Method and device for analyzing activation mechanism of biological brain neuron for visual search

Abstract

The application provides a visual search-oriented biological brain neuron activation mechanism analysis method and device. The application takes a mouse as a model animal, (1) proposes a neuron activation analysis method based on target stimulation correlation, solves the representation problem of the sparse activation mechanism of the visual cortex neuron, (2) proposes a neuron activity pattern analysis method based on the functional hierarchical division of the mouse visual brain area, solves the representation problem of the hierarchical activation mechanism of the visual cortex neuron, and (3) introduces a background noise control experiment in visual stimulation, and solves the representation problem of the visual cortex neuron activation distribution law under the influence of background stimulation noise.

Description

Technical Field

[0001] This invention relates to the field of neuronal activity pattern analysis technology, and in particular to a method and apparatus for analyzing the activation mechanism of biological brain neurons for visual search. Background Technology

[0002] The emergence and development of real-time, high-resolution, wide-field-of-view microscopic mesoscopic imaging technology has made it possible to observe biological brain neurons, accelerated the research on the mechanism of neuronal activity patterns, and further promoted the development of research on the neurocognitive mechanism of visual search.

[0003] The principle behind neuronal fluorescent labeling technology is the specific expression of fluorescent protein indicators by nerve cells, which allows us to observe fluorescence signals in specific wavelengths under a mesoscopic fluorescence microscope. By analyzing the intensity of the fluorescence signals, we can indirectly obtain information about the intensity of individual and collective neuronal activity.

[0004] The study of visual search mechanisms in rodents, especially mice, has always been a hot topic and a challenging area in the field of biological visual processing mechanisms. Research on visual search mechanisms includes two aspects: attention mechanism research and the study of neuronal activity patterns in the visual cortex. This invention primarily focuses on the study of neuronal activity patterns in the visual cortex, specifically the group coding mechanisms of neurons in the primary and higher visual cortex during the visual search process of mice for target objects / images.

[0005] Existing technical solutions:

[0006] Different visual behavioral experiments employ different analytical methods. Some experiments typically use full-screen checkerboard or interference fringe stimuli, focusing on the physiological perception of visual brightness and color changes in normal or diseased mice. Other experiments use visual stimuli involving differences in image shape or changes in position, focusing on memory and decision-making mechanisms. Because visual search is a fundamental visual ability in animals, visual behavioral experiments, except for those involving full-screen stimuli, all involve the animal's visual search process. However, few experiments conduct in-depth, focused analysis of visual search results. The core reason is that their experimental protocols lack specific quantitative evaluation indicators from the perspective of visual search research.

[0007] Brain imaging analysis methods corresponding to these visual behavioral experiments focus primarily on the basic cognitive functions of different brain regions under target visual stimuli, including brain regions or sub-brain regions such as the functional columns of the visual cortex, prefrontal cortex, hippocampus, dorsal striatum, and ventral striatum. Less attention is paid to the process by which mice find target stimuli through visual search, and the analysis of neuronal population activity patterns during the visual search process.

[0008] Existing theoretical research on brain mapping proposes the concept of functional levels in the visual cortex from the perspective of dendrites and axonal connections of visual neurons. This divides the visual cortex into four functional levels: V1, LM, RL, LP, AL, PM, and AM. These four levels are positioned in the feedforward and feedback layers of complex neural networks, precisely coordinating to complete the transmission and processing of visual information. However, current research on the activity patterns of visual cortical neurons in visual search treats different visual cortex areas as isolated information islands, with limited research on the interconnected activation patterns between multiple visual cortex areas. Furthermore, it rarely conducts grouped statistical studies of related brain regions of the visual cortex from a functional perspective, failing to uncover the activation patterns of different functional levels during the visual search process.

[0009] In addition, the real-world environment is filled with visual background noise. Therefore, simulating real-world environments and studying how model animals search for target visual stimuli under background noise interference is valuable. However, existing visual behavioral experiments typically only present target stimuli to model animals without a background (usually a pure black background), thus failing to investigate the neurocognitive mechanisms of the influence of background noise. Therefore, there is an urgent need for a method and system for analyzing the activity patterns of neurons in the mouse visual cortex. By analyzing the functional hierarchy and association patterns of the primary and higher visual cortex during the mouse visual search process, the activation mechanism of neurons in the visual cortex during this visual cognitive processing can be studied. Summary of the Invention

[0010] The present invention aims to at least partially solve one of the technical problems in the related art.

[0011] Therefore, the first objective of this invention is to propose a method for analyzing the activation mechanism of biological brain neurons for visual search.

[0012] The second objective of this invention is to propose a device for analyzing the activation mechanism of biological brain neurons for visual search.

[0013] To achieve the above objectives, a first aspect of the present invention proposes a method for analyzing the activation mechanism of biological brain neurons for visual search, comprising:

[0014] S1, based on the functional hierarchy of brain atlas, divides the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex and higher visual cortex.

[0015] S2, introduce background image noise into the visual stimulus to obtain neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions;

[0016] S3, perform non-rigid motion correction, cell segmentation and trace extraction on the calcium ion fluorescence signal to obtain the activity characteristics of each neuron at different stimulation stages;

[0017] S4. Based on the activity characteristics, cluster analysis and significance testing methods are used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level.

[0018] In one embodiment of the present invention, S1 includes:

[0019] S11 divides the primary visual cortex into the V1 brain region, the secondary visual cortex into the LM and RL brain regions, the intermediate visual cortex into the LP and AL brain regions, and the higher visual cortex into the AM and PM brain regions.

[0020] S12, based on the functional localization of each brain region in visual information processing, the LLA and LI brain regions are assigned to the intermediate visual cortex, while the MMP, MMA, and RS brain regions are assigned to the higher visual cortex.

[0021] In one embodiment of the present invention, S2 includes:

[0022] S21, superimpose the target stimulus image with the background noise image, wherein the background noise image is a random texture or natural scene image, and its brightness and contrast are adjustable;

[0023] S22 quantifies the impact of background noise on visual search efficiency and neuron activation distribution by comparing the activation ratios of neurons at each level in noise-free and noisy experiments.

[0024] In one embodiment of the present invention, S3 includes:

[0025] S31 uses the NoRMCorre algorithm to perform non-rigid motion correction on the imaging data, splits the field of view into overlapping patches and registers them separately, and then merges them through smooth interpolation to eliminate non-rigid artifacts caused by the natural movement of the mouse.

[0026] S32 uses the CNMF algorithm to perform cell segmentation and trace extraction on the corrected image, decomposes the spatiotemporal activity of neurons into spatial and temporal components, and introduces a structured calcium dynamics model in low signal-to-noise ratio regions to improve the accuracy of trace extraction.

[0027] In one embodiment of the present invention, it further includes:

[0028] S5 calculates the z-score normalized standard deviation of neurons at different stimulation stages and combines it with the K-means clustering method to divide neurons into multiple clusters to identify their functional roles in the visual search task.

[0029] To achieve the above objectives, a second aspect of the present invention provides a biological brain neuron activation mechanism analysis device for visual search, comprising:

[0030] The visual cortex hierarchy module is used to divide the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex and higher visual cortex, based on the functional hierarchy of brain atlas.

[0031] The noise stimulation data acquisition module is used to introduce background image noise into visual stimuli to acquire neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions.

[0032] The neuronal activity feature extraction module is used to perform non-rigid motion correction, cell segmentation, and trace extraction on the calcium ion fluorescence signal to obtain the activity features of each neuron at different stimulation stages.

