Separation of signal and noise from in vivo optical recording in Guinea pigs using independent component analysis

Evaluation and Optimization of Methods for Generating High-Resolution Retinotopic Maps Using Visual Cortex Voltage-Sensitive Dye Imaging

The localization and measurement of neuronal activity magnitude at high spatial and temporal resolution are essential for mapping and better understanding neuronal systems and mechanisms. One such example is the generation of retinotopic maps, which correlates localized retinal stimulation with the corresponding specific visual cortex responses. Here we evaluated and compared seven different methods for extracting and localizing cortical responses from voltage-sensitive dye imaging recordings, elicited by visual stimuli projected directly on the rat retina by a customized projection system. The performance of these methods was evaluated both qualitatively and quantitatively by means of two cluster separation metrics, namely, the (adjusted) Silhouette Index (SI) and the (adjusted) Davies-Bouldin Index (DBI). These metrics were validated using simulated data, which showed that Temporally Structured Component Analysis (TSCA) outperformed all other analysis methods for localizing cortical responses and generating high-resolution retinotopic maps. The analysis methods, as well as the use of cluster separation metrics proposed here, can facilitate future research aiming to localize specific activity at high resolution in the visual cortex or other brain areas.

Separable gain control of ongoing and evoked activity in the visual cortex by serotonergic input

Controlling gain of cortical activity is essential to modulate weights between internal ongoing communication and external sensory drive. Here, we show that serotonergic input has separable suppressive effects on the gain of ongoing and evoked visual activity. We combined optogenetic stimulation of the dorsal raphe nucleus (DRN) with wide-field calcium imaging, extracellular recordings, and iontophoresis of serotonin (5-HT) receptor antagonists in the mouse visual cortex. 5-HT1A receptors promote divisive suppression of spontaneous activity, while 5-HT2A receptors act divisively on visual response gain and largely account for normalization of population responses over a range of visual contrasts in awake and anesthetized states. Thus, 5-HT input provides balanced but distinct suppressive effects on ongoing and evoked activity components across neuronal populations. Imbalanced 5-HT1A/2A activation, either through receptor-specific drug intake, genetically predisposed irregular 5-HT receptor density, or change in sensory bombardment may enhance internal broadcasts and reduce sensory drive and vice versa.

Improving voltage-sensitive dye imaging: with a little help from computational approaches

Voltage-sensitive dye imaging (VSDI) is a key neurophysiological recording tool because it reaches brain scales that remain inaccessible to other techniques. The development of this technique from in vitro to the behaving nonhuman primate has only been made possible thanks to the long-lasting, visionary work of Amiram Grinvald. This work has opened new scientific perspectives to the great benefit to the neuroscience community. However, this unprecedented technique remains largely under-utilized, and many future possibilities await for VSDI to reveal new functional operations. One reason why this tool has not been used extensively is the inherent complexity of the signal. For instance, the signal reflects mainly the subthreshold neuronal population response and is not linked to spiking activity in a straightforward manner. Second, VSDI gives access to intracortical recurrent dynamics that are intrinsically complex and therefore nontrivial to process. Computational approaches are thus necessary to promote our understanding and optimal use of this powerful technique. Here, we review such approaches, from computational models to dissect the mechanisms and origin of the recorded signal, to advanced signal processing methods to unravel new neuronal interactions at mesoscopic scale. Only a stronger development of interdisciplinary approaches can bridge micro- to macroscales.

Spatially Structured Sparse Morphological Component Separation for voltage-sensitive dye optical imaging

BACKGROUND

Voltage-sensitive dye optical imaging is a promising technique for studying in vivo neural assemblies dynamics where functional clustering can be visualized in the imaging plane. Its practical potential is however limited by many artifacts.

NEW METHOD

We present a novel method, that we call "SMCS" (Spatially Structured Sparse Morphological Component Separation), to separate the relevant biological signal from noise and artifacts. It extends Generalized Linear Models (GLM) by using a set of convex non-smooth regularization priors adapted to the morphology of the sources and artifacts to capture.

