Resting state local brain function index description method based on T2*time sequence and abnormal brain activity positioning device

Abstract

The invention discloses a resting state local brain function index description method based on a T2 * time sequence and an abnormal brain activity positioning device. According to the resting state local brain function index description method, a multi-echo gradient echo sequence is adopted for magnetic resonance data collection, multiple signals of different echo time are collected after radio frequency pulse excitation each time, T2 * values at all time points are fitted by applying a single-index attenuation model, T2 * time sequences are spliced in sequence, and the rest state local brain function indexes are obtained. And then low-frequency amplitude ALFF and a local consistency ReHo index are calculated, a resting state local brain function is described, and support is provided for accurate and reliable positioning of abnormal brain activities. The abnormal brain activity positioning device positions abnormal brain activities by comparing resting state local brain function indexes of different subjects. According to the method, the signals are automatically separated into the nerve correlation signals and the physiological noise through a physical model, and the method has specificity discovery capability and higher repeatability in a deep brain area and an area with serious magnetic sensitive artifacts.

Description

Technical Field

[0001] This invention belongs to the field of magnetic resonance imaging technology and relates to brain function identification, specifically to a method for characterizing resting-state local brain function indicators based on T2* time series and a device for locating abnormal brain activity. Background Technology

[0002] In neuroscience research and clinical practice, accurately characterizing resting-state regional brain function indicators is crucial for locating abnormally active areas in the brain, understanding the pathophysiological mechanisms of various neurological and psychiatric diseases, assisting in clinical diagnosis, and guiding treatment plans. Currently, functional magnetic resonance imaging (fMRI) is the main non-invasive technique for achieving this goal, with mainstream implementation methods including single-echo and multi-echo imaging techniques.

[0003] The single-echo functional magnetic resonance imaging (fMRI) method, after each radiofrequency pulse excitation, waits for a preset fixed echo time designed to have good sensitivity to the BOLD (blood oxygen level dependent) effect across the entire brain before acquiring a complete T2*-weighted brain functional image. By repeating this acquisition process, a single signal intensity time series is generated for each brain voxel, used for subsequent statistical analysis to locate brain activity. Due to its simple data structure and high computational efficiency, it is one of the most widely used fMRI techniques. However, it has unavoidable fundamental limitations when facing the inherent T2* spatial heterogeneity of the brain. Catastrophic signal loss occurs in magnetically sensitive areas with extremely short T2* values ​​in the orbitofrontal cortex and temporal lobe, as well as in deep brain regions such as the hippocampus and amygdala. These areas are crucial for higher cognitive and emotional functions and are often the origin of abnormal brain activity. Therefore, severe signal attenuation directly leads to extremely low sensitivity in detecting neural activity in these key areas, or even the formation of signal blind spots, resulting in inaccurate characterization of brain functional indicators and making subsequent abnormal brain activity localization results unreliable.

[0004] The Multi-Echo Optimal Combination (ME-OC) method improves the quality of the BOLD signal by acquiring multiple echo signals and performing post-processing to address the spatial heterogeneity of brain T2* values. This scheme acquires T2*-weighted signals at multiple different echo time points after a single radiofrequency pulse excitation. Based on the T2* characteristics of each brain voxel, the signals from different echoes are assigned corresponding weights, and then combined into an optimized, single signal intensity time series to improve the signal loss problem in single-echo functional magnetic resonance imaging (fMRI). However, the hybrid processing method in the ME-OC method cannot fundamentally distinguish the different physical sources of the signal, inevitably leading to the amplification of the S0 component, the main noise carrier, in the final signal, thus severely contaminating the true BOLD signal.

[0005] In summary, the integrity and purity of signals from existing mainstream magnetic resonance imaging (MRI) techniques cannot be guaranteed, and they cannot accurately characterize brain function indicators, especially when applied to magnetically sensitive and deep brain regions, where the defects are even more pronounced. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a method for characterizing resting-state local brain function indicators based on T2* time series and a device for locating abnormal brain activity. By directly quantifying the biophysical source of the BOLD effect, it systematically solves the problems of signal loss and noise amplification faced by existing technologies in accurately locating abnormal brain activity in magnetically sensitive and deep brain regions, providing more valuable and accurate indicators for subsequent abnormal brain activity localization.

[0007] A method for characterizing resting-state local brain function indicators based on T2* time series includes the following steps:

[0008] Step 1: Multi-echo data acquisition

[0009] Magnetic resonance data acquisition was performed using a multi-echo gradient-recalled echo (ME-GRE) sequence, acquiring signals at N different echo times after each radio frequency pulse excitation.

[0010] Step Two: Core Quantification and Separation

[0011] For each voxel, a set of multi-echo signal data (TE) is acquired after each radio frequency pulse excitation. t1 TE t2 …TE tN The initial signal strength at time point t is obtained by fitting the signal using a single exponential decay model. and T2* value :

[0012]

[0013] Where n = 1, 2, ..., N, Represents the nth echo signal data TE tn The signal strength.