[0033] The activation pattern quantification analysis module is used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level based on the activity characteristics, using cluster analysis and significance testing methods.

[0034] This invention provides a method and apparatus for analyzing the activation mechanism of biological brain neurons in visual search. It can effectively analyze the sparse activation and hierarchical activation mechanism of neurons in the visual search process of mice, quantitatively evaluate the influence of background noise on the activation distribution of neurons in the visual cortex, and thus reveal the neural coding rules of the biological visual cortex in complex environments.

[0035] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0036] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0037] Figure 1 This is a flowchart of a method for analyzing the activation mechanism of biological brain neurons for visual search according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of trace extraction and dimensionality reduction clustering according to an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the sparse activation mechanism of mouse visual cortex neurons according to an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the hierarchical activation mechanism of mouse visual cortex neurons according to an embodiment of the present invention;

[0041] Figure 5This is a schematic diagram illustrating the distribution pattern of visual cortical neuron activation caused by background noise according to an embodiment of the present invention;

[0042] Figure 6 This is a structural block diagram of a biological brain neuron activation mechanism analysis system for visual search according to an embodiment of the present invention;

[0043] Figure 7 This is a structural diagram of a biological brain neuron activation mechanism analysis device for visual search according to an embodiment of the present invention. Detailed Implementation

[0044] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0046] The following description, with reference to the accompanying drawings, describes a method and apparatus for analyzing the activation mechanism of biological brain neurons for visual search, according to an embodiment of the present invention.

[0047] Example 1

[0048] Figure 1 This is a flowchart of a method for analyzing the activation mechanism of biological brain neurons for visual search according to an embodiment of the present invention, such as... Figure 1 As shown, it includes:

[0049] S1, based on the functional hierarchy of brain atlas, divides the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex, and higher visual cortex.

[0050] Specifically, the step in this invention of "dividing the mouse visual cortex into four levels—primary visual cortex, secondary visual cortex, intermediate visual cortex, and higher visual cortex—based on the functional hierarchy of brain atlases" is a crucial step in systematically analyzing the neuronal population coding mechanism during visual search. This step is technically implemented through cross-validation of neuroanatomy and functional neural atlases, combining the spatial distribution of microscopic imaging data with the temporal series characteristics of neuronal activity to classify and statistically analyze the functional hierarchy of multiple visual cortical brain regions.

[0051] In some implementations, this step first uses research findings on the anatomical structure and functional connectivity of the visual cortex in mouse brain atlases (such as the Allen Brain Atlas) to divide brain regions such as V1 (primary visual cortex), LM / RL (secondary visual cortex), LP / AL (intermediate visual cortex), and AM / PM (advanced visual cortex) according to their functional hierarchy in visual information processing. Specifically, V1, as the initial processing area for visual information, is responsible for basic edge, orientation, and motion detection; LM / RL participates in more complex shape recognition and spatial integration; LP / AL further integrates multimodal information and supports the regulation of visual attention; and AM / PM is involved in higher cognitive functions, such as target recognition and decision-making.

[0052] Furthermore, in this embodiment of the invention, neuronal activity data extracted by microscopic imaging is used to group and statistically analyze neurons in the aforementioned brain regions according to their activation characteristics in the visual search task. For example, in the Visual_E1 experiment, 85.7% of neurons in the V1 brain region were significantly activated, while the proportion in the AM / PM brain region was 57.8%, showing a progressively decreasing activation trend. In the Visual_E2 experiment, after adding background noise, the activation proportion of neurons at all levels increased, but the increase in the activation rate of the higher visual cortex was relatively small, indicating that it can still maintain a high selective processing ability under noise interference.

[0053] The statistical methods used in this step include t-tests and Cohen's d effect size analysis to assess the significant differences between neurons at different levels in the task state and the resting state. Through this functional hierarchical division, this invention can systematically reveal the hierarchical activation mechanism of neurons during visual search, providing a quantitative basis for understanding the step-by-step processing of visual information in the brain.

[0054] In practical applications, this step can be widely used for decoding brain region functions, modeling neural networks, and designing brain-like computing systems in neuroscience experiments. By dividing the visual cortex into four levels, this invention not only improves the structuring of neuronal activity pattern analysis but also provides a clear framework for subsequent brain network modeling and neural coding mechanism research, demonstrating significant innovation and practicality.

[0055] Furthermore, S1 includes:

[0056] S11 divides the primary visual cortex into the V1 brain region, the secondary visual cortex into the LM and RL brain regions, the intermediate visual cortex into the LP and AL brain regions, and the higher visual cortex into the AM and PM brain regions.

[0057] Specifically, this invention divides the mouse visual cortex into four functional levels (V1, LM / RL, LP / AL, AM / PM), a technique combining brain mapping functional hierarchy theory with experimental data-driven analysis of neuronal activity patterns. This step plays a crucial role in this invention, aiming to reveal the collaborative activation mechanism between different brain regions during visual search, thereby overcoming the limitation of treating the visual cortex as an isolated information island in existing technologies.

[0058] In some implementations, the V1 brain region, acting as the primary visual cortex, is responsible for basic visual feature extraction, such as edge, orientation, and motion detection. The secondary visual cortex regions LM and RL primarily participate in local feature integration and spatial attention regulation. The intermediate visual cortex regions LP and AL further integrate visual information, participating in target recognition and scene understanding. The higher visual cortex regions AM and PM are closely related to spatial memory, target selection, and decision-making. This invention constructs a neuronal activation path model from low-level perception to high-level cognition by dividing these brain regions according to functional hierarchy.

[0059] Furthermore, this classification was statistically validated based on the activation characteristics of neurons in microscopic imaging data. For example, in the Visual_E1 experiment, the activation rate of neurons in the V1 brain region reached 86.1%, while the activation rates in the AM and PM brain regions were 71.7% and 90.4%, respectively. In the Visual_E2 experiment, after adding background noise, the activation rate of V1 decreased to 81.9%, while the activation rates in the AM and PM brain regions were 67.7% and 67.7%, respectively. These data indicate that the activation rate of neurons decreases with increasing functional level, while background noise significantly increases the activation rate at each level, validating the effectiveness of this classification method in revealing the neural mechanisms of visual search.

[0060] This step provides a structured foundation for subsequent neuronal population coding analysis, brain network modeling, and sparse activation mechanism research in this invention, and is a core prerequisite for realizing multi-level collaborative analysis of the visual cortex.

[0061] S12, based on the functional localization of each brain region in visual information processing, the LLA and LI brain regions are assigned to the intermediate visual cortex, while the MMP, MMA, and RS brain regions are assigned to the higher visual cortex.

[0062] Specifically, in this invention, the LLA and LI brain regions are classified into the intermediate visual cortex, and the MMP, MMA, and RS brain regions are classified into the higher visual cortex. This is based on the systematic classification and statistical analysis of the functional hierarchy of the mouse visual cortex, thereby realizing the hierarchical modeling and quantitative evaluation of the neuronal population activation mechanism during visual search.

[0063] In this embodiment of the invention, the functional hierarchy of the visual cortex is based on the anatomical features of neuronal dendrites and axonal connections in brain atlases, as well as their functional roles in visual information processing. Specifically, the LLA (Lateral Lateral Amygdala) and LI (Lateral Interparietal) brain regions primarily undertake spatial attention and preliminary feature integration functions in visual information processing. Their neuronal response patterns lie between those of the primary visual cortex (e.g., V1) and the higher visual cortex (e.g., AM, PM), and are therefore classified as the intermediate visual cortex. The MMP (Medial Medial Prefrontal), MMA (Medial Amygdala), and RS (Retrosplenial) brain regions participate in more complex cognitive integration, memory retrieval, and goal-oriented behavior. Their neuronal responses exhibit stronger selectivity and integrativeness, and are therefore classified as the higher visual cortex. This classification process incorporates multidimensional information such as neuronal activation intensity, response time windows, and spatial distribution characteristics from microscopic imaging data, and is validated through K-means clustering and Pearson correlation analysis.

