Publication

Abstract

Established methods for imaging the living mammalian brain have, to date, taken the brain’s optical properties as fixed; we here demonstrate that it is possible to modify the optical properties of the brain itself to significantly enhance at-depth imaging while preserving native physiology. Using a small amount of any of several biocompatible materials to raise the refractive index of solutions superfusing the brain prior to imaging, we could increase several-fold the signals from the deepest cells normally visible and, under both one-photon and two-photon imaging, visualize cells previously too dim to see. The enhancement was observed for both anatomical and functional fluorescent reporters across a broad range of emission wavelengths. Importantly, visual tuning properties of cortical neurons in awake mice, and electrophysiological properties of neurons assessed ex vivo, were not altered by this procedure.

Abstract

Spatially resolved proteomics requires a highly multiplexed imaging modality. Cyclic imaging techniques, which repeat staining, imaging, and signal erasure, have been adopted for this purpose. However, due to tissue distortion, it is challenging to obtain high fluorescent signal intensities and complete signal erasure in thick tissue with cyclic imaging techniques. Here, we propose an erasureless cyclic imaging method named IMPASTO. In IMPASTO, specimens are iteratively stained and imaged without signal erasure. Then, images from two consecutive rounds are unmixed to retrieve the images of single proteins through self-supervised machine learning without any prior training. Using IMPASTO, we demonstrate 30-plex imaging from brain slices in 10 rounds, and when used in combination with spectral unmixing, in five rounds. We show that IMPASTO causes negligible tissue distortion and demonstrate 3D multiplexed imaging of brain slices. Further, we show that IMPASTO can shorten the signal removal processes of existing cyclic imaging techniques.

Abstract

Nanoscale resolution imaging of whole vertebrates is required for a systematic understanding of human diseases, but this has yet to be realized. Expansion microscopy (ExM) is an attractive option for achieving this goal, but the expansion of whole vertebrates has not been demonstrated due to the difficulty of expanding hard body components. Here, we demonstrate whole-body ExM, which enables nanoscale resolution imaging of anatomical structures, proteins, and endogenous fluorescent proteins (FPs) of whole zebrafish larvae and mouse embryos by expanding them fourfold. We first show that post-digestion decalcification and digestion kinetics matching are critical steps in the expansion of whole vertebrates. Then, whole-body ExM is combined with the improved pan-protein labeling approach to demonstrate the three-dimensional super-resolution imaging of antibody- or FP-labeled structures and all major anatomical structures surrounding them. We also show that whole-body ExM enables visualization of the nanoscale details of neuronal structures across the entire body.

Abstract

Recent advancements in image denoising have leveraged neural networks to enhance performance, particularly in scenarios where clean-noisy image pairs are unavailable. In this context, self-supervised image denoising methods have gained prominence, centered around the principle of J-invariance — ensuring that the output pixel is not influenced by its corresponding input pixel. Traditionally, enforcing J-invariance has constrained blind spot network (BSN) designs, requiring even core operations such as upsampling or downsampling to follow complex rules. This limitation has led to the exclusion of efficient multi-resolution architectures such as U-net, increasing computational complexity. To address these constraints, we introduce generalized design principles for multi-scale J-invariant networks that allow for the flexible incorporation of nearly any architectural elements. This approach challenges the prevailing notion that J-invariance must be maintained throughout the entire process. Based on our design principles, we present U-BSN, a novel J-invariant network design that utilizes the versatile U-Net architecture, adapting it to accommodate self-supervised learning effectively. We also propose randomized PD, an advanced technique that enhances denoising of real-world images with structured noise. Experimental results validate that U-BSN surpasses existing BSNs in handling real-world noise scenarios and achieves the lowest computational complexity among comparable networks, thus confirming the effectiveness of our design principles and proposed methodologies. 

