The Most Informative Neural Code Accounts For Population Heterogeneity

Population Heterogeneity ◽

Motion Direction ◽

Temporal Area ◽

Firing Rates ◽

Preferred Direction

AbstractPerception is produced by ‘reading out’ the representation of a sensory stimulus contained in the firing rates of a population of neurons. To examine experimentally how populations code information, a common approach is to decode a linearly-weighted sum of the neurons’ firing rates. This approach is popular because of its biological validity: weights in a computational decoder are analogous to synaptic strengths. For neurons recorded in vivo, weights are highly variable when derived through machine learning methods, but it is unclear what neuronal properties explain this variability, and how the variability affects decoding performance. To address this, we recorded from neurons in the middle temporal area (MT) of anesthetized marmosets (Callithrix jacchus) viewing stimuli comprising a sheet of dots that moved coherently in one of twelve different directions. We found that high gain and direction selectivity both predicted that a neuron would be weighted more highly in an optimised decoding model. Although learned weights differed markedly from weights chosen according to a priori rules based on a neuron’s tuning profile, decoding performance was only marginally better for the learned weights. In the models with a priori rules, selectivity is the best predictor of weighting, and defining weights according to a neuron’s preferred direction and selectivity improves decoding performance to very near the maximum level possible, as defined by the learned weights.New & NoteworthyWe examined which aspects of a neuron’s tuning account for its contribution to sensory coding. Strongly direction-selective neurons were weighted most highly by machine learning algorithms trained to discriminate motion direction. Models with a priori defined decoding weights demonstrate that the learned weighting scheme causally improved direction representation by a neuronal population. Optimising decoders (using machine learning) lead to only marginally better performance than decoders based purely on a neuron’s preferred direction and selectivity.

Weighting neurons by selectivity produces near-optimal population codes

10.1152/jn.00504.2018 ◽

2019 ◽

Vol 121 (5) ◽

pp. 1924-1937

Author(s):

Elizabeth Zavitz ◽

Nicholas S. C. Price

Keyword(s):

A Priori ◽

Maximum Level ◽

Optimization Methods ◽

Neuronal Population ◽

Direction Selectivity ◽

Motion Direction ◽

Temporal Area ◽

Linear Discriminant ◽

Preferred Direction

Perception is produced by “reading out” the representation of a sensory stimulus contained in the activity of a population of neurons. To examine experimentally how populations code information, a common approach is to decode a linearly weighted sum of the neurons’ spike counts. This approach is popular because of the biological plausibility of weighted, nonlinear integration. For neurons recorded in vivo, weights are highly variable when derived through optimization methods, but it is unclear how the variability affects decoding performance in practice. To address this, we recorded from neurons in the middle temporal area (MT) of anesthetized marmosets ( Callithrix jacchus) viewing stimuli comprising a sheet of dots that moved coherently in 1 of 12 different directions. We found that high peak response and direction selectivity both predicted that a neuron would be weighted more highly in an optimized decoding model. Although learned weights differed markedly from weights chosen according to a priori rules based on a neuron’s tuning profile, decoding performance was only marginally better for the learned weights. In the models with a priori rules, selectivity is the best predictor of weighting, and defining weights according to a neuron’s preferred direction and selectivity improves decoding performance to very near the maximum level possible, as defined by the learned weights. NEW & NOTEWORTHY We examined which aspects of a neuron’s tuning account for its contribution to sensory coding. Strongly direction-selective neurons are weighted most highly by optimal decoders trained to discriminate motion direction. Models with predefined decoding weights demonstrate that this weighting scheme causally improved direction representation by a neuronal population. Optimizing decoders (using a generalized linear model or Fisher’s linear discriminant) led to only marginally better performance than decoders based purely on a neuron’s preferred direction and selectivity.

Brain tumor diagnostic model and dietary effect based on extracellular vesicle microbiome data in serum

Experimental & Molecular Medicine ◽

10.1038/s12276-020-00501-x ◽

2020 ◽

Vol 52 (9) ◽

pp. 1602-1613

Author(s):

Jinho Yang ◽

Hyo Eun Moon ◽

Hyung Woo Park ◽

Andrea McDowell ◽

Tae-Seop Shin ◽

...