[0064] Furthermore, in the Visual_E1 experiment, the percentages of significantly activated neurons in the LLA and LI brain regions were 77.8% and 80.0%, respectively, while in the Visual_E2 experiment, the percentages were 28.9% and 57.9%, respectively. In contrast, the activation rates of the MMP, MMA, and RS brain regions were 91.6%, 46.2%, and 23.5% in Visual_E1, and 35.8%, 35.9%, and 28.9% in Visual_E2, respectively. These data indicate that higher visual cortical regions exhibit higher response selectivity in the presence of background noise, while the intermediate visual cortex assumes a transitional integration function during visual search.

[0065] This step plays a crucial role in the visual search experiments of this invention, particularly in constructing a multi-level neuronal activation model, providing a structured foundation for subsequent brain network analysis and sparse coding research. By mapping different brain regions to functional levels, the transmission paths and collaborative mechanisms of information between primary, intermediate, and higher cortices during visual search can be revealed more clearly. This is applicable to research on neurocognitive mechanisms, brain-computer interface development, and optimization of bio-inspired visual algorithms. This invention effectively solves the problem of treating different brain regions as isolated information islands in existing technologies by scientifically dividing the functional levels of the visual cortex, enabling a systematic analysis of neuronal population activation patterns during visual search. This step provides structured data support for subsequent neuronal correlation analysis, sparse activation ratio calculation, and background noise impact assessment, thereby significantly improving the depth of understanding of biological visual search mechanisms and the accuracy of model construction.

[0066] S2 introduces background image noise into the visual stimulus to obtain neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions.

[0067] Specifically, introducing background image noise into visual stimuli and acquiring neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions is one of the key steps in this invention to analyze the activation mechanism of neurons in the visual cortex. This step aims to simulate the complex visual environment in the real world, thereby studying the impact of background noise on the visual search behavior and neuronal population coding mechanism of mice.

[0068] In this embodiment of the invention, background image noise is introduced by superimposing random textures or natural scene noise onto the background of the target stimulus image. Specifically, using OpenCV or MATLAB image processing tools, the target stimulus image (such as a pattern of a specific shape or color) is embedded into a noisy background with high spatial frequency and low contrast. The noise intensity is adjusted by changing the variance (σ) of the grayscale values ​​of the background pixels. 2 The contrast is controlled, typically set to 0.1–0.3 times that of the target stimulus. Under noise-free conditions, the target stimulus is placed against a pure black or uniform grayscale background to eliminate interference from non-target information. The visual stimulus is presented through a high refresh rate (≥120Hz) LED backlit display to ensure the timing accuracy and spatial stability of the stimulus.

[0069] Specifically, in this embodiment of the invention, the parameters of background noise include spatial frequency (0.5–2.0 cycles / degree), contrast (0.1–0.3), and noise type (Gaussian noise, natural texture noise, or dynamic noise). The acquisition frequency of calcium ion fluorescence signals is typically 30–60 Hz. GCaMP6f is used as a calcium indicator, and its fluorescence signal is acquired through a 525 / 50 nm bandpass filter under 488 nm laser excitation. The significance of neuronal activity is determined based on the z-score-normalized fluorescence trace, and statistically validated using a T-test (p<0.05) and Cohen's d effect size (d>0.5).

[0070] This procedure is applicable to visual search experiments in mice with the head fixed, particularly for studying the neural coding mechanisms of the visual cortex (such as V1, LM, RL, AM, PM, etc.) in complex visual environments. By comparing neuronal activation patterns under noisy and noise-free conditions, the differences in response of the visual cortex at different functional levels can be revealed, providing experimental evidence for understanding the mechanisms of visual attention, target detection, and inhibition.

[0071] The technical advantage of this step lies in the fact that introducing background noise significantly improves the ecological validity of the experiment, making the neuronal activation data more closely resemble real-world visual search tasks. Experimental results show that under noisy conditions, the activation rate of neurons in the mouse visual cortex significantly increases (e.g., V1 brain region increases from 81.9% to 85.7%), revealing the promoting effect of background noise on neuronal population encoding. This step provides high-quality, comparable raw data for subsequent sparse activation analysis, functional hierarchy division, and brain network modeling, and is the core foundation for this invention to achieve the analysis of the neural mechanisms of visual search.

[0072] Furthermore, S2 includes:

[0073] S21, the target stimulus image is superimposed on the background noise image, wherein the background noise image is a random texture or natural scene image, and its brightness and contrast are adjustable.

[0074] Specifically, in this invention, the step of overlaying the target stimulus image with a background noise image is one of the key steps in analyzing the neuronal activation mechanism in mouse visual search experiments. The core technical principle of this step lies in introducing controllable background noise to simulate the complex visual environment of the real world, thereby studying the impact of background noise on the activation patterns of neurons in the mouse visual cortex. The specific implementation is as follows:

[0075] In some implementations, the target stimulus image is typically a visual target with well-defined contours and features (such as a specific shape, color, or texture), and its presentation position on the screen is randomly distributed with equal probability to avoid mice forming fixed search paths. Background noise images are random texture images or natural scene images to enhance the ecological validity of the visual stimulus. Random texture images can be generated using Gaussian white noise or fractal noise, with adjustable spatial frequency and contrast to simulate visual disturbance environments of varying complexity. Natural scene images can be selected from publicly available image databases (such as the van Hateren Natural Image Dataset), and used as background after preprocessing (such as grayscale conversion and contrast normalization).

[0076] Furthermore, during the overlay display process, the brightness and contrast parameters of the target stimulus image and the background noise image can be adjusted independently. The brightness adjustment range is typically 0-255 (8-bit grayscale), while the contrast adjustment is achieved by adjusting the grayscale histogram distribution of the image to make the target stimulus discernible in the background while controlling its salience. For example, the contrast of the target stimulus can be set to 1.5-3 times that of the background noise to ensure that mice can recognize the target without making it too prominent, thus avoiding distortion of experimental results.

[0077] In practical applications, this step is typically performed in a head-fixed experimental environment, where mice lick the spittoon as a feedback signal for correctly recognizing the target stimulus. The superimposed image is presented on a high refresh rate (≥120Hz) LED display to ensure the temporal accuracy of the visual stimulus, meeting the temporal resolution requirements in neuroscience experiments.

[0078] By introducing background noise, this invention effectively solves the problem of overly idealized backgrounds in existing visual stimulation experiments, making the experiments more closely resemble real-world visual search scenarios. This step provides a crucial experimental data foundation for subsequent sparsity analysis of neuronal activity patterns, functional hierarchy classification, and brain network modeling, thereby revealing the neural coding mechanism of the visual cortex in noisy environments, and has significant neuroscience research value and application prospects.

[0079] S22 quantifies the impact of background noise on visual search efficiency and neuron activation distribution by comparing the activation ratios of neurons at each level in noise-free and noisy experiments.

[0080] Specifically, this step quantifies the impact of background noise on the visual search efficiency and neuronal activation distribution in mice by comparing the activation ratio of neurons at each level in the noise-free experiment (Visual_E1) and the noisy experiment (Visual_E2). This is a key analytical step in this invention to reveal the neural coding mechanism of the visual cortex.