Abstract

A temporal sequential pattern (TSP) is defined as an ordered collection of events and the corresponding time gaps between consecutive event pairs. When analyzing a sequence of temporal events, identifying frequent TSPs is essential with applications across diverse domains, including neuronal activity analysis, stock trading, and transportation management systems. However, existing mining techniques are often sensitive to hyperparameter settings and may not be scalable for large datasets. Moreover, in practical scenarios, the time gaps in different occurrences of the same TSP may vary to some extent, posing a challenge to the accurate detection of TSPs. In this work, we propose RASP, a robust and resource-adaptive method for mining frequent TSPs. RASP incorporates (a) duplicated matching between TSPs and instances, based on a novel concept of relaxed TSP, for robustness against variations in time gaps, (b) resource-adaptive automatic hyperparameter tuning for enhancing usability, and (c) a tree-based concise data structure for achieving space efficiency. In our experiments, RASP outperforms four state-of-the-art competitors, offering up to 854× faster speed with similar accuracy and up to 342% greater accuracy at similar speeds.

Abstract

This review explores how artificial intelligence (AI) is transforming fluorescence microscopy, providing an overview of its fundamental principles and recent advancements. The roles of AI in improving image quality and introducing new imaging modalities are discussed, offering a comprehensive perspective on these changes. Additionally, a unified framework is introduced for comprehending AI-driven microscopy methodologies and categorizing them into linear inverse problem-solving, denoising, and nonlinear prediction. Furthermore, the potential of self-supervised learning techniques that address the challenges associated with training the networks are explored, utilizing unlabeled microscopy data to enhance data quality and expand imaging capabilities. It is worth noting that while the specific examples and advancements discussed in this review focus on fluorescence microscopy, the general approaches and theories are directly applicable to other optical microscopy methods.

Abstract

As a vertebrate model animal, larval zebrafish are widely used in neuroscience and provide a unique opportunity to monitor whole-brain activity at the cellular resolution. Here, we provide an optimized protocol for performing whole-brain imaging of larval zebrafish using three-dimensional fluorescence microscopy, including sample preparation and immobilization, sample embedding, image acquisition, and visualization after imaging. The current protocol enables in vivo imaging of the structure and neuronal activity of a larval zebrafish brain at a cellular resolution for over 1 h using confocal microscopy and custom-designed fluorescence microscopy. The critical steps in the protocol are also discussed, including sample mounting and positioning, preventing bubble formation and dust in the agarose gel, and avoiding motion in images caused by incomplete solidification of the agarose gel and paralyzation of the fish. The protocol has been validated and confirmed in multiple settings. This protocol can be easily adapted for imaging other organs of a larval zebrafish.

Abstract

Accurate alignment of calcium imaging data, which is critical for the extraction of neuronal activity signals, is often hindered by the image noise and the neuronal activity itself. To address the problem, we propose an algorithm named REALS for robust and efficient batch image alignment through simultaneous transformation and low rank and sparse decomposition. REALS is constructed upon our finding that the low rank subspace can be recovered via linear projection, which allows us to perform simultaneous image alignment and decomposition with gradient-based updates. REALS achieves orders-of-magnitude improvement in terms of accuracy and speed compared to the state-of-the-art robust image alignment algorithms.

Abstract

We report a bio-inspired approach for training a neural network through reinforcement learning to induce high level functions within the network. Based on the interpretation that animals have gained their cognitive functions such as object recognition — without ever being specifically trained for — as a result of maximizing their fitness to the environment, we place our agent in a custom environment where developing certain functions may facilitate decision making; the custom environment is designed as a partially observable Markov decision process in which an input image and the initial value of hidden variables are given to the agent at each time step. We show that our agent, which consists of a convolutional neural network, a recurrent neural network, and a multilayer perceptron, learns to classify the input image and to predict the hidden variables. The experimental results show that high level functions, such as image classification and hidden variable estimation, can be naturally and simultaneously induced without any pre-training or specifying them.

