Kavli Blog

Method to overcome light scatter in optical imaging distinguishes moving objects with high fidelity… Novel analysis method improves accuracy of neural network models… Improved imaging technique reveals synaptic transmission at quantal resolution in fruit fly larvae …

Phase retrieval methods in optical imaging allow for successful imaging of moving targets through scattered media 

A major challenge of optical imaging is the scattering of light, which leads to lower resolution, poorer image quality, and shallower depths of images, especially those involving biological tissue. Researchers have tried to overcome these effects by filtering out repeated scattering of light to measure only the un-scattered or minimally scattered photons. However, such methods fail to address limitations in depth of imaging because they detect only the minimally scattered photons, which are difficult to observe at greater distances. Optical imaging approaches that incorporate information from the scattered photons often require long acquisition times, a reference source (also known as a “guide star”), and dark-field conditions. Recently, Dr. Changhuei Yang and his group at California Institute of Technology have developed a successful method to image moving objects through scattering media. Using temporal and angular correlational analyses from the scattering process, they distinguished objects from their backgrounds, permitting imaging of moving targets that would otherwise have been hidden due to the scattering media. This technique can be used in both dark-field and light-field scenarios, allowing the imaging of non-emitting, transmissive, and reflective non-static samples. This exciting new method could be extended to biological tissue, such as imaging deep brain structures in freely moving animals, and possibly other applications, such as moving objects in dense environments like fog or underwater.

A) Objects were hidden behind a scattering medium and moved 1.5 mm between each acquisition (4 acquisitions per image). B) The raw camera images were blurry and indistinguishable due to the scattering of light from the background. Images remain indistinct based on C) object autocorrelations and D) speckle autocorrelation analyses. E) Using a phase retrieval algorithm, each object was successfully reconstructed.

Functional imaging analysis incorporates connectivity and structural relatedness to improve modeling of brain functional networks

Functional network modeling from resting-state functional magnetic resonance imaging (fMRI) helps clarify how the brain functions normally and could identify biomarkers of neurological and psychiatric disorders. Often, models represent regions of interest (ROIs) in the brain as “nodes” with the functional connectivity between these regions delineated “edges.” At the University of North Carolina, Chapel Hill, Dr. Dinggang Shen’s group developed a new method, connectivity strength-weighted sparse group representation, to model brain functional networks. Like other methods, the team relied on pairwise correlations between ROIs to infer connectivity and made the neurologically reasonable assumption of sparse networks. From BOLD signals, the model assumed that brain regions have ‘first order/direct” interactions with only a few other regions and reduced spurious connections. However, unlike previous methods, the model made use of information on variations in connectivity strength and groupings of similar ROIs. Thus, the team combined pairwise correlations to measure interactions across multiple ROIs, weighted functional connectivity strength to account for network sparsity, and incorporated ROI group structures (a set of regions with similar characteristics) into one model. Using fMRI data from individuals with mild cognitive impairment (MCI), an early indicator of Alzheimer’s disease or another dementia, and from normal controls, the group tested the model on seven performance metrics, including accuracy, sensitivity, and specificity. The team found that their model successfully classified the data as MCI or control better than other brain network connection models with almost 85% accuracy compared to 65%. These results provide validation that this model can identify biomarkers of MCI, which could guide early interventions for Alzheimer’s disease. Shen’s group intends to improve the grouping of related structures in the model further and apply to different brain disorders and diseases.

This figure shows the pattern of nodes and edges that are most likely to indicate MCI classification, an early marker of Alzheimer’s disease. The nodes, represented by green spheres, mark regions of interest and the edges, represented by red and blue lines, identify the connectivity between the regions. These nodes and edges were consistently selected by the new method of connectivity strength-weighted sparse group representation. Note that increased line thickness is represented by increased connectivity as measured by fMRI BOLD response.

Novel quantal resolution imaging technique advances understanding of input-specific plasticity and homeostasis at the Drosophila larval neuromuscular junction

Synapses, including those using glutamate as a neurotransmitter, vary greatly in pre- and post-synaptic transmission properties. How differences in pre-synaptic release characteristics, short term plasticity, and homeostatic stability-promoting regulation relate to one another and to this diversity is poorly understood. At the University of California, Berkeley, Dr. Ehud Isacoff and his group used the Drosophila larval neuromuscular junction (NMJ), a model system for studying glutamatergic transmission, to better understand these mechanisms. They used novel quantal resolution imaging to study the role of input and synapse specificity in the regulation of basal synaptic strength, plasticity, and homeostasis. This technique relies on transgenic Drosophila that express a genetically-encoded calcium indicator, SynapGCaMP6f, allowing for the imaging and quantification of post-synaptic transmission without voltage clamping. Glutamate is released during basal and evoked synaptic events following increases in calcium, revealed by SynapGCaMP6f as changes in fluorescence. Thus, the researchers used fluorescent imaging to examine the probability, frequency, and amplitude of quantal spontaneous, and activity-dependent synaptic transmission of two different synapes onto muscle, those from Ib and Is neurons. By applying this technique with and without motor input, the researchers discovered that Ib and Is synapses have different basal synaptic release characteristics and activity-dependent synaptic modulation at the NMJ. Strikingly, homeostatic compensation in synaptic strength occurred only in the 1b synapses. BRAIN-supported advances in imaging techniques such as this continue to give unprecedented insight into how the nervous system carries out computations and can be expanded across model systems.