[0081] Further, this step involves statistical analysis of neuronal spike activity sequences (traces) extracted from calcium ion imaging data. First, the mouse visual cortex was divided into four functional levels: Level I (V1), Level II (LM / RL), Level III (LP / AL / LLA / LI), and Level IV (AM / PM / MMP / MMA / RS), ensuring functional grouping of neurons according to brain atlas functional hierarchy theory. Subsequently, the activation state of neurons extracted from the Visual_E1 and Visual_E2 experiments was statistically analyzed during the presentation of the target stimulus (Go stimulus). The activation state was determined based on the z-score-normalized fluorescence signal intensity; if its standard deviation within a specific time window exceeded a set threshold (e.g., 2σ), the neuron was considered significantly activated.

[0082] Specifically, this step uses the activation ratio (number of significantly activated neurons / total number of extracted neurons) as the core quantitative indicator. For example, in Visual_E1, the activation ratio of brain region V1 was 81.9%, while in Visual_E2 this ratio increased to 85.7%. Similarly, the activation ratio of brain region IV was 57.8% in Visual_E1, but increased to 67.7% in Visual_E2. The statistical significance of the changes in activation ratio was verified by t-test and Cohen's d effect size analysis (p<0.01), thus ensuring the reliability of the experimental results.

[0083] At the application level, this step is suitable for studying the neural coding mechanisms of mice under different visual environments, especially in visual search tasks that simulate real-world background noise. By introducing background image noise, the cognitive processing of animals in complex visual environments can be more realistically reflected, providing a neural mechanism basis for the development of brain-like visual systems.

[0084] The technical effect of this step is that it reveals the significant impact of background noise on the activation patterns of neurons in the visual cortex, verifies that the higher visual cortex has stronger adaptability and selective activation ability under noise interference, provides key data support for understanding the attention regulation mechanism in visual search, and has important value for neuroscience and brain-like computing applications.

[0085] S3, perform non-rigid motion correction, cell segmentation and trace extraction on the calcium ion fluorescence signal to obtain the activity characteristics of each neuron at different stimulation stages.

[0086] Specifically, performing non-rigid motion correction, cell segmentation, and trace extraction on the calcium ion fluorescence signal is one of the steps in this invention to achieve quantitative analysis of neuronal activity characteristics. The technical implementation principle is based on the preprocessing and feature extraction process of microscopic imaging data, aiming to extract biologically significant neuronal activity time series (trace) from the raw calcium ion imaging data, providing basic data support for subsequent neuronal population coding analysis and functional hierarchy classification.

[0087] In some implementations, non-rigid motion correction employs the NoRMCorre algorithm. This algorithm divides the field of view (FOV) into multiple overlapping local patches, performs motion correction on each patch based on normalized mutual correlation, and then stitches the corrected local images into a global image through smooth interpolation. Compared to traditional rigid registration methods, NoRMCorre effectively handles non-uniform deformations caused by natural mouse movement, and is particularly suitable for large field-of-view, long-term calcium ion imaging data. Its computational complexity is relatively low, making it suitable for online or batch processing.

[0088] In the cell segmentation stage, the CNMF (Constrained Non-negative Matrix Factorization) or CaImAn algorithm is used to decompose neuronal activity in the image into spatial components (cell footprint) and temporal components (trace) through non-negative matrix factorization. The CNMF algorithm introduces sparsity constraints and spatiotemporal consistency constraints during the segmentation process, which can effectively distinguish the fluorescence signals of adjacent neurons. Especially in low signal-to-noise ratio (SNR) regions, the segmentation accuracy and trace extraction stability are further improved by applying a structured model of calcium ion dynamics (such as the double exponential decay model).

[0089] After trace extraction, z-score normalization was applied to standardize the fluorescence signal of each neuron into a time series with a mean of 0 and a standard deviation of 1. Subsequently, the trace was divided into 10 phases, including 9 different target stimulus phases and a pure background noise stimulus phase, and the standard deviation of each phase was calculated as a neuronal activity feature. This step provides a quantifiable input vector for subsequent neuronal population classification, sparse activation ratio calculation, and functional hierarchy analysis.

[0090] This step plays a crucial role in the overall technical solution, directly impacting the decodeability and statistical significance of the neuronal population encoding. Through high-precision motion correction and cell segmentation, artifact interference can be effectively removed, improving the signal-to-noise ratio of the trace, thereby enhancing the accuracy and reliability of neuronal activity pattern analysis and laying a data foundation for revealing the sparse and hierarchical activation mechanisms in the visual search process.

[0091] Furthermore, S3 includes:

[0092] S31 uses the NoRMCorre algorithm to perform non-rigid motion correction on the imaging data, splits the field of view into overlapping patches and registers them separately, and then merges them through smooth interpolation to eliminate non-rigid artifacts caused by the natural movement of the mouse.

[0093] Specifically, in this invention, the NoRMCorre algorithm is used to perform non-rigid motion correction on imaging data, which is one of the key steps in achieving high-quality neuronal activity pattern analysis. This step mainly addresses the problem of non-rigid motion artifacts caused by the presence of minute natural movements (such as breathing, heartbeat, and muscle tremors) in mice with their heads fixed, thereby ensuring the accuracy and stability of subsequent neuron segmentation and spike sequence extraction.

[0094] Furthermore, NoRMCorre (Non-Rigid Motion Correction) is a non-rigid registration algorithm based on local image patches. Its core idea is to divide the entire field of view (FOV) into multiple patches with overlapping regions, and each patch undergoes independent motion correction. Specifically, each patch is typically set to 512×512 pixels, with the overlapping area being 20%–30% of the patch width to ensure the continuity of motion information between adjacent patches. Each patch obtains its local displacement field by calculating its cross-correlation coefficient with a reference frame and employing sub-pixel precision displacement estimation methods (such as interpolation cross-correlation). Finally, smooth interpolation (such as bicubic interpolation or Gaussian smoothing) is used to merge the displacement fields of all patches into a global non-rigid deformation field, performing non-rigid correction on the original image.

[0095] Specifically, the motion correction effect of NoRMCorre is typically evaluated by metrics such as improved signal-to-noise ratio (SNR) of the registered image, smoothness of the motion displacement field (e.g., smoothing using a Gaussian kernel σ = 2–4 pixels), and registration error (RMSE). In this invention, the reference frame used is the frame with the most stable fluorescence signal in the entire experimental sequence, typically the average image of the first 100 frames. A sliding window strategy is used for patch division to ensure coverage of the entire field of view and to retain sufficient overlap to avoid edge distortion. Furthermore, to meet the online processing requirements of large-capacity streaming data, the computational complexity of NoRMCorre is controlled at O(N·P), where N is the total number of frames and P is the number of patches, ensuring both real-time performance and computational efficiency.

[0096] This step is widely used in mesoscopic imaging experiments of the mouse visual cortex under head fixation, especially when using calcium ion indicators such as GCaMP6f to monitor neuronal activity. Non-rigid motion artifacts can significantly affect the spatiotemporal consistency of fluorescence signals. Non-rigid correction by NoRMCorre can effectively eliminate local deformations caused by natural mouse movement, improving the accuracy of subsequent CNMF and CaImAn algorithms in neuron segmentation and spike sequence extraction. This provides a reliable data foundation for revealing the sparse and hierarchical activation mechanisms in the visual search process.