[0014] T2* values ​​under multiple TRs By piecing them together in order, we obtain... Time series data directly quantifies the Bollinger Band effect, reflecting neural activity. It is itself a purified signal from a physical model, readily usable for subsequent analysis. Initial signal strength. The captured signal fluctuations unrelated to neural activity are a quantitative representation of the system's physiological noise, which can be discarded or used separately in the analysis. The physical model described above, through separation and denoising, mathematically forces the signal to be decomposed into two independent parameters: the initial signal strength and the effective transverse relaxation time. This completes the purification of the BOLD signal when generating data.

[0015] Step 3: Characterization of resting-state regional brain function indicators

[0016] For the data generated in step two, which has already been purified by the physical model Statistical analysis of time series data was performed to calculate low-frequency amplitude ALFF and local consistency ReHo index, characterizing resting-state local brain function and providing data support for the accurate and reliable localization of abnormal brain activity.

[0017] A device for localizing abnormal brain activity based on T2* time-series resting-state local brain function indices includes:

[0018] The multi-echo data acquisition module uses a multi-echo gradient echo sequence to acquire magnetic resonance data, acquiring N magnetic resonance signals with different echo times after each radio frequency pulse excitation.

[0019] The data quantization and separation module is used to collect a set of multi-echo signal data (TE) after a radio frequency pulse excitation. t1 TE t2 …TE tN A single exponential decay model was applied for fitting to obtain the T2* value at a given time point, which was then concatenated sequentially to form... Time series.

[0020] The resting-state local brain function index characterization module receives the output from the data quantization and separation module. Time series analysis was used to calculate low-frequency amplitude and local consistency indices to characterize resting-state local brain function.

[0021] The abnormal brain activity localization module locates abnormal brain activity by comparing resting-state local brain function indices of different subjects.

[0022] The present invention has the following beneficial effects:

[0023] 1. To address the challenges of fMRI analysis in magnetically sensitive and deep brain regions, a novel technical approach is proposed. Its fundamental paradigm shift lies in the fact that the measurement target is no longer a signal intensity proxy that indirectly reflects the BOLD effect, but rather a direct quantification of its biophysical source, namely the T2* relaxation time itself.

[0024] 2. By using a unified physical model to mathematically model the complete signal attenuation curve, rather than relying on the signal amplitude at fixed time points, the signal lost in the short T2* region can be effectively recovered, thus solving the signal attenuation problem of the single-echo (SE) method. At the same time, the physical model automatically separates the signal into neural-related signals and physiological noise, thereby avoiding the noise pollution problem caused by signal mixing in the multi-echo optimal combination (ME-OC) method.

[0025] 3. This method has specific detection capabilities and higher reproducibility in deep brain regions and areas with severe magnetic susceptibility artifacts. Attached Figure Description

[0026] Figure 1 A flowchart of a method for characterizing resting-state local brain function indicators based on T2* time series;

[0027] Figure 2 This is a comparison of the average signal retention rate of different methods in different brain regions in the embodiments;

[0028] Figure 3 The results of the test-retest reliability distribution of the ALFF index in the subcortical region are shown in the figure. (a) is based on single echo SE, (b) is based on multi-echo ME-OC, and (c) is based on T2* fluctuation.

[0029] Figure 4 The diagram shows the test-retest reliability distribution of the ReHo index in the subcortical region, where (a) is based on single echo SE, (b) is based on multi-echo ME-OC, and (c) is based on T2* fluctuations.

[0030] Figure 5 The test-retest reliability distribution of the ALFF index in different subcortical regions;

[0031] Figure 6 The test-retest reliability distribution of the ReHo index in different subcortical regions;

[0032] Figure 7 A system block diagram of an abnormal brain activity localization device based on T2* time series resting-state local brain function indicators;

[0033] Figure 8 The results of paired-samples t-tests for the eye-opening and closing task under different methods for the ALFF index are shown. Detailed Implementation

[0034] The present invention will be further explained below with reference to the accompanying drawings;

[0035] like Figure 1 As shown, the method for characterizing resting-state local brain function indicators based on T2* time series specifically includes the following steps:

[0036] Step 1: Multi-echo signal acquisition

[0037] Magnetic resonance data acquisition was performed using a multi-echo gradient-recalled echo (ME-GRE) sequence, acquiring signals at four different echo times (TE) after each radio frequency pulse excitation. t1 TE t2 TE t3 TE t4 Unlike the single-echo (SE) method, which acquires signals at only a single fixed time point, the method captures the entire process of signal decay over time by acquiring data from multiple TE points. This is the physical basis for overcoming the signal loss problem that inevitably occurs in the short T2* region due to single-point sampling in the single-echo method.