Abstract

Three-dimensional fluorescence microscopy has an intrinsic performance limit set by the number of photons that can be collected from the sample in a given time interval. Here, we extend our earlier work – a recursive light propagation network (RLP-Net) – which is a computational microscopy technique that overcomes such limitations through virtual refocusing that enables volume reconstruction from two adjacent 2-D wide-field fluorescence images. RLP-Net employs a recursive inference scheme in which the network progressively predicts the subsequent planes along the axial direction. This recursive inference scheme reflects that the law of physics for the light propagation remains spatially invariant and therefore a fixed function (i.e., a neural network) for a short distance light propagation can be recursively applied for a longer distance light propagation. In addition, we employ a self-supervised denoising method to enable accurate virtual light propagation over a long distance. We demonstrate the capability of our method through high-speed volumetric imaging of neuronal activity of a live zebrafish brain. The source code used in the paper is available at https://github.com/NICALab/rlpnet.

Abstract

We report the development of deep decomposition and deconvolution microscopy (3DM), a computational microscopy method for the volumetric imaging of neural activity. 3DM overcomes the major challenge of deconvolution microscopy, the ill-posed inverse problem. We take advantage of the temporal sparsity of neural activity to reformulate and solve the inverse problem using two neural networks which perform sparse decomposition and deconvolution. We demonstrate the capability of 3DM via in vivo imaging of the neural activity of a whole larval zebrafish brain with a field of view of 1040 µm × 400 µm × 235 µm and with estimated lateral and axial resolutions of 1.7 µm and 5.4 µm, respectively, at imaging rates of up to 4.2 volumes per second.

Abstract

Calcium imaging is an essential tool to study the activity of neuronal populations. However, the high level of background fluorescence in images hinders the accurate identification of neurons and the extraction of neuronal activities. While robust principal component analysis (RPCA) is a promising method that can decompose the foreground and background in such images, its computational complexity and memory requirement are prohibitively high to process large-scale calcium imaging data. Here, we propose BEAR, a simple bilinear neural network for the efficient approximation of RPCA which achieves an order of magnitude speed improvement with GPU acceleration compared to the conventional RPCA algorithms. In addition, we show that BEAR can perform foreground-background separation of calcium imaging data as large as tens of gigabytes. We also demonstrate that two BEARs can be cascaded to perform simultaneous RPCA and non-negative matrix factorization for the automated extraction of spatial and temporal footprints from calcium imaging data. The source code used in the paper is available at https://github.com/NICALab/BEAR.

Abstract

High-speed optical 3-D fluorescence microscopy is an essential tool for capturing the rapid dynamics of biological systems such as cellular signaling and complex movements. Designing such an optical system is constrained by the inherent trade-off among resolution, speed, and noise which comes from the limited number of photons that can be collected. In this paper, we propose a recursive light propagation network (RLP-Net) that infers the 3-D volume from two adjacent 2-D wide-field fluorescence images via virtual refocusing. Specifically, we propose a recursive inference scheme in which the network progressively predicts the subsequent planes along the axial direction. This recursive inference scheme reflects that the law of physics for the light propagation remains spatially invariant and therefore a fixed function (i.e., a neural network) for a short distance light propagation can be recursively applied for a longer distance light propagation. Experimental results show that the proposed method can faithfully reconstruct the 3-D volume from two planes in terms of both quantitative measures and visual quality. The source code used in the paper is available at https://github.com/NICALab/rlpnet.

Abstract

Methods for one-photon fluorescent imaging of calcium dynamics can capture the activity of hundreds of neurons across large fields of view at a low equipment complexity and cost. In contrast to two-photon methods, however, one-photon methods suffer from higher levels of crosstalk from neuropil, resulting in a decreased signal-to-noise ratio and artifactual correlations of neural activity. We address this problem by engineering cell-body-targeted variants of the fluorescent calcium indicators GCaMP6f and GCaMP7f. We screened fusions of GCaMP to natural, as well as artificial, peptides and identified fusions that localized GCaMP to within 50 μm of the cell body of neurons in mice and larval zebrafish. One-photon imaging of soma-targeted GCaMP in dense neural circuits reported fewer artifactual spikes from neuropil, an increased signal-to-noise ratio, and decreased artifactual correlation across neurons. Thus, soma-targeting of fluorescent calcium indicators facilitates usage of simple, powerful, one-photon methods for imaging neural calcium dynamics.