[0097] This step significantly improves the spatiotemporal alignment quality of imaging data, making neuronal activity traces at different time points more comparable. Experimental data show that after NoRMCorre correction, the signal-to-noise ratio of neuronal activity is improved by an average of 15%–20%, and the signal drift caused by motion artifacts is reduced to less than 0.5 pixels, thus providing high-quality input data for subsequent neuronal population coding analysis, brain network construction, and sparse activation ratio calculation. The introduction of this step is an important guarantee for achieving high accuracy and high stability in the analysis of visual cortical neuronal activity patterns in this invention.

[0098] S32 uses the CNMF algorithm to perform cell segmentation and trace extraction on the corrected image, decomposes the spatiotemporal activity of neurons into spatial and temporal components, and introduces a structured calcium dynamics model in low signal-to-noise ratio regions to improve the accuracy of trace extraction.

[0099] Specifically, in this invention, the CNMF (Constrained Non-negative Matrix Factorization) algorithm is used for cell segmentation and trace extraction of the corrected calcium ion imaging image, which is one of the key steps in realizing neuronal activity pattern analysis. CNMF is a blind source separation method based on non-negative matrix factorization. Its core idea is to decompose the spatiotemporal activity signal of neurons into spatial components (representing the spatial distribution of neurons) and temporal components (representing the dynamic changes of calcium fluorescence in neurons), thereby realizing the automatic identification and activity trajectory extraction of individual neurons.

[0100] In its implementation, the CNMF algorithm first preprocesses the image sequence after NoRMCorre non-rigid motion correction, including background removal, motion artifact suppression, and signal-to-noise ratio (SNR) estimation. Subsequently, the algorithm iteratively optimizes the decomposition of the image data into multiple spatial components and corresponding temporal components. Each spatial component represents the morphological contour of a neuron, while the temporal components reflect the trace of calcium fluorescence changes of that neuron during the experiment. During the decomposition process, CNMF introduces sparsity and non-negativity constraints to ensure that the extracted components have biological significance and avoid the appearance of negative or non-localized spurious components.

[0101] In regions with low signal-to-noise ratios (SNR < 10), traditional CNMF methods are prone to missegmentation or inaccurate trace extraction. To address this, this invention introduces a structured calcium dynamics model into the CNMF framework. This model, based on the physiological characteristics of calcium ion fluorescence signals, parameterizes the temporal components. For example, it employs a double exponential model to describe the rise and decay of calcium fluorescence, with parameters including the rise time constant τ_rise, the decay time constant τ_decay, and the baseline fluorescence intensity F_0. By introducing these prior dynamic parameters, the algorithm can more accurately fit traces in low SNR regions, thereby significantly improving the extraction accuracy and stability of neuronal activity signals.

[0102] This step plays a crucial role in the overall technical solution, serving as a bridge between preceding and subsequent steps. Its output (spatial distribution and temporal trace of neurons) forms the basis for subsequent neuronal population coding analysis, functional hierarchy classification, and brain network modeling. Through high-precision trace extraction, this invention can effectively identify neurons associated with target stimuli and further reveal their sparse and hierarchical activation mechanisms during visual search, providing key data support for understanding the neural basis of biological visual cognition.

[0103] S4. Based on the activity characteristics, cluster analysis and significance testing methods are used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level.

[0104] Specifically, in this invention, based on the aforementioned activity characteristics, cluster analysis and significance testing are used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level. This is a crucial step in achieving an in-depth analysis of the population coding mechanism of neurons in the mouse visual cortex. This step, through multidimensional data processing and statistical modeling, reveals the activation characteristics and hierarchical organizational patterns of neurons in visual search tasks.

[0105] In some implementations, this step first standardizes the neuronal spike activity sequence (trace) extracted from calcium ion imaging, using z-score normalization to eliminate baseline differences between different neurons. Subsequently, the trace is divided into 10 phases, including 9 different target stimulus phases and a pure background noise stimulus phase. The standard deviation of neuronal activity in each phase is extracted as a feature vector to characterize its response intensity under different stimulus conditions. Based on these features, K-means clustering is used to group neurons, forming clusters of neurons with similar response patterns. Further, the response correlation between neurons is calculated using the Pearson correlation coefficient, and the false positive rate in multiple comparisons is controlled using the FDR (False Discovery Rate) correction method, thereby screening out a subset of neurons significantly correlated with the target stimulus.

[0106] In this invention, the K-means algorithm is used for cluster analysis. The number of clusters K is set according to the total number of neurons in the brain region and the diversity of response patterns, with selectable values ​​ranging from 6 to 10. A two-sample t-test is used for significance testing, with a significance threshold set at p < 0.01, and Cohen's d effect size is combined to assess the statistical significance of activation differences. The sparse activation ratio is defined as the ratio of the number of significantly activated neurons to the total number of extracted neurons; for example, in the Visual_E2 experiment, the sparse activation ratio was 32.6%. Furthermore, by calculating the activation ratios of different brain regions (I–IV), the response distribution of neurons at functional levels can be quantified, revealing their progressively decreasing activation characteristics.

[0107] This step is applied to the post-processing stage of mouse visual search experiments, particularly in mesoscale brain imaging data under head-fixed conditions. By analyzing neuronal responses to different visual stimuli (such as target images versus background noise), the information processing mechanisms of the visual cortex in complex environments can be assessed. For example, in the Visual_E1 and Visual_E2 experiments, comparing changes in activation ratios with and without background noise reveals the impact of visual noise on neuronal population encoding.

[0108] This step enables quantitative analysis of neuronal activation patterns, overcoming the limitations of traditional research that treats different brain regions as isolated information islands. It reveals the sparse activation mechanism and hierarchical response patterns of visual cortex neurons in visual search tasks. The combined application of clustering and significance testing not only improves the accuracy of neuronal response recognition but also provides a reliable data foundation for subsequent brain network modeling and cognitive mechanism deduction, demonstrating significant neuroscience research value and application prospects.

[0109] S5 calculates the z-score normalized standard deviation of neurons at different stimulation stages and combines it with the K-means clustering method to divide neurons into multiple clusters to identify their functional roles in the visual search task.

[0110] Specifically, in this invention, the normalized standard deviation of the z-score of neurons at different stimulation stages is calculated, and K-means clustering is used to divide neurons into multiple clusters to identify their functional roles in the visual search task. This step is a crucial link in the entire neuronal activation mechanism analysis process, used to reveal the population coding characteristics of neurons under different visual stimulation conditions, thereby supporting in-depth research on sparse activation and hierarchical activation mechanisms in the visual search process.

[0111] First, the calcium fluorescence trace of each neuron was divided into 10 phases, including 9 different target stimulus phases and a pure background noise stimulus phase. The time window length of each phase was set to a fixed value (e.g., 4 s) according to the experimental design, ensuring that the temporal resolution matched the neuronal response characteristics. Subsequently, the trace of each phase was z-score normalized. The normalized data eliminated the differences in baseline activity of individual neurons, making the response intensity of different neurons comparable under the same stimulus conditions.

[0112] Furthermore, the standard deviation of each neuron across the four stages is calculated as a characteristic of its response variability under different stimulus conditions. This standard deviation reflects the change in activation intensity of the neuron at a specific stimulus stage and is an important indicator of its sensitivity to the target stimulus. In some implementations, sliding window analysis can be combined to capture the dynamic changes in neuronal responses.

[0113] Specifically, this invention employs the K-means clustering algorithm to group neurons, where the number of clusters K can be set to 10, corresponding to the response patterns of 9 different target stimuli and pure background noise stimuli. The similarity between neurons is measured using Euclidean distance, and the optimal number of clusters is determined using the Elbow Method or Silhouette Score. The clustering results can be further used to analyze the differences in response patterns of neurons at different functional levels (such as V1, LM / RL, LP / AL, AM / PM).