RESULTS

We make use of first order proximal splitting algorithms to solve the corresponding large scale optimization problem. We also propose an automatic parameters selection procedure based on statistical risk estimation methods.

COMPARISON WITH EXISTING METHODS

We compare this method with blank subtraction and GLM methods on both synthetic and real data. It shows encouraging perspectives for the observation of complex cortical dynamics.

CONCLUSIONS

This work shows how recent advances in source separation can be integrated into a biophysical model of VSDOI. Going beyond GLM methods is important to capture transient cortical events such as propagating waves.

Vobi One: a data processing software package for functional optical imaging

Optical imaging is the only technique that allows to record the activity of a neuronal population at the mesoscopic scale. A large region of the cortex (10-20 mm diameter) is directly imaged with a CCD camera while the animal performs a behavioral task, producing spatio-temporal data with an unprecedented combination of spatial and temporal resolutions (respectively, tens of micrometers and milliseconds). However, researchers who have developed and used this technique have relied on heterogeneous software and methods to analyze their data. In this paper, we introduce Vobi One, a software package entirely dedicated to the processing of functional optical imaging data. It has been designed to facilitate the processing of data and the comparison of different analysis methods. Moreover, it should help bring good analysis practices to the community because it relies on a database and a standard format for data handling and it provides tools that allow producing reproducible research. Vobi One is an extension of the BrainVISA software platform, entirely written with the Python programming language, open source and freely available for download at https://trac.int.univ-amu.fr/vobi_one.

Thoracotomy reduces intrinsic brain movement caused by heartbeat and respiration: a simple method to prevent motion artifact for in vivo experiments

Recent technical advances in electrophysiological recording and functional imaging from the brain of living animals have promoted our understandings of the brain function, but these in vivo experiments are still technically demanding and often suffer from spontaneous pulsation, i.e., brain movements caused by respiration and heartbeat. Here we report that thoracotomy suppresses the motion artifact to a practically negligible level. This simple method will be useful in a wide variety of in vivo experiments, such as patch-clamp physiology, and optical imaging of neurons, glial cell, and blood vessels.

Study of the origin of short- and long-latency SSEP during recovery from brain ischemia in a rat model

Somatosensory evoked potentials (SSEPs) have been established as an electrophysiological tool for the prognostication of neurological outcome in patients with hypoxic-ischemic brain injury. The early and late responses in SSEPs reflect the sequential activation of neural structures along the somatosensory pathway. This study reports that the SSEP can be separated into early (short-latency, SL) and late (long-latency, LL) responses using independent component analysis (ICA), based on the assumption that these components are generated from different neural sources. Moreover, this source separation into the SL and LL components allows analysis of electrophysiological response to brain injury, even when the SSEPs are severely distorted and SL and LL components get mixed. With the help of ICA decomposition and corrected peak estimation, the latency of LL-SSEP is shown to be predictive of long-term neurological outcome. Further, it is shown that the recovery processes of SL- and LL-SSEPs follow different dynamics, with the SL-SSEP restored earlier than LL-SSEP. We predict that the SL- and LL-SSEPs reflect the timing of the progression of evoked response through the thalamocortical pathway and as such respond differently depending upon injury and recovery of the thalamic and cortical regions, respectively.