Abstract

We developed a new way to engineer complex proteins toward multidimensional specifications using a simple, yet scalable, directed evolution strategy. By robotically picking mammalian cells that were identified, under a microscope, as expressing proteins that simultaneously exhibit several specific properties, we can screen hundreds of thousands of proteins in a library in just a few hours, evaluating each along multiple performance axes. To demonstrate the power of this approach, we created a genetically encoded fluorescent voltage indicator, simultaneously optimizing its brightness and membrane localization using our microscopy-guided cell-picking strategy. We produced the high-performance opsin-based fluorescent voltage reporter Archon1 and demonstrated its utility by imaging spiking and millivolt-scale subthreshold and synaptic activity in acute mouse brain slices and in larval zebrafish in vivo. We also measured postsynaptic responses downstream of optogenetically controlled neurons in C. elegans.

Abstract

We here introduce and study the properties, via computer simulation, of a candidate automated approach to algorithmic reconstruction of dense neural morphology, based on simulated data of the kind that would be obtained via two emerging molecular technologies—expansion microscopy (ExM) and in-situmolecular barcoding. We utilize a convolutional neural network to detect neuronal boundaries from protein-tagged plasma membrane images obtained via ExM, as well as a subsequent supervoxel-merging pipeline guided by optical readout of information-rich, cell-specific nucleic acid barcodes. We attempt to use conservative imaging and labeling parameters, with the goal of establishing a baseline case that points to the potential feasibility of optical circuit reconstruction, leaving open the possibility of higher-performance labeling technologies and algorithms. We find that, even with these conservative assumptions, an all-optical approach to dense neural morphology reconstruction may be possible via the proposed algorithmic framework. Future work should explore both the design-space of chemical labels and barcodes, as well as algorithms, to ultimately enable routine, high-performance optical circuit reconstruction.

Abstract

We recently developed a method called expansion microscopy, in which preserved biological specimens are physically magnified by embedding them in a densely crosslinked polyelectrolyte gel, anchoring key labels or biomolecules to the gel, mechanically homogenizing the specimen, and then swelling the gel–specimen composite by ∼4.5× in linear dimension. Here we describe iterative expansion microscopy (iExM), in which a sample is expanded ∼20×. After preliminary expansion a second swellable polymer mesh is formed in the space newly opened up by the first expansion, and the sample is expanded again. iExM expands biological specimens ∼4.5 × 4.5, or ∼20×, and enables ∼25-nm-resolution imaging of cells and tissues on conventional microscopes. We used iExM to visualize synaptic proteins, as well as the detailed architecture of dendritic spines, in mouse brain circuitry.

Abstract

A first-order time-based ΔΣ modulator using voltage-controlled oscillator (VCO) is presented. The proposed modulator employs phase interpolation technique to enhance the time resolution of the VCO and digital calibration to improve the linearity of the VCO tuning curve. The proposed modulator, implemented in 0.13 μm CMOS process, achieves 55 dB peak signal-to-noise ratio and 52.5 dB peak signal-to-noise-and-distortion ratio at 600 MHz sampling frequency for 20 MHz input bandwidth and consumes 14.3 mW.