[0114] Specifically, this step primarily identifies neuronal populations with different functional roles in visual search tasks, such as neurons sensitive to target stimuli, neurons responding to background noise, and neurons maintaining baseline activity during gray-screen periods. Combining this functional hierarchical classification further reveals the synergistic mechanism of the visual cortex in target detection and background suppression.

[0115] This method effectively improves the interpretability of neuronal population response patterns, providing quantitative evidence for studying neural coding mechanisms in visual search. Through cluster analysis, neuronal clusters with significant differences in different stimulus stages can be identified, thereby revealing their potential functional roles in visual attention, target recognition, and background suppression.

[0116] The biological brain neuron activation mechanism analysis method for visual search in this invention combines z-score normalized standard deviation and K-means clustering to classify neurons by function, further improving the recognition accuracy of neuronal group behavior patterns and functional roles in visual search tasks, and realizing dynamic analysis of neural coding mechanisms from the individual to the group level.

[0117] Example 2

[0118] This invention utilizes large-field-of-view, high-resolution biological brain imaging and data processing techniques based on calcium ion fluorescent labeling, and analyzes the sparse activation and hierarchical activation mechanisms of visual search in mice. The specific implementation process is as follows:

[0119] Mice (TIT2L-GC6s-ICL-tTA2, Ai162, male, 8-12 weeks old) were anesthetized with isoflurane. The mouse heads were fixed under an Olympus 25× / 1.05NA microscope. A cranial window was created by removing the visual cortex (Interaural 0.2 mm, Bregma 5.2 mm) and a 10 mm diameter cranial flap (AP-5.2, ML7.1) above the surrounding cortical tissue, leaving the dura mater intact. The exposed cortical surface was covered with a 12 mm diameter glass coverslip (No.1 Thickness, 64-0700, CS-5R, Warner Instruments), fixed to the skull with UV-cured cyanoacrylate adhesive (Loctite, 4305), and a matching metal headpiece was glued to the skull using dental cement acrylic to secure the imaging instrument interface to the head. Neurons were fluorescently labeled with GCaMP6f as a calcium ion indicator. Brain imaging studies should begin at least 14 days after surgery.

[0120] Image registration, such as Figure 2 As shown in (1), brain regions are segmented as follows: Figure 2As shown in (2), non-rigid motion artifacts are generated due to the mechanical displacement between the imaging instrument and the mouse brain caused by the natural movement of the mouse. NoRMCorre overcomes the shortcomings of rigid motion alignment and has low computational cost, thus enabling online real-time registration of large-volume streaming data. By splitting the field of view (FOV) into overlapping patches, which are registered separately and then merged by smooth interpolation, NoRMCorre can account for non-uniform artifacts along all axes, thus capturing non-rigid brain motion very well compared to other available non-rigid registration methods.

[0121] After image registration, cell locations are identified and their neural activity is deconvolved. Estimates of neural activity are extracted from fluorescent traces, and then traces are extracted, such as... Figure 2 As shown in (3) of the paper. The CNMF algorithm is used to decompose spatiotemporal activity into spatial components with local structure and temporal components simulating calcium dynamics. In the low signal-to-noise ratio (SNR) region, by applying a structured model of calcium index dynamics, a significant improvement in the spatial trace of observed neurons can be obtained.

[0122] To analyze neural responses to visual stimuli, neurons associated with the visual stimulus can be screened based on collective measurements of neuronal reliability. Neurons observed at specific locations during each trial can be used. For example... Figure 2 As shown in (4), in order to obtain the correlation map between different rounds, the Pearson correlation coefficient of the average stimulus response intensity of each neuron was calculated, and N was used. relation The average value of each neuron is used to capture the central tendency of the entire population. Principal component analysis (PCA) is used to reduce the dimensionality of the features, and then neurons that have significant responses during stimulation are grouped. Figure 2 As shown in (5) above, trace is used as the input vector. For N trial ×N time Data within a 3×3 time window is analyzed using t-distributed random neighborhood embedding (tSNE) to select the most relevant N. relation Visualize and analyze individual neurons.

[0123] To compare changes in neural activity intensity, the curves were normalized using the mean in each round. This experiment divided the time trajectory into 10 phases, including 9 different target stimulus phases and a pure background noise stimulus phase, and used their standard deviations in the corresponding phases as features for each neuron. Then, K-means clustering was used to divide the neurons into four clusters based on these features, and the percentage of activated neurons was calculated.

[0124] To better describe the implementation process of the neuronal activation mechanism analysis in this invention, the embodiments of this invention are as follows: Figure 3 (1) and Figure 3 (4) in this paper cites some images from the invention patent "A Visual Search Experiment Method for Head-Fixed Brain Imaging" that our unit has submitted to the State Patent Office. The trace data analysis of this invention patent is a follow-up operation of the visual search experiment mentioned in the above patent. It is mainly to explore all neurons related to the target stimulus and the neuronal activity patterns under the influence of background noise, and does not involve patent conflict.

[0125] like Figure 3 As shown in (1), in the visual search experiment of this embodiment of the invention, the screen is divided into 3×3 squares (the light white solid lines are not visible during the experiment and only serve as an illustration). The target stimulus image appears randomly in a certain square with equal probability. The mouse performs a visual search on the screen to find the target stimulus image. When the target stimulus is presented on the screen and the mouse licks the water outlet, it is recorded as a correct response.

[0126] In the Visual_E1 experiment, the distribution of all relevant activated neurons is as follows: Figure 3 As shown in (2) of the figure, taking Ai162#07 mice as an example, statistical analysis revealed that: a total of 1957 neurons were extracted from the V1 brain region, of which 2547 neurons were significantly activated, accounting for 86.1%; a total of 460 neurons were extracted from the PM brain region, of which 416 neurons were significantly activated, accounting for 90.4%; a total of 452 neurons were extracted from the AM brain region, of which 324 neurons were significantly activated, accounting for 71.7%; a total of 463 neurons were extracted from the RL brain region, of which 407 neurons were significantly activated, accounting for 87.9%; a total of 309 neurons were extracted from the AL brain region, of which 260 neurons were significantly activated, accounting for 84.1%; and the LM brain region... A total of 347 neurons were extracted from the first brain region, of which 300 were significantly activated, accounting for 86.5%; 250 neurons were extracted from the MMP brain region, of which 229 were significantly activated, accounting for 91.6%; 327 neurons were extracted from the RS brain region, of which 255 were significantly activated, accounting for 78.0%; 169 neurons were extracted from the MMA brain region, of which 78 were significantly activated, accounting for 46.2%; 130 neurons were extracted from the LI brain region, of which 104 were significantly activated, accounting for 80.0%; and 167 neurons were extracted from the LLA brain region, of which 130 were significantly activated, accounting for 77.8%. To find all relevant neurons, a total of 6031 neurons were extracted in the Visual_E1 experiment, of which 5050 were significantly activated, accounting for 83.7%, and the specific distribution is as follows: Figure 3 As shown in (3) of the text.

[0127] like Figure 3 As shown in (4) above, with Figure 3 There are two differences in (1) of this invention. First, background image noise is added to the embodiment of this invention. Second, the embodiment of this invention only counts the average number of neurons activated in a single experiment. Similarly, in the visual search experiment of this invention, the screen is divided into 3×3 squares (the light white solid lines are not visible during the experiment and only serve an illustrative purpose). The target stimulus image appears randomly in a certain square with equal probability. The mouse performs a visual search on the screen to find the target stimulus image. If the target stimulus is presented on the screen and the mouse licks the water outlet, it is recorded as a correct response.