Keyword(s):

Machine Learning ◽

Brown Rice ◽

Metagenomic Data ◽

Diagnostic Model ◽

Tissue Samples ◽

In Vivo Models ◽

Linear Discriminant ◽

Difficult Condition

Abstract The human microbiome has been recently associated with human health and disease. Brain tumors (BTs) are a particularly difficult condition to directly link to the microbiome, as microorganisms cannot generally cross the blood–brain barrier (BBB). However, some nanosized extracellular vesicles (EVs) released from microorganisms can cross the BBB and enter the brain. Therefore, we conducted metagenomic analysis of microbial EVs in both serum (152 BT patients and 198 healthy controls (HC)) and brain tissue (5 BT patients and 5 HC) samples based on the V3–V4 regions of 16S rDNA. We then developed diagnostic models through logistic regression and machine learning algorithms using serum EV metagenomic data to assess the ability of various dietary supplements to reduce BT risk in vivo. Models incorporating the stepwise method and the linear discriminant analysis effect size (LEfSe) method yielded 12 and 29 significant genera as potential biomarkers, respectively. Models using the selected biomarkers yielded areas under the curves (AUCs) >0.93, and the model using machine learning resulted in an AUC of 0.99. In addition, Dialister and [Eubacterium] rectale were significantly lower in both blood and tissue samples of BT patients than in those of HCs. In vivo tests showed that BT risk was decreased through the addition of sorghum, brown rice oil, and garlic but conversely increased by the addition of bellflower and pear. In conclusion, serum EV metagenomics shows promise as a rich data source for highly accurate detection of BT risk, and several foods have potential for mitigating BT risk.

Comparison of the Spatial Limits on Direction Selectivity in Visual Areas MT and V1

10.1152/jn.00767.2004 ◽

2005 ◽

Vol 93 (3) ◽

pp. 1235-1245 ◽

Cited By ~ 25

Author(s):

Mark M. Churchland ◽

Nicholas J. Priebe ◽

Stephen G. Lisberger

Keyword(s):

Apparent Motion ◽

Great Majority ◽

Spatial Separation ◽

Direction Selectivity ◽

Temporal Separation ◽

Temporal Area ◽

Mean Values ◽

V1 Neurons ◽

Preferred Direction ◽

The Difference

We recorded responses to apparent motion from directionally selective neurons in primary visual cortex (V1) of anesthetized monkeys and middle temporal area (MT) of awake monkeys. Apparent motion consisted of multiple stationary stimulus flashes presented in sequence, characterized by their temporal separation (Δ t) and spatial separation (Δ x). Stimuli were 8° square patterns of 100% correlated random dots that moved at apparent speeds of 16 or 32°/s. For both V1 and MT, the difference between the response to the preferred and null directions declined with increasing flash separation. For each neuron, we estimated the maximum flash separation for which directionally selective responses were observed. For the range of speeds we used, Δ x provided a better description of the limitation on directional responses than did Δ t. When comparing MT and V1 neurons of similar preferred speed, there was no difference in the maximum Δ x between our samples from the two areas. In both V1 and MT, the great majority of neurons had maximal values of Δ x in the 0.25–1° range. Mean values were almost identical between the two areas. For most neurons, larger flash separations led to both weaker responses to the preferred direction and increased responses to the opposite direction. The former mechanism was slightly more dominant in MT and the latter slightly more dominant in V1. We conclude that V1 and MT neurons lose direction selectivity for similar values of Δ x, supporting the hypothesis that basic direction selectivity in MT is inherited from V1, at least over the range of stimulus speeds represented by both areas.