Spontaneous activity resembling tone-evoked activity in the primary auditory cortex of guinea pigs

In the primary auditory cortex (AI), a pure tone evokes propagating activity along a strip of the cortex. We have previously shown that focal activation of AI triggers autonomously propagating activity that resembles tone-evoked activity (Song et al., 2006). Because a focal spontaneous activity is expected to trigger similar activity propagation, spontaneous activity resembling tone-evoked activity may exist in AI. Here we tested this possibility by optical imaging of AI in guinea pigs. After obtaining tone-evoked activities, we made long-duration optical recordings (9-40s) and isolated spontaneous activities from respiration and heartbeat noises using independent component analyses. Spontaneous activities were found all over AI, in all animals examined. Of all spontaneous events, 33.6% showed significant correlation in spatio-temporal pattern with tone-evoked activities. Simulation using a model that captures the temporal feature of spontaneous response in single channels but sets no constraint among channels, generated no spontaneous events that resembled tone-evoked activations. These results show the existence of spontaneous events similar in spatio-temporal pattern to tone-evoked activations in AI. Such spontaneous events are likely a manifestation of cortical structures that govern the pattern of distributed activation in AI.

Percutaneous fiber-optic sensor for chronic glucose monitoring in vivo

We are developing a family of fiber-optic sensors called Sencils (sensory cilia), which are disposable, minimally invasive, and can provide in vivo monitoring of various analytes for several weeks. The key element is a percutaneous optical fiber that permits reliable spectroscopic measurement of chemical reactions in a nano-engineered polymeric matrix attached to the implanted end of the fiber. This paper describes its first application to measure interstitial glucose based on changes in fluorescence resonance energy transfer (FRET) between fluorophores bound to betacyclodextrin and Concanavalin A (Con A) in a polyethylene glycol (PEG) matrix. In vitro experiments demonstrate a rapid and precise relationship between the ratio of the two fluorescent emissions and concentration of glucose in saline for the physiological range of concentrations (0-500mg/dl) over seven weeks. Chronic animal implantation studies have demonstrated good biocompatibility and durability for clinical applications.

Independent component analysis of high-resolution imaging data identifies distinct functional domains

In the vertebrate brain external stimuli are often represented in distinct functional domains distributed across the cortical surface. Fast imaging techniques used to measure patterns of population activity record movies with many pixels and many frames, i.e., data sets with high dimensionality. Here we demonstrate that principal component analysis (PCA) followed by spatial independent component analysis (sICA), can be exploited to reduce the dimensionality of data sets recorded in the olfactory bulb and the somatosensory cortex of mice as well as the visual cortex of monkeys, without loosing the stimulus-specific responses. Different neuronal populations are separated based on their stimulus-specific spatiotemporal activation. Both, spatial and temporal response characteristics can be objectively obtained, simultaneously. In the olfactory bulb, groups of glomeruli with different response latencies can be identified. This is shown for recordings of olfactory receptor neuron input measured with a calcium-sensitive axon tracer and for network dynamics measured with the voltage-sensitive dye RH 1838. In the somatosensory cortex, barrels responding to the stimulation of single whiskers can be automatically detected. In the visual cortex orientation columns can be extracted. In all cases artifacts due to movement, heartbeat or respiration were separated from the functional signal by sICA and could be removed from the data set. sICA following PCA is therefore a powerful technique for data compression, unbiased analysis and dissection of imaging data of population activity, collected with high spatial and temporal resolution.

Cortical Intrinsic Circuits Can Support Activity Propagation through an Isofrequency Strip of the Guinea Pig Primary Auditory Cortex

A pure tone evokes propagating activities in a strip of the primary auditory cortex (AI), an isofrequency strip (IS). A fundamental issue concerns the roles that thalamocortical input and intracortical connectivity play in generating the activities. Here we addressed this issue in guinea pigs using in vivo and in vitro real-time optical imaging techniques. As reported previously, tone-evoked activity propagated dorsoventrally along a strip (an IS) in AI. We found that an electrical pulse applied focally within the strip, triggered activity propagation with a spatiotemporal pattern highly similar to tone-evoked activation. The propagation velocity of electrically evoked activity was significantly slower than that of tone-evoked activity, but was comparable to the velocity of lateral activity propagation in cortical slices, suggesting that the electrically evoked activity propagation in vivo is mediated by intracortical circuits. To test this notion, we lesioned the auditory thalamus chemically; in such animals, electrically evoked activity in AI was not affected, although tone-evoked activity was abolished. Further, in slices of the AI, the extent of electrically evoked activity propagation in layer II/III was significantly larger in coronal slices than in horizontal slices. Together, our results suggest that intracortical connectivity in AI enables a focally evoked activity to propagate throughout an IS.