Panel A. Baseline fluorescence of the calcium indicator in Ib (blue) relative to Is (orange) at the NMJ. Panel B. Evoked synaptic activity at Ib (blue) relative to Is (orange) at the NMJ. Warmer colors indicate an increased probability of glutamate release. Panels E and F. High-magnification comparison of quantal evoked activity for each neuron.

BRAIN Initiative team pushes the limits of functional magnetic resonance imaging (fMRI) for human brain… A novel tool to manipulate gene function in specific cell types… Understanding the functional diversity of retinal bipolar cells…

Advancements to functional imaging technique result in ultra-high resolution capture of human cortical columns

Despite numerous advances in fMRI technology, most components are optimized for the entire body. This makes safe, ultra-high resolution (UHR) imaging of columnar organization throughout the cortex of the human brain nearly impossible. At the University of California, Berkeley, Dr. David Feinberg and colleagues applied updates to magnetic gradients, receiver arrays, and pulse sequences of simultaneous multi-slice echo planar imaging fMRI to achieve UHR imaging of human ocular dominance columns. Focusing particularly on a prototype receiver array (8 channels with 4 cm diameter coils), the group systematically describes the changes necessary to achieve the higher signal-to-noise ratio required to attain ~0.5 mm imaging resolution in 3 dimensions. Finally, the researchers display their updated UHR system compared to commercially available technology when mapping ocular dominance columns in three subjects shown visual stimuli in the scanner. The group notes their findings are part of a growing set of 3D imaging studies, moving to leverage fMRI to understand neural circuitry by revealing activity of distinct cell populations in different cortical layers. They postulate that this 3D imaging technique could eventually progress from 0.5 mm to 300-400 µm resolution fMRI of the entire human brain.

Brief fMRI scans from a visual activation paradigm in a human subject reveal enhanced cortical activation at 0.5 mm spatial resolution in the prototype receiver array (8-channel with 4 cm diameter loops) compared to commercially available technology (32-channel). The prototype array accurately measured activation at 0.45 mm resolution as well, further illustrating improved signal-to-noise ratio.

New technique developed for controlling gene function in distinct cell types in the fruit fly

The ability to manipulate genes in specific cell types is critical for understanding circuit function and dysfunction. Unfortunately, there are limitations for the currently available tools, including off-target effects (i.e., modifying genes other than the targeted gene), incomplete inactivation of the targeted gene, requiring cell division (which restricts the time during development when the gene can be targeted), and incompatibility with model systems like the fruit fly Drosophila melanogaster, which is a principle model for studying neural development and function. At Stanford University, Dr. Thomas Clandinin and colleagues have developed a tool called FlpStop, which can completely disrupt targeted genes, as well as rescue gene expression of the disrupted genes, in differentiated and undifferentiated cells in Drosophila. FlpStop uses endogenous mechanisms within cells and a process called insertional mutagenesis to completely inactivate/disrupt the normal function of a gene. The FlpStop insertion is tagged with a fluorescent protein so that mutant cells can be visualized, and it can be inserted in both non-disrupting and disrupting orientations. The team successfully disrupted gene function in six out of eight of the genes that were tested and confirmed that the non-disrupting orientation did not interfere with gene function in any case. They successfully labeled the FlpStop inserted genes in three different cell types in the Drosophila visual system (Mi1, Tm1, and T4). Finally, the group effectively combined FlpStop with in vivo calcium imaging. The findings suggest that FlpStop represents a promising and powerful new tool for investigating gene function in specific cell types.

(a) Schematic of the experimental design for testing the effectiveness of the FlpStop insertion. Drosophila Gal4 driver lines were used, whereby three distinct cell types in the visual system were targeted: (b) Mi1, (c) Tm1, and (d) T4. The full expression of each Gal4 driver line is labeled green while the combination of Gal4 (in green) and the successfully-expressed FlpStop gene (in red) together appear yellow, demonstrating 70%-93% overlap depending on the cell type.

Studying the functional diversity of bipolar cells improves understanding of neuronal processing in the visual system

One core goal of the BRAIN Initiative is to better understand brain circuitry and the role of different cell types in these brain circuits, in both healthy and diseased brains. A relevant area of focus is retinal neurons and their ability to encode visual stimuli for the brain. While the anatomy and genetics of retinal bipolar cells are well characterized, their functional diversity is incompletely understood. In an article in Nature, BRAIN awardee Dr. Thomas Euler and colleagues used two-photon imaging to examine the effects of amacrine cell activity (a type of retinal interneuron) on the output of bipolar cells in the mouse retina. Through exposing different areas of the retina to light, the researchers found that functionally opposite signals, such as those used to describe ON and OFF bipolar cells, exist at the same layer of the retina, suggesting that the retina structure is more complicated than previously thought. Furthermore, inhibition of amacrine cell activity led to increases in the functional diversity of bipolar cells. The team determined that a bipolar cell’s output is determined by a combination of excitatory input to the dendrite and amacrine cell input to the axon, which ultimately allows for temporal encoding in the visual system. These important findings give us a better understanding of the visual system and of the mechanisms through which different neuronal cell types communicate with one another.

Local (gray) and full-field (black) output responses of bipolar cells in both control and drug conditions. TPMPA/Gbz blocks GABAergic amacrine cell activity, while strychnine blocks glycinergic amacrine cell activity. Blocking amacrine cell activity led to opposite responses from bipolar cells compared to control conditions.