Neural responses to visual texture patterns in middle temporal area of the macaque monkey

10.1152/jn.1992.68.1.164 ◽

1992 ◽

Vol 68 (1) ◽

pp. 164-181 ◽

Cited By ~ 49

Author(s):

J. F. Olavarria ◽

E. A. DeYoe ◽

J. J. Knierim ◽

J. M. Fox ◽

D. C. van Essen

Keyword(s):

Macaque Monkey ◽

Direction Selectivity ◽

Neural Responses ◽

Temporal Area ◽

Middle Temporal ◽

Tuning Curves ◽

Preferred Direction ◽

Orientation Contrast ◽

Background Texture ◽

Moving Texture

1. We studied how neurons in the middle temporal visual area (MT) of anesthetized macaque monkeys responded to textured and nontextured visual stimuli. Stimuli contained a central rectangular ,figure- that was either uniform in luminance or consisted of an array of oriented line segments. The figure moved at constant velocity in one of four orthogonal directions. The region surrounding the figure was either uniform in luminance or contained a texture array (whose elements were identical or orthogonal in orientation to those of the figure), and it either was stationary or moved along with the figure. 2. A textured figure moving across a stationary textured background (,texture bar- stimulus) often elicited vigorous neural responses, but, on average, the responses to texture bars were significantly smaller than to solid (uniform luminance) bars. 3. Many cells showed direction selectivity that was similar for both texture bars and solid bars. However, on average, the direction selectivity measured when texture bars were used was significantly smaller than that for solid bars, and many cells lost significant direction selectivity altogether. The reduction in direction selectivity for texture bars generally reflected a combination of decreased responsiveness in the preferred direction and increased responsiveness in the null (opposite to preferred) direction. 4. Responses to a texture bar in the absence of a texture background (,texture bar alone-) were very similar to the responses to solid bars both in the magnitude of response and in the degree of direction selectivity. Conversely, adding a static texture surround to a moving solid bar reduced direction selectivity on average without a reduction in response magnitude. These results indicate that the static surround is largely responsible for the differences in direction selectivity for texture bars versus solid bars. 5. In the majority of MT cells studied, responses to a moving texture bar were largely independent of whether the elements in the bar were of the same orientation as the background elements or of the orthogonal orientation. Thus, for the class of stimuli we used, orientation contrast does not markedly affect the responses of MT neurons to moving texture patterns. 6. The optimum figure length and the shapes of the length tuning curves determined with the use of solid bars and texture bars differed significantly in most of the cells examined. Thus neurons in MT are not simply selective for a particular figure shape independent of whatever cues are used to delineate the figure.

Advances in Deep Neuropathological Phenotyping of Alzheimer Disease: Past, Present, and Future

Journal of Neuropathology & Experimental Neurology ◽

10.1093/jnen/nlab122 ◽

2022 ◽

Author(s):

Mustafa N Shakir ◽

Brittany N Dugger

Keyword(s):

Machine Learning ◽

Alzheimer Disease ◽

Neurodegenerative Disorder ◽

Lewy Body Disease ◽

Brain Diseases ◽

Disease Heterogeneity ◽

Technological Advances ◽

Alois Alzheimer

Abstract Alzheimer disease (AD) is a neurodegenerative disorder characterized pathologically by the presence of neurofibrillary tangles and amyloid beta (Aβ) plaques in the brain. The disease was first described in 1906 by Alois Alzheimer, and since then, there have been many advancements in technologies that have aided in unlocking the secrets of this devastating disease. Such advancements include improving microscopy and staining techniques, refining diagnostic criteria for the disease, and increased appreciation for disease heterogeneity both in neuroanatomic location of abnormalities as well as overlap with other brain diseases; for example, Lewy body disease and vascular dementia. Despite numerous advancements, there is still much to achieve as there is not a cure for AD and postmortem histological analyses is still the gold standard for appreciating AD neuropathologic changes. Recent technological advances such as in-vivo biomarkers and machine learning algorithms permit great strides in disease understanding, and pave the way for potential new therapies and precision medicine approaches. Here, we review the history of human AD neuropathology research to include the notable advancements in understanding common co-pathologies in the setting of AD, and microscopy and staining methods. We also discuss future approaches with a specific focus on deep phenotyping using machine learning.