Novel naphthylstyryl-pyridium potentiometric dyes offer advantages for neural network analysis

The submucous plexus of the guinea pig intestine is a quasi-two-dimensional mammalian neural network that is particularly amenable to study using multiple site optical recording of transmembrane voltage (MSORTV) [Biol. Bull. 183 (1992) 344; J. Neurosci. 19 (1999) 3073]. For several years the potentiometric dye of choice for monitoring the electrical activity of its individual neurons has been di-8-ANEPPS [Neuron 9 (1992) 393], a naphthylstyryl-pyridinium dye with a propylsulfonate headgroup that provides relatively large fluorescence changes during action potentials and synaptic potentials. Limitations to the use of this dye, however, have been its phototoxicity and its low water solubility which requires the presence of DMSO and Pluronic F-127 in the staining solution. In searching for less toxic and more soluble dyes exhibiting larger fluorescence signals, we first tried the dienylstyryl-pyridinium dye RH795 [J. Neurosci. 14 (1994) 2545] which is highly soluble in water. This dye yielded relatively large signals, but it was internalized quickly by the submucosal neurons resulting in rapid degradation of the signal-to-noise ratio. We decided to synthesize a series of naphthylstyryl-pyridinium dyes (di-n-ANEPPDHQ) having the same chromophore as di-8-ANEPPS and the quaternary ammonium headgroup (DHQ) of RH795 (resulting in two positive charges versus the neutral propylsulfonate-ring nitrogen combination), and we tested the di-methyl (JPW3039), di-ethyl (JPW2081), di-propyl (JPW3031), di-butyl (JPW5029), and di-octyl (JPW5037) analogues, all of them soluble in ethanol. We found that the di-propyl (di-3-ANEPPDHQ) and the di-butyl (di-4-ANEPPDHQ) forms yielded the best combination of signal-to-noise ratio, moderate phototoxicity and absence of dye internalization.

Isolation of neural activities from respiratory and heartbeat noises for in vivo optical recording in guinea pigs using independent component analysis

Optical recording in vivo is severely interfered by heartbeat and respiratory noises. Here we tested if these noises can be removed from in vivo optical recordings from the primary auditory cortex of the guinea pig, using independent component analysis (ICA). We applied a fast ICA algorithm combined with principal component analysis to optical recordings of long durations (9-40 s). Our results show that ICA can be successfully used to separate sound-evoked neural activities from heartbeat and respiratory noises.

Independent Component Analysis as a Tool to Eliminate Artifacts in EEG: A Quantitative Study

Independent component analysis (ICA) is a novel technique that calculates independent components from mixed signals. A hypothetical clinical application is to remove artifacts in EEG. The goal of this study was to apply ICA to standard EEG recordings to eliminate well-known artifacts, thus quantifying its efficacy in an objective way. Eighty samples of recordings with spikes and evident artifacts of electrocardiogram (EKG), eye movements, 50-Hz interference, muscle, or electrode artifact were studied. ICA components were calculated using the Joint Approximate Diagonalization of Eigen-matrices (JADE) algorithm. The signal was reconstructed excluding those components related to the artifacts. A normalized correlation coefficient was used as a measure of the changes caused by the suppression of these components. ICA produced an evident clearing-up of signals in all the samples. The morphology and the topography of the spike were very similar before and after the removal of the artifacts. The correlation coefficient showed that the rest of the signal did not change significantly. Two examiners independently looked at the samples to identify the changes in the morphology and location of the discharge and the artifacts. In conclusion, ICA proved to be a useful tool to clean artifacts in short EEG samples, without having the disadvantages associated with the digital filters. The distortion of the interictal activity measured by correlation analysis was minimal.