[0128] In the Visual_E2 experiment, the distribution of neurons activated under background interference was found as follows: Figure 3 As shown in (5) of the figure, taking Ai162#07 mice as an example, statistical analysis revealed that: a total of 3011 neurons were extracted from the V1 brain region, of which 1016 neurons were significantly activated, accounting for 33.7%; a total of 649 neurons were extracted from the PM brain region, of which 206 neurons were significantly activated, accounting for 31.7%; a total of 557 neurons were extracted from the AM brain region, of which 173 neurons were significantly activated, accounting for 30.0%; a total of 272 neurons were extracted from the RL brain region, of which 98 neurons were significantly activated, accounting for 36.0%; a total of 201 neurons were extracted from the AL brain region, of which 89 neurons were significantly activated, accounting for 44.3%; and the LM brain region was significantly activated. A total of 261 neurons were extracted from the brain regions, of which 93 were significantly activated, accounting for 35.6%; 302 neurons were extracted from the MMP brain region, of which 108 were significantly activated, accounting for 35.8%; 898 neurons were extracted from the RS brain region, of which 211 were significantly activated, accounting for 23.5%; 451 neurons were extracted from the MMA brain region, of which 162 were significantly activated, accounting for 35.9%; 19 neurons were extracted from the LI brain region, of which 11 were significantly activated, accounting for 57.9%; and 90 neurons were extracted from the LLA brain region, of which 26 were significantly activated, accounting for 28.9%. To find all relevant neurons, a total of 6729 neurons were extracted in the Visual_E2 experiment, of which 2193 were significantly activated, accounting for 32.6%, and the specific distribution is as follows: Figure 3 As shown in (6) of the text.

[0129] The sparse activation ratio is defined in this embodiment of the invention as follows: Calculations show that neurons in the biological visual cortex use sparse activation for visual searching, with a single activation rate of only 32.6 / 83.7 = 38.9%.

[0130] There are numerous brain regions related to visual function, but studies on the functional hierarchy of brain atlases show that the visual cortex can be divided into four layers, such as... Figure 4 As shown in (1) of the present invention. Based on the brain regions of interest in the microscopic imaging of this invention, the neurons of the visual cortex can be divided into the following four levels: I (V1), II (LM / RL), III (LP / AL), IV (AM / PM). Figure 2 Some of the brain regions mentioned in the text are also classified into corresponding levels, such as LLA and LI being classified into level III, and MMP, MMA, and RS being classified into level IV.

[0131] To investigate the neuronal activation mechanisms at different functional levels of the cortex, this invention grouped and statistically analyzed neurons extracted from the Visual_E1 and Visual_E2 experiments according to their functional levels. In the Visual_E1 experiment, the neuronal activation proportions were as follows: I(V1): 2579 / 3011 = 81.9%, II(LM / RL): 370 / 533 = 69.4%, III(LP / AL): 143 / 220 = 65%, IV(AM / PM): 709 / 1226 = 57.8%. In the Visual_E2 experiment, the neuronal activation proportions were as follows: I(V1): 2071 / 2530 = 85.7%, II(LM / RL): 550 / 754 = 72.9%, III(LP / AL): 131 / 185 = 70.8%, IV(AM / PM): 847 / 1251 = 67.7%. Longitudinal comparisons reveal that as functional levels increase, the activation rate of neurons in higher visual cortex decreases progressively compared to the primary visual cortex. Lateral comparisons show that, due to the inclusion of background image noise in the Visual_E2 experiment compared to the Visual_E1 experiment, the activation rate of neurons at the same level is higher in Visual_E2 than in Visual_E1. This indirectly confirms that the visual noise in the experiment effectively elicited a response in the mouse visual cortex.

[0132] To further analyze the impact mechanism of interference noise on visual search, this embodiment of the invention analyzes the proportion of all neurons related to the target stimulus and the proportion of all extracted neurons. For example... Figure 5 As shown in (1) of the diagram, the result in the Visual_E1 experiment is: 4327 / 6031 = **70.7% ± 0.113. (See diagram (1) for details.) Figure 5As shown in (2), the proportion of all relevant neurons to all extracted neurons in the Visual_E2 experiment was: Visual_E_2=5570 / 6676=**78.4%±0.096. Here, `*` represents significance, specifically ***Sig.<0.001; **Sig.<0.01; *Sig.<0.05. Grouped statistical analysis revealed that the activation rate of neurons under background noise interference was significantly higher than that without background noise, and this was observed at different levels.

[0133] like Figure 6 As shown, this invention also designs a biological brain neuron activation mechanism analysis system for visual search, including: a calcium ion imaging module, a non-rigid motion correction module, a cell segmentation and trace extraction module, a dimensionality reduction and clustering analysis module, a neuron correlation analysis module, and a functional hierarchy division and saliency analysis module.

[0134] Among them, the calcium ion imaging module is usually observed using single-photon or two-photon mesoscopic microscopic imaging equipment. When the neurotransmitters of neurons are active, the movement of calcium ions in the nerve cells is also active. At this time, a laser of a specific wavelength will cause fluorescent proteins to bind to calcium ions and emit fluorescence of a specific wavelength, which is captured by the microscopic imaging equipment, thereby realizing real-time observation of neuronal activity in the mouse brain and obtaining neuronal activation image data.

[0135] The non-rigid motion correction module is used to eliminate image jitter caused by mouse movement or equipment vibration. Many existing algorithms, such as NoRMCorre, can eliminate non-rigid displacement in raw microscopic imaging.

[0136] The cell segmentation and trace extraction module converts the corrected microscopic imaging data into trace data of neuronal activation, which is a real-number time series. First, supervised learning algorithms are used to label and identify neurons. Then, signal-to-noise ratio and calcium ion dynamics are estimated from the image to achieve trace extraction.

[0137] The dimensionality reduction and clustering analysis module is used to test the effectiveness of target stimuli in visual search experiments. Factors such as mouse strain differences, screen size, and the distance between the screen and the mouse's eyeballs can all affect the mouse's visual search. The controlled variable method requires eliminating irrelevant variables, which need to be eliminated through multiple rounds of fine-tuning. This invention uses dimensionality reduction and clustering analysis algorithms such as tSNE, PCA, and K-means to test visual stimuli. If neurons can be well divided into N... trail ×N trail This demonstrates the effectiveness of visual stimulation, ensuring the validity of subsequent analyses of neuronal activation mechanisms.

[0138] The neuron correlation analysis module is used to perform Pearson correlation statistical analysis on the Visual_E1 and Visual_E2 experiments mentioned above, and to correct false positive results through FDR correction, thereby quantifying the sparse activation mechanism of neurons in the visual cortex.

[0139] The functional hierarchy segmentation and significance analysis module is used to segment multiple brain regions of the visual cortex mentioned above into functional levels, and to perform significance difference analysis through data statistical analysis methods such as T-test and Cohen's d effect size analysis, thereby quantifying the impact of background noise on visual search and analyzing the hierarchical activation mechanism of neurons in the visual cortex.

[0140] The biological brain neuron activation mechanism analysis method for visual search in this invention combines z-score normalized standard deviation and K-means clustering to classify neurons by function, further improving the recognition accuracy of neuronal group behavior patterns and functional roles in visual search tasks, and realizing dynamic analysis of neural coding mechanisms from the individual to the group level.

