrane-bound alkaline phosphatase (ALP), AMPase, PPiase into the media. 2. About 20% hydrolytic activity was not released from MV membranes by PI-PLC treatment. 3. SDS-polyacrylamide gel electrophoresis and Western blot analysis showed only one immunoreactive protein corresponding to the molecular weight of ALP present in the soluble fraction after PI-PLC treatment. 4. The specific activity of the released ALP was at least 5-fold higher than the residual activity. 5. After PI-PLC treatment, MVs also demonstrated an 80% reduction of AMP- or beta GP-dependent calcium deposition. 6. The soluble fraction containing 80% of ALP activity was unable to support calcium deposition. The mixing of the soluble and insoluble fractions after PI-PLC treatment failed to fully restore calcium-depositing activity.

A neurobiological model of visual attention and invariant pattern recognition based on dynamic routing of information.

We present a biologically plausible model of an attentional mechanism for forming position- and scale-invariant representations of objects in the visual world. The model relies on a set of control neurons to dynamically modify the synaptic strengths of intracortical connections so that information from a windowed region of primary visual cortex (V1) is selectively routed to higher cortical areas. Local spatial relationships (i.e., topography) within the attentional window are preserved as information is routed through the cortex. This enables attended objects to be represented in higher cortical areas within an object-centered reference frame that is position and scale invariant. We hypothesize that the pulvinar may provide the control signals for routing information through the cortex. The dynamics of the control neurons are governed by simple differential equations that could be realized by neurobiologically plausible circuits. In preattentive mode, the control neurons receive their input from a low-level "saliency map" representing potentially interesting regions of a scene. During the pattern recognition phase, control neurons are driven by the interaction between top-down (memory) and bottom-up (retinal input) sources. The model respects key neurophysiological, neuroanatomical, and psychophysical data relating to attention, and it makes a variety of experimentally testable predictions.

Estrous cycle modulation of O6-alkylguanine-DNA alkyltransferase expression in rat mammary epithelial cells.

N-methyl-N-nitrosourea (MNU)-induced rat mammary tumor incidence and tumor number per rat, is directly correlated with an increase in the circulating level of estrogen(s) at the time of carcinogen administration and subsequent mammary epithelial O6-methylguanine content. We report that, expression of O6-alkyltransferase (AGT) is also regulated by reproductive hormones in a tissue specific manner. The level of mammary epithelial cell AGT activity on estrus (0.47 pmol/mg protein) and proestrus (0.32) was significantly higher than on metestrus (0.14) (P < 0.05, estrus vs. metestrus). However, no change was observed in liver AGT activity (0.52 pmol/mg protein). In contrast, the mean level of AGT protein was not significantly different between tumors from rats injected with MNU on different days of the estrous cycle. In conclusion, the different tumor biologies resulting from carcinogen injection on different days of the estrous cycle may be partially explained by variation in levels of DNA repair activity. However, the cells in the resulting tumors did not continue an obligatory differential expression of the AGT activity consistent with their stage of initiation.

Carboxypeptidase M in brain and peripheral nerves.

Carboxypeptidase M (CPM), a plasma membrane-bound enzyme, cleaves C-terminal basic amino acids with a neutral pH optimum. We studied its distribution in human, baboon, and dog brain and in dog peripheral nerves. Areas were dissected, homogenized, centrifuged, and assayed for activity with dansyl-Ala-Arg. The corpus callosum and the pyramidal and optic tract were especially rich in CPM, whereas basal ganglia and cortex had low activity. The identity of the basic carboxypeptidase activity with CPM was shown by similarities in subcellular localization, membrane attachment, substrate hydrolysis, inhibition by a specific basic carboxypeptidase inhibitor, and cross-reaction with anti-human CPM antiserum. This antiserum immunoprecipitated an average of 85% of the activity in human and baboon brain and approximately 66% in dog brain. CPM co-purified with myelin extracted from the brain. Consistent with results obtained in placenta and cultured kidney cells, CPM in the brain appears to be membrane-bound via a phosphatidylinositol glycan anchor. In the peripheral nerves, the specific activity in dog sciatic nerve and in vagus was high (98 and 149 nmol/h/mg of protein, respectively). In immunohistochemical studies, glia in the brain, which appear to be oligodendrocytes or astrocytes, and the outer aspects of myelin sheaths and Schwann cells in sciatic and vagus nerves were stained. We conclude that in some areas of the CNS and the PNS, CPM is closely associated with myelin and myelin-forming cells. Northern blot analysis revealed the presence of mRNA coding for CPM in the brain, showing that the enzyme is indeed synthesized there.

Cellular kinetics of rat mammary gland terminal end bud epithelium exposed to N-methyl-N-nitrosourea in vivo.