Abstract

A voltage-controlled oscillator (VCO) based analog-to-digital converter (ADC) is a time-based architecture with a first-order noise-shaping property, which can be implemented using a VCO and digital circuits. This paper analyzes the performance of VCO-based ADCs in the presence of nonidealities such as jitter, nonlinearity, mismatch, and the metastability of D flip-flops. Based on this analysis, design criteria for determining parameters for VCO-based ADCs are described. In addition, a digital calibration technique to enhance the spurious-free dynamic range degraded by the nonlinearity is also introduced. To verify the theoretical analysis, a prototype chip is implemented in a 0.13-μm CMOS process. With a 500-MHz sampling frequency, the prototype achieves a signal-to-noise ratio ranging from 71.8 to 21.3 dB for an input bandwidth of 100 kHz-247 MHz, while dissipating 12.6 mW and occupying an area of 0.078 mm2 .

Abstract

In this paper, a bandpass analog-to-digital converter (ADC) based on time-interleaved oversampled ADC is introduced. Unlike previous delta-sigma bandpass ADCs that require accurate digital-to-analog converters and high-speed analog circuits, the proposed architecture provides bandpass function by time-interleaving first-order voltage-controlled-oscillator (VCO)-based ADCs. The use of VCO-based ADC has the advantage that its resolution is determined by the time resolution rather than the voltage resolution, thus making it attractive for future low-voltage CMOS processes. The performance of the proposed ADC is theoretically analyzed and simulated in ideal condition, as well as in nonideal condition, in the presence of nonlinearity, sampling clock jitter, and mismatch.

Abstract

Light-field microscopy (LFM) can support imaging of the entire C. elegans nervous system and the entire larval zebrafish (Danio rerio) brain at high speeds (Nature Methods (2014) 11:727-730), but the spatial resolution obtained was insufficient to yield single-cell resolution imaging in many contexts, such as in behaving zebrafish larvae. More recently, eXtended field of view LFM (XLFM) (eLife (2017) 6:e28158), which simultaneously optimizes imaging volume and spatial resolution, and avoids square-shaped artifacts near the focal plane by placing the micro-lens array on the pupil plane of the system, was developed to offer better spatial resolution, but separating signals from nearby neurons is still a challenge. Here we introduce sparse decomposition light-field microscopy (SDLFM), a computational imaging technique that further improves the ability of XLFM, and that allows for the imaging of neuronal activity with very high spatiotemporal resolution. With this technique, the spatial resolution can be improved and hence neuronal activity can be accurately recovered, even from nearby neurons. We show the power of SDLFM by demonstrating in vivo imaging of neuronal activity of whole brains of larval zebrafish and adult fruit flies (Drosophila melanogaster) with high volume rates up to 50 Hz. 

Abstract

We developed a new way to engineer complex proteins toward multidimensional specifications using a simple, yet scalable, directed evolution strategy. By robotically picking mammalian cells that were identified, under a microscope, as expressing proteins that simultaneously exhibit several specific properties, we can screen hundreds of thousands of proteins in a library in just a few hours, evaluating each along multiple performance axes. To demonstrate the power of this approach, we created a genetically encoded fluorescent voltage indicator, simultaneously optimizing its brightness and membrane localization using our microscopy-guided cell-picking strategy. We produced the high-performance opsin-based fluorescent voltage reporter Archon1 and demonstrated its utility by imaging spiking and millivolt-scale subthreshold and synaptic activity in acute mouse brain slices and in larval zebrafish in vivo. We also measured postsynaptic responses downstream of optogenetically controlled neurons in C. elegans.

Abstract

Methods for one-photon fluorescent imaging of calcium dynamics in vivo are popular due to their ability to simultaneously capture the dynamics of hundreds of neurons across large fields of view, at a low equipment complexity and cost. In contrast to two-photon methods, however, one-photon methods suffer from higher levels of crosstalk between cell bodies and the surrounding neuropil, resulting in decreased signal-to-noise and artifactual correlations of neural activity. Here we address this problem by engineering cell body-targeted variants of the fluorescent calcium indicator GCaMP6f. We screened fusions of GCaMP6f to both natural as well as engineered peptides, and identified fusions that localized GCaMP6f to within approximately 50 microns of the cell body of neurons in live mice and larval zebrafish. One-photon imaging of soma-targeted GCaMP6f in dense neural circuits in larval zebrafish and in mice reported a decrease in artifactual spikes from neuropil, an increased signal-to-noise ratio, and decreased artifactual correlation across neurons. Thus, soma-targeting of GCaMP may facilitate even greater usage of simple, powerful, one-photon methods of population imaging of neural calcium dynamics.