Dendritic Calcium Accumulation Associated With Direction-Selective Adaptation in Visual Motion-Sensitive Neurons In Vivo

10.1152/jn.2000.84.4.1914 ◽

2000 ◽

Vol 84 (4) ◽

pp. 1914-1923 ◽

Cited By ~ 26

Author(s):

Rafael Kurtz ◽

Volker Dürr ◽

Martin Egelhaaf

Keyword(s):

Visual Motion ◽

Selective Adaptation ◽

Direction Selectivity ◽

Motion Adaptation ◽

Stimulus Velocity ◽

Calcium Accumulation ◽

Calcium Dependent ◽

Preferred Direction ◽

Cell Class

Motion adaptation in directionally selective tangential cells (TC) of the fly visual system has previously been explained as a presynaptic mechanism. Based on the observation that adaptation is in part direction selective, which is not accounted for by the former models of motion adaptation, we investigated whether physiological changes located in the TC dendrite can contribute to motion adaptation. Visual motion in the neuron's preferred direction (PD) induced stronger adaptation than motion in the opposite direction and was followed by an afterhyperpolarization (AHP). The AHP subsides in the same time as adaptation recovers. By combining in vivo calcium fluorescence imaging with intracellular recording, we show that dendritic calcium accumulation following motion in the PD is correlated with the AHP. These results are consistent with a calcium-dependent physiological change in TCs underlying adaptation during continuous stimulation with PD motion, expressing itself as an AHP after the stimulus stops. However, direction selectivity of adaptation is probably not solely related to a calcium-dependent mechanism because direction-selective effects can also be observed for fast moving stimuli, which do not induce sizeable calcium accumulation. In addition, a comparison of two classes of TCs revealed differences in the relationship of calcium accumulation and AHP when the stimulus velocity was varied. Thus the potential role of calcium in motion adaptation depends on stimulation parameters and cell class.

Computation of motion direction in the vertebrate retina

e-Neuroforum ◽

10.1007/s13295-012-0033-x ◽

2012 ◽

Vol 18 (3) ◽

Cited By ~ 1

Author(s):

T. Euler ◽

S.E. Hausselt

Keyword(s):

Large Scale ◽

Petri Dish ◽

Cell Types ◽

Direction Selectivity ◽

Dendritic Morphology ◽

Image Motion ◽

Target Area ◽

Motion Direction ◽

Preferred Direction ◽

The Brain

AbstractHow direction of image motion is detected as early as at the level of the vertebrate eye has been intensively studied in retina research. Although the first direction-selective (DS) retinal ganglion cells were already described in the 1960s and have since then been in the focus of many studies, scientists are still puzzled by the intricacy of the neuronal circuits and computational mechanisms underlying retinal direction selectivity. The fact that the retina can be easily isolated and studied in a Petri dish-by presenting light stimuli while recording from the various cell types in the retinal circuits-in combination with the extensive anatomical, molecular and physiological knowledge about this part of the brain presents a unique opportunity for studying this intriguing visual circuit in detail. This article provides a brief overview of the history of research on retinal direction selectivity, but then focuses on the past decade and the progress achieved, in particular driven by methodological advances in optical recording techniques, molecular genetics approaches and large-scale ultrastructural reconstructions. As it turns out, retinal direction selectivity is a complex, multi-tiered computation, involving dendrite-intrinsic mechanisms as well as several types of network interactions on the basis of highly selective, likely genetically predetermined synaptic connectivity. Moreover, DS ganglion cell types appear to be more diverse than previously thought, differing not only in their preferred direction and response polarity, but also in physiology, DS mechanism, dendritic morphology and, importantly, the target area of their projections in the brain.

Machine Learning Approaches in Parkinson’s Disease

Current Medicinal Chemistry ◽

10.2174/0929867328999210111211420 ◽

2021 ◽

Vol 28 ◽

Author(s):

Annamaria Landolfi ◽

Carlo Ricciardi ◽

Leandro Donisi ◽

Giuseppe Cesarelli ◽

Jacopo Troisi ◽

...