[0141] Example 3

[0142] To achieve the above embodiments, such as Figure 7 As shown, this embodiment also provides a biological brain neuron activation mechanism analysis device 10 for visual search, including:

[0143] The visual cortex hierarchical division module 100 is used to divide the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex and higher visual cortex based on brain atlas functional hierarchical division.

[0144] The noise stimulus data acquisition module 200 is used to introduce background image noise into visual stimuli and acquire neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions.

[0145] The neuronal activity feature extraction module 300 is used to perform non-rigid motion correction, cell segmentation, and trace extraction on the calcium ion fluorescence signal to obtain the activity features of each neuron at different stimulation stages.

[0146] The activation pattern quantification analysis module 400 is used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation pattern of neurons at each level based on the activity characteristics, using cluster analysis and significance testing methods.

[0147] Furthermore, the visual cortex hierarchy segmentation module 100 is also used for:

[0148] The primary visual cortex is divided into the V1 brain region, the secondary visual cortex into the LM and RL brain regions, the intermediate visual cortex into the LP and AL brain regions, and the higher visual cortex into the AM and PM brain regions.

[0149] Based on the functional localization of each brain region in visual information processing, the LLA and LI brain regions are assigned to the intermediate visual cortex, while the MMP, MMA, and RS brain regions are assigned to the higher visual cortex.

[0150] Furthermore, the noise stimulus data acquisition module 200 is also used for:

[0151] The target stimulus image is overlaid with a background noise image, where the background noise image is a random texture or a natural scene image, and its brightness and contrast are adjustable.

[0152] By comparing the activation ratios of neurons at each level in noise-free and noisy experiments, the impact of background noise on visual search efficiency and neuron activation distribution is quantified.

[0153] Furthermore, the neuronal activity feature extraction module 300 is also used for:

[0154] The NoRMCorre algorithm was used to perform non-rigid motion correction on the imaging data. The field of view was split into overlapping patches and registered separately. Then, smooth interpolation was used to merge the patches to eliminate non-rigid artifacts caused by the natural movement of the mouse.

[0155] The CNMF algorithm is used to perform cell segmentation and trace extraction on the corrected image, decomposing the spatiotemporal activity of neurons into spatial and temporal components, and introducing a structured calcium dynamics model in low signal-to-noise ratio regions to improve the accuracy of trace extraction.

[0156] Furthermore, device 10 also includes:

[0157] The functional role recognition module is used to divide neurons into multiple clusters by calculating the z-score normalized standard deviation of neurons at different stimulus stages and combining it with the K-means clustering method, so as to identify their functional roles in visual search tasks.

[0158] The biological brain neuron activation mechanism analysis device for visual search in this invention combines z-score normalized standard deviation and K-means clustering to classify neurons by function, further improving the recognition accuracy of neuronal group behavior patterns and functional roles in visual search tasks, and realizing dynamic analysis of neural coding mechanisms from individual to group level.

[0159] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0160] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for analyzing the activation mechanism of biological brain neurons for visual search, characterized in that, include: S1, based on the functional hierarchy of brain atlas, divides the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex and higher visual cortex. S2, introduce background image noise into the visual stimulus to obtain neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions; S3, perform non-rigid motion correction, cell segmentation and trace extraction on the calcium ion fluorescence signal to obtain the activity characteristics of each neuron at different stimulation stages; S4. Based on the activity characteristics, cluster analysis and significance testing methods are used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level.

2. The method as described in claim 1, characterized in that, S1 includes: S11 divides the primary visual cortex into the V1 brain region, the secondary visual cortex into the LM and RL brain regions, the intermediate visual cortex into the LP and AL brain regions, and the higher visual cortex into the AM and PM brain regions. S12, based on the functional localization of each brain region in visual information processing, the LLA and LI brain regions are assigned to the intermediate visual cortex, while the MMP, MMA, and RS brain regions are assigned to the higher visual cortex.

3. The method as described in claim 1, characterized in that, S2 includes: S21, superimpose the target stimulus image with the background noise image, wherein the background noise image is a random texture or natural scene image, and its brightness and contrast are adjustable; S22 quantifies the impact of background noise on visual search efficiency and neuron activation distribution by comparing the activation ratios of neurons at each level in noise-free and noisy experiments.

4. The method as described in claim 1, characterized in that, The S3 includes: S31 uses the NoRMCorre algorithm to perform non-rigid motion correction on the imaging data, splits the field of view into overlapping patches and registers them separately, and then merges them through smooth interpolation to eliminate non-rigid artifacts caused by the natural movement of the mouse. S32 uses the CNMF algorithm to perform cell segmentation and trace extraction on the corrected image, decomposes the spatiotemporal activity of neurons into spatial and temporal components, and introduces a structured calcium dynamics model in low signal-to-noise ratio regions to improve the accuracy of trace extraction.

5. The method as described in claim 1, characterized in that, Also includes: S5 calculates the z-score normalized standard deviation of neurons at different stimulation stages and combines it with the K-means clustering method to divide neurons into multiple clusters to identify their functional roles in the visual search task.

6. A device for analyzing the activation mechanism of biological brain neurons for visual search, characterized in that, include: The visual cortex hierarchy module is used to divide the mouse visual cortex into four levels: primary visual cortex, secondary visual cortex, intermediate visual cortex, and higher visual cortex, based on the brain atlas functional hierarchy classification method. The noise stimulation data acquisition module is used to introduce background image noise into visual stimuli to acquire neuronal calcium ion fluorescence signal data of mice under noisy and noise-free conditions. The neuronal activity feature extraction module is used to perform non-rigid motion correction, cell segmentation, and trace extraction on the calcium ion fluorescence signal to obtain the activity features of each neuron at different stimulation stages. The activation pattern quantification analysis module is used to quantify the sparse activation ratio of neurons related to the target stimulus and the activation patterns of neurons at each level based on the activity characteristics, using cluster analysis and significance testing methods.

7. The apparatus as claimed in claim 6, characterized in that, The visual cortex hierarchy segmentation module is also used for: The primary visual cortex is divided into the V1 brain region, the secondary visual cortex into the LM and RL brain regions, the intermediate visual cortex into the LP and AL brain regions, and the higher visual cortex into the AM and PM brain regions. Based on the functional localization of each brain region in visual information processing, the LLA and LI brain regions are assigned to the intermediate visual cortex, while the MMP, MMA, and RS brain regions are assigned to the higher visual cortex.

8. The apparatus as claimed in claim 6, characterized in that, The noise stimulation data acquisition module is also used for: The target stimulus image is overlaid with a background noise image, where the background noise image is a random texture or a natural scene image, and its brightness and contrast are adjustable. By comparing the activation ratios of neurons at each level in noise-free and noisy experiments, the impact of background noise on visual search efficiency and neuron activation distribution is quantified.

9. The apparatus as claimed in claim 6, characterized in that, The neuronal activity feature extraction module is also used for: The NoRMCorre algorithm was used to perform non-rigid motion correction on the imaging data. The field of view was split into overlapping patches and registered separately. Then, smooth interpolation was used to merge the patches to eliminate non-rigid artifacts caused by the natural movement of the mouse. The CNMF algorithm is used to perform cell segmentation and trace extraction on the corrected image, decomposing the spatiotemporal activity of neurons into spatial and temporal components, and introducing a structured calcium dynamics model in low signal-to-noise ratio regions to improve the accuracy of trace extraction.

10. The apparatus as claimed in claim 6, characterized in that, Also includes: The functional role recognition module is used to divide neurons into multiple clusters by calculating the z-score normalized standard deviation of neurons at different stimulus stages and combining it with the K-means clustering method, so as to identify their functional roles in visual search tasks.

Images