Abstract

Understanding how biomolecules such as proteins are architected in 3-D throughout synapses, neurons, and brain circuits is essential to understanding how such molecular and cellular machines work together to support the operation of normal and diseased neural networks. Earlier, we discovered that biological specimens could be physically magnified by ~4.5-fold by embedding them in a dense swellable polyelectrolyte gel, anchoring key biomolecules to the polymer network, and then adding water to osmotically swell the gel - a process we call ‘expansion microscopy’ (ExM; Science 347(6221):534-548). We are now creating an improved, next-generation form of expansion microscopy, which can expand specimens up to 20-fold, enabling ~20-nm spatial resolution imaging, using conventional optics. Here, we report the optimization and application of this next-generation expansion microscopy to reveal extremely fine structures that make up brain circuits, including components of the synaptic cleft in intact brain circuits, and Brainbow-labeled neural circuits. Given that conventional, diffraction-limited optics can proceed with very high throughput (e.g., in lightsheet microscopy), the organization of brain circuitry can be revealed with nanoscopic precision over large volumes relevant to behavior and disease, using next-generation ExM. We anticipate that large-volume nanoscopic imaging via next-generation expansion microscopy, because it does not require any special hardware, may support widescale and democratized investigation of the molecular configurations of brain circuits.

Abstract

We recently developed an approach using light-field microscopy (LFM) to image the entire C. elegans nervous system and the entire zebrafish brain in 3D, at high speeds (e.g., ~20 Hz), possible since no moving parts are required for light-field microscopes to acquire 3-D images (Prevedel et al., 2014). However, the spatial resolution was limited, and the activity of only a subset of the neurons could be extracted as it relied on a computational method based on independent component analysis (ICA) to extract the activity of neurons during post-processing. For ground-truth studies of the nervous system, ideally it would be possible to pick up the activity of all neurons, even those that are quiet during the recording (whereas ICA tends to select for highly active neurons). And, ideally, it would be possible to assign neural activity to defined sites in the circuit, in order to link anatomical and dynamical descriptions of neural circuits. We have developed a novel sparse reconstruction lightfield microscopy (SRLFM) strategy that can, in simulation, accurately reconstruct neural activity throughout the entire brain of the larval zebrafish, expressing a GCaMP variant, with single cell resolution, at up to 100 Hz. No post-processing (e.g., via ICA) is needed with SRLFM, meaning that even quiet neurons can be detected, and also neural activity can be accurately assigned to specific neurons in the neural network, important for linking neural circuit dynamics to the underlying network architecture. The effective resolution of light-field microscopy is improved by >2-fold with our new algorithm. We are currently implementing this algorithm on an optimized light-field microscope, aiming to assess the technology in the context of living brain dynamics. 