Keyword(s):

Machine Learning ◽

Parkinson’S Disease ◽

Parkinson's Disease ◽

Neurodegenerative Disorder ◽

Learning Algorithms ◽

Disease Diagnosis ◽

Clinical Aspects ◽

Learning Approaches

Background:: Parkinson’s disease is the second most frequent neurodegenerative disorder. Its diagnosis is challenging and mainly relies on clinical aspects. At present, no biomarker is available to obtain a diagnosis of certainty in vivo. Objective:: The present review aims at describing machine learning algorithms as they have been variably applied to different aspects of Parkinson’s disease diagnosis and characterization. Methods:: A systematic search was conducted on PubMed in December 2019, resulting in 230 publications obtained with the following search query: “Machine Learning” “AND” “Parkinson Disease”. Results:: the obtained publications were divided into 6 categories, based on different application fields: “Gait Analysis - Motor Evaluation”, “Upper Limb Motor and Tremor Evaluation”, “Handwriting and typing evaluation”, “Speech and Phonation evaluation”, “Neuroimaging and Nuclear Medicine evaluation”, “Metabolomics application”, after excluding the papers of general topic. As a result, a total of 166 articles were analyzed, after elimination of papers written in languages other than English or not directly related to the selected topics. Conclusion:: Machine learning algorithms are computer-based statistical approaches which can be trained and are able to find common patterns from big amounts of data. The machine learning approaches can help clinicians in classifying patients according to several variables at the same time.

Mechanism of Motion Direction Detection Based on Barlow’s Retina Inhibitory Scheme in Direction-Selective Ganglion Cells

Electronics ◽

10.3390/electronics10141663 ◽

2021 ◽

Vol 10 (14) ◽

pp. 1663

Author(s):

Mianzhe Han ◽

Yuki Todo ◽

Zheng Tang

Keyword(s):

Ganglion Cells ◽

Direction Selectivity ◽

Detection Accuracy ◽

Local Motion ◽

Motion Direction ◽

Preferred Direction ◽

Directional Detection ◽

Direction Detection ◽

Series Of Experiments ◽

Better Than

Previous studies have reported that directionally selective ganglion cells respond strongly in their preferred direction, but are only weakly excited by stimuli moving in the opposite null direction. Various studies have attempted to elucidate the mechanisms underlying direction selectivity with cellular basis. However, these studies have not elucidated the mechanism underlying motion direction detection. In this study, we propose the mechanism based on Barlow’s inhibitory scheme for motion direction detection. We described the local motion-sensing direction-selective neurons. Next, this model was used to construct the two-dimensional multi-directional detection neurons which detect the local motion directions. The information of local motion directions was finally used to infer the global motion direction. To verify the validity of the proposed mechanism, we conducted a series of experiments involving a dataset with a number of images. The proposed mechanism exhibited good performance in all experiments with high detection accuracy. Furthermore, we compare the performance of our proposed system and traditional Convolution Neural Network (CNN) on motion direction prediction. It is found that the performance of our system is much better than that of CNN in terms of accuracy, calculation speed and cost.

Attack and defence in cellular decision-making: lessons from machine learning

10.1101/366724 ◽

2018 ◽

Author(s):

Thomas J. Rademaker ◽

Emmanuel Bengio ◽

Paul François

Keyword(s):

Machine Learning ◽

Neural Networks ◽

Decision Making ◽

Critical Point ◽

Nonlinear Models ◽

Immune Recognition ◽

Cellular Decision Making ◽

Decision Making Models

Machine learning algorithms can be fooled by small well-designed adversarial perturbations. This is reminiscent of cellular decision-making where ligands (called antagonists) prevent correct signalling, like in early immune recognition. We draw a formal analogy between neural networks used in machine learning and models of cellular decision-making (adaptive proofreading). We apply attacks from machine learning to simple decision-making models, and show explicitly the correspondence to antagonism by weakly bound ligands. Such antagonism is absent in more nonlinear models, which inspired us to implement a biomimetic defence in neural networks filtering out adversarial perturbations. We then apply a gradient-descent approach from machine learning to different cellular decision-making models, and we reveal the existence of two regimes characterized by the presence or absence of a critical point for the gradient. This critical point causes the strongest antagonists to lie close to the decision boundary. This is validated in the loss landscapes of robust neural networks and cellular decision-making models, and observed experimentally for immune cells. For both regimes, we explain how associated defence mechanisms shape the geometry of the loss landscape, and why different adversarial attacks are effective in different regimes. Our work connects evolved cellular decision-making to machine learning, and motivates the design of a general theory of adversarial perturbations, both for in vivo and in silico systems.