Administration of the direct acting carcinogen N-methyl-N-nitrosourea (NMU) to 50-55-day-old virgin female rats on different days of the estrous cycle yields differential breast tumor biology (T. A. Ratko and C. W. Beattie, Cancer Res., 45: 3042-3047). One basis for these estrous cycle-dependent differences may be the duration of cell cycle stages of susceptible structures such as mammary terminal end buds or the quantity and duration of repair effected following adduct formation within these structures. The terminal end bud (TEB) epithelial cell cycle was characterized using pulse injections of [3H]thymidine (0.5 mCi/g body weight). On estrus, TEB epithelial cell cycle was significantly shorter (15.5 h) than on proestrus (19.9 h) and diestrus (18.8 h). The shorter duration in TEB cell cycle on estrus was likely due to a shorter TG1 (3-4 h) (P less than 0.05) since TS and TG2 did not differ between estrous cycle days. When NMU was injected 1 h after [3H]thymidine, the labeled mitotic wave within TEB of diestrus rats recovered approximately 2-3 h sooner than those given injections during proestrus (P less than 0.01), suggesting less initial damage or a slightly faster rate of DNA adduct repair. When [3H]thymidine was injected 1-5 days after the NMU, the percentage of labeled mitoses of rats given injections during diestrus and proestrus recovered to near normal 48 h after NMU, although the proportion of all cells labeled was still low compared to non-NMU-treated rats. The percentage of labeled mitoses and labeling of cells were normal 3 and 5 days after NMU. Rats receiving a carcinogenic but sublethal dose of NMU (5 mg/100 g body weight), followed by [3H]thymidine injection within 1 min, had one-half the intensity of thymidine incorporation into the terminal end bud DNA of non-NMU-treated rats. Unscheduled DNA synthesis was not demonstrable within the first 48 h following injection of NMU. The results support and extend the finding that rat mammary epithelial cell carcinogenicity of NMU is estrous cycle dependent and appears to be correlated with a differential response in the cell cycle of TEB (shorter at estrus) or delayed recovery in response to NMU (proestrus versus diestrus).

Information processing in the primate visual system: an integrated systems perspective.

The primate visual system contains dozens of distinct areas in the cerebral cortex and several major subcortical structures. These subdivisions are extensively interconnected in a distributed hierarchical network that contains several intertwined processing streams. A number of strategies are used for efficient information processing within this hierarchy. These include linear and nonlinear filtering, passage through information bottlenecks, and coordinated use of multiple types of information. In addition, dynamic regulation of information flow within and between visual areas may provide the computational flexibility needed for the visual system to perform a broad spectrum of tasks accurately and at high resolution.

Estrous cycle dependence of nitrosomethylurea (NMU)-induced preneoplastic lesions in rat mammary gland.

Virgin 50-55-day-old rats exhibiting regular estrous cycles were injected i.v. with the direct acting carcinogen 1-nitroso-1-methylurea (NMU) on the morning of proestrus, estrus, or diestrus. Rats were killed at weekly intervals following NMU administration to identify source and number of microscopically identifiable dysplasias. Terminal end bud (TEB) abnormalities appeared within 1 week following NMU administration, with a significantly greater number of abnormal TEBs in mammary glands of rats injected on proestrus (PE) and estrus (E) than on diestrus (DE). Ductal (DH) and ductal alveolar hyperplasias (DAH) and hyperplastic alveolar nodules (HAN) appeared during week 3, with significantly more of each type of lesion appearing by 6 weeks after NMU injection. HAN were most numerous in glands from rats injected on estrus. Adenocarcinomas arose from both the proximal and distal ductal network; at 10 and 12 weeks post NMU, significantly more tumors were found in rats injected on proestrus than diestrus and estrus. These results support the theory that the hormonal environment at the time of NMU administration significantly alters early development of mammary tumors in the rat.

Estrous cycle status alters N-methyl-N-nitrosourea (NMU)-induced rat mammary tumor growth and regression.

The relationship between mammary carcinoma growth, ovariectomy-induced regression and estrogen receptor status were determined in Sprague-Dawley rats with 5-day estrous cycles after injection of N-methyl-N-nitrosourea (NMU) on metestrus (ME), diestrus-1 (DE-1), proestrus (PE) or estrus (E). Rats exposed to NMU on PE had a shorter tumor latency than those injected on ME and E, as well as more carcinomas per rat than those exposed on ME and DE-1. Mammary carcinomas grew faster in rats injected on ME (doubling time, 6.4 days) and DE-1 (6.9 days) compared with PE (15.2 days) and E (16.3 days). Tumor regression was also significantly faster in rats injected on ME (time to 50% vol., 5.5 days) and DE-1 (5.3 days) compared with PE (8.2 days) and E (8.5 days) following bilateral-ovariectomy during log phase growth. Significantly, total nuclear estrogen receptor (ERN) content was increased in carcinomas from rats injected on PE compared with DE-1 (70.8 +/- 11.3 vs. 32.9 +/- 7.3 fm/mg DNA) (P less than 0.05) and DE-1 and ME combined (P less than 0.01). These observations generalize the concept that estrous cycle stage at the time of NMU injection alters subsequent mammary carcinoma biology, and represents the first experimental evidence that slower growing and responding estrogen receptor positive rat mammary carcinomas may be associated with an increase in circulating estrogen prior to carcinogen exposure.

Shifter circuits: a computational strategy for dynamic aspects of visual processing.

We propose a general strategy for dynamic control of information flow between arrays of neurons at different levels of the visual pathway, starting in the lateral geniculate nucleus and the geniculorecipient layers of cortical area V1. This strategy can be used for resolving computational problems arising in the domains of stereopsis, directed visual attention, and the perception of moving images. In each of these situations, some means of dynamically controlling how retinal outputs map onto higher-level targets is desirable--in order to achieve binocular fusion, to allow shifts of the focus of attention, and to prevent blurring of moving images. The proposed solution involves what we term "shifter circuits," which allow for dynamic shifts in the relative alignment of input and output arrays without loss of local spatial relationships. The shifts are produced in increments along a succession of relay stages that are linked by diverging excitatory inputs. The direction of shift is controlled at each stage by inhibitory neurons that selectively suppress appropriate sets of ascending inputs. The shifter hypothesis is consistent with available anatomical and physiological evidence on the organization of the primate visual pathway, and it offers a sensible explanation for a variety of otherwise puzzling facts, such as the plethora of cells in the geniculorecipient layers of V1.