Abstract

We present the design of a system to image the neural activity of many individual neurons simultaneously, at arbitrary locations in the brain. We use numerical simulations to estimate the proposed system’s performance, and predict how future advances would improve performance. Simulating widely available fluorescent reporters and current CMOS pixel designs, our results indicate that a microprobe shank (e.g., 10 microns thin, 60 microns wide) placed in the cortex could image the activity of many hundreds of neurons. While fluorescent indicators of physiological activity are powerful tools for monitoring brain activity, it remains difficult to use them to image the activity of individual cells at high speeds, especially in deep tissue. We propose a lensless microimager that is thin enough to implant with minimal tissue displacement. An array of such probes can record from many arbitrary sites in the mammalian brain. The system consists of a silicon probe that is densely arrayed with CMOS imaging pixels. On top of the CMOS pixels are a standoff layer, a light-modulating mask layer, and fluorescence emission filters. This probe is implanted in conjunction with an excitation light delivery mechanism--such as a waveguide or optical fiber to excite the fluorescent signal to be recorded. Whereas the lenses in a traditional imager alter the angles of incoming light such that the recorded image contains purely spatial information, our mask causes the imager to collect spatial and angular information, sampling from the 4D light field. The recorded angular information allows us to determine depth information; the imager records a full 3D image every frame. This reduces problems related to out of focal plane light without requiring techniques like confocal or two photon microscopy. While our technique does not allow diffraction-limited resolution, it does allow resolving to within a few microns, which is sufficient to correctly assign fluorescence to its cell of origin in most brain structures. Reconstructing the 3D volume from the recorded image can be cast as a straightforward linear algebra problem. While this problem appears to be underconstrained, there are several techniques for introducing additional information—such as sparsity, knowledge of cell locations, or positive-definiteness—that allow us to constrain the problem to a single valid solution. The end result is an activity trace for each neuron in the probe’s field of view. We characterize the correlation between the fluorescent activity and the probe recording, and find that with full fluorescent labeling in the cortex, a single probe would record hundreds of neurons with over 90% fidelity to the underlying signal.

Abstract

3D functional imaging of neuronal activity of entire organisms at the single cell level and on physiologically relevant time scales poses a major challenge in neuroscience. Here, using light-field microscopy in combination with 3D deconvolution, we demonstrate intrinsically simultaneous volumetric functional imaging of neuronal population activity at single neuron resolution for an entire organism, the nematode Caenorhabditis elegans, at up to 50Hz. This owes to a system-level optimized light field microscope, a fast deconvolution algorithm, and an appropriate genetically-encoded calcium indicator (GECI). This data combined with the full wiring diagram, or connectome, of C. elegans may allow the study of the nervous system at the circuit level in a systematic way. By performing whole-brain imaging of neural activity in larval zebrafish, we demonstrate the ability of our technique to capture dynamics of spiking neurons in volumes of ~700x700x200μm at 20Hz. The optical resolution for this volume is below single neuron resolution, but computational techniques based on independent component analysis enabled us to extract the activity of large numbers of neurons. The simplicity of our technique makes it an attractive tool for high-speed volumetric calcium imaging, and potentially voltage imaging, of neural activities in intact neural networks. 

Abstract

In this paper, a cepstral analysis based approach to measuring the depth of anesthesia (DoA) is presented. Cepstral analysis is a signal processing technique widely used especially for speech recognition in order to extract speech information regardless of vocal cord characteristics. The resulting index for the DoA is called index based on cepstral analysis (ICep). The Fisher criterion is engaged to evaluate the performance of indices. All analyses are based on a single-channel electroencephalogram (EEG) of 10 human subjects. To validate the proposed technique, ICep is compared with bispectral index (BIS), which is the most commonly used method to estimate the level of consciousness via EEG during general anesthesia. The results show that ICep has high correlation with BIS, and is outstanding in terms of the Fisher criterion and offers faster tracking than BIS in the transition from consciousness to unconsciousness.

Abstract

In this paper, an entropy based method for quantifying the depth of anesthesia from rat EEG is presented. The proposed index for the depth of anesthesia called modified Shannon entropy (MShEn) is based on Shannon entropy (ShEn) and spectral entropy (SpEn) which are widely used for analyzing non-stationary signals. Discrimination power (DP), as a performance indicator for indexes, is defined and used to derive the final index for the depth of anesthesia. For experiment, EEG from anesthetized rats are measured and analyzed by using MShEn. MShEn shows both high stability and high correlation with other indexes for depth of anesthesia.

Abstract

In this paper, a time-based analog-to-digital converter (ADC) with first order noise shaping property is presented. The ADC consists of a pulse-width modulator (PWM) and a time-to-digital converter (TDC) based on a gated-ring oscillator. A prototype is designed in 0.18μm CMOS and simulated to verify the concept. The ADC shows high linearity despite the open loop architecture since the linearity does not depend on the frequency tuning characteristic of oscillators unlike conventional noise shaping time-based ADCs. With the sampling frequency of 10MHz, the prototype achieves spurious-free dynamic range (SFDR) and signal-to-noise-and-distortion ratio (SNDR) of 66dB and 56dB, respectively, when the signal bandwidth is 500kHz.

Abstract

In this paper, a linearization technique for voltage-controlled oscillator (VCO)-based analog-to-digital converter (ADC) is presented. Even order harmonics are canceled by using pseudo-differential architecture with two identical VCO-based ADCs. The effect of cancelation technique is verified through simulation with a prototype designed in 0.18μm CMOS technology and verilog. The CMOS ring VCO consumes about 280μW of power in average and the digital counter is implemented in verilog. With the sampling frequency of 100MHz, the prototype achieves signal-to-noise ratio (SNR), spurious-free dynamic range (SFDR) and signal-to-noise-and-distortion ratio (SNDR) of 58.2dB, 60dB and 56.5dB, respectively, when the signal bandwidth is 5MHz. Although the power consumption of the overall ADC is doubled compared to a single channel ADC, the figure-of-merit (FOM) of the ADC is improved by 2.95 times due to the SNDR improvement by linearization technique.

Abstract

In this paper, a highly programmable bio-impedance measurement system for multi-purpose health monitoring application is presented. Complex and area consuming analog components are replaced by a high resolution ADC and digital processing providing lots of flexibility to both designers and users. The prototype of the proposed system architecture is designed in 0.18μm CMOS technology. The tuning range of signal generator is from 10kHz∼10MHz and current variation of voltage-controlled current source (VCCS) is below 1%. The gain, bandwidth, CMRR and input referred noise of instrumentation amplifier (IA) are 18dB, 1.5Hz∼9.5MHz, 96.7dB at 100kHz and 10.3nV/√Hz, respectively. ADC is implemented using pseudo-differential VCO-based architecture. Signal-to-noise ratio (SNR), spurious-free dynamic range (SFDR) and signal-to-noise-and-distortion ratio (SNDR) of ADC are 87.5dB, 60dB and 60dB, respectively. The total power consumption of prototype is 6.7mW.

Abstract

In this paper, a novel pulse transit time (PTT) measurement method which employs electrocardiography and bioimpedance measurement (ECG-BIM) is proposed. Unlike conventional method based on ECG and photoplethysmography (ECG-PPG), the proposed method offers a promising technique for continuous and portable monitoring of cardiovascular diseases, as it requires only electrical components that can be fully integrated in a low-power and low-cost silicon circuit. To verify the validity of the proposed ECG-BIM method, experiments have been conducted on a human body in various conditions, using both the ECG-PPG and the ECG-BIM methods which is implemented using discrete off-the-shelf components. Experimental results show that the PTT data achieved from ECG-PPG and ECG-BIM methods are highly correlated, with correlation coefficient of 0.93, hence indicating the validity of the proposed method which can offer high level of integration and low power consumption for portable continuous cardiovascular signal monitoring. 

Abstract

This paper presents a bandpass ADC which exploits enhanced time-resolution of a deep submicron CMOS process. Unlike conventional bandpass ADCs that rely on voltage resolution and Gm-LC filters, the proposed ADC employs time-interleaved voltage-controlled oscillators that enable frequency tunable bandstop noise shaping property without a feedback loop. The ADC implemented in 65nm CMOS achieves SNR of 63.3dB for 1MHz signal located at 1.5GHz, while consuming 19.6mW from 1.2V supply.