[0038] As a preferred embodiment, the number of echo signals acquired after each radiofrequency pulse excitation can be adjusted according to clinical or research requirements for time resolution and signal-to-noise ratio. Fewer acquired echo signals result in shorter acquisition times, suitable for clinical scenarios requiring high patient cooperation. More acquired echo signals lead to higher fitting accuracy and more accurately reflect signal changes caused by spontaneous brain activity.

[0039] Step Two: Core Quantification and Separation

[0040] For each voxel, a set of multi-echo signal data is acquired after each RF pulse excitation, and a single exponential decay model is applied for fitting to obtain the initial signal strength at time point t. and T2* value :

[0041]

[0042] Where n=1,2,3,4. The difference between this application and the Multi-Echo Optimal Combination (ME-OC) method is that this application does not perform weighted averaging and synthesis of multi-echo signals, but directly fits the above physical model through an optimization algorithm. During the fitting process, the signal is mathematically forcibly decomposed into two independent parameters: initial signal strength and effective transverse relaxation time.

[0043] As a preferred embodiment, to enhance the robustness of fitting in low signal-to-noise ratio environments such as magnetically sensitive and deep brain regions, a weighted nonlinear least squares method is used to fit the single exponential decay model. Weighting is applied based on the signal-to-noise ratio of different echoes to emphasize more reliable signal points. When fitting fails to converge, the Levenberg-Marquardt algorithm, combining gradient descent and the Gauss-Newton method, is employed to achieve better stability. Furthermore, the physiological prior distribution of T2* values ​​can be introduced, and a Bayesian fitting method can be used to constrain the fitting results through prior knowledge, improving its accuracy under noise interference.

[0044] As a preferred embodiment, the multi-echo data acquired in step one is decomposed and denoised using principal component analysis (PCA) before fitting. Alternatively, the initial signal intensity obtained from the fitting is... and T2* value Independent component analysis (ICA) was performed separately to further separate and identify artifact components.

[0045] Step 3: Calculation and Detection of Abnormal Brain Activity

[0046] The result obtained through step two Time series computation is based on the low-frequency amplitude ALFF (T2*-ALFF) of T2* fluctuations and the local consistency ReHo (T2*-ReHo) index to characterize resting-state local brain function, providing data support for the accurate and reliable localization of abnormal brain activity.

[0047] This method captures the complete signal attenuation curve by acquiring multiple echo points and performing mathematical fitting, directly solving the severe signal attenuation problem caused by the fixed echo time in the single-echo (SE) method. This allows for the recovery of signals from magnetically sensitive and deep brain regions with extremely short T2* values. The average signal retention rate of each brain region was compared between single-echo sequences and T2* wave sequences in 27 subjects. Figure 2 As shown in the figure, the medial orbitofrontal cortex is marked in green, the lateral orbitofrontal cortex in blue, the lower temporal lobe in red, the brainstem in dark green, and the cerebellum in yellow. It can be seen that the signal preservation rate of this method is significantly better than that of the single echo method in multiple regions such as the orbitofrontal cortex, temporal lobe, brainstem and cerebellum. In particular, the signal preservation rate advantage exceeds 40% at a specific apex of the lateral orbitofrontal cortex.

[0048] This method uses a physical model to forcibly separate stable neural signals from unstable noise carriers, significantly improving the repeatability of measurements in brain regions with severe magnetic susceptibility artifacts and deep brain regions, and enhancing the sensitivity of neural activity detection in these areas. Repeated multi-echo and single-echo data were collected from 27 participants. ALFF and ReHo regional brain activity indices were calculated based on single-echo SE, multi-echo ME-OC, and T2* fluctuations, respectively. The test-retest reliability of the three methods was compared, and the results are shown in Table 1.

[0049] Table 1

[0050]

[0051] Figure 3 , 4 This paper compares the average test-retest reliability of the ALFF and ReHo indices in the subcortical region under three time series. The average results show that the method based on the T2* fluctuation time series outperforms the single-echo SE and multi-echo ME-OC time series in overall test-retest reliability. Figure 5 , 6 Radar plots comparing the test-retest reliability of the ALFF and ReHo indices in different subcortical regions under three time series. Both ALFF and ReHo results show that T2* fluctuations have a significant advantage in test-retest reliability in 7 / 10 deep regions, demonstrating that this method exhibits optimal test-retest reliability in both the overall brain and most local regions, confirming its stability in deep brain region measurements.

[0052] like Figure 7 As shown, the device for localizing abnormal brain activity based on T2* time-series resting-state local brain function indices includes:

[0053] The multi-echo data acquisition module uses a multi-echo gradient echo sequence to acquire magnetic resonance data, acquiring N magnetic resonance signals with different echo times after each radio frequency pulse excitation.

[0054] The data quantization and separation module is used to collect a set of multi-echo signal data (TE) after a radio frequency pulse excitation. t1 TE t2 …TE tN A single exponential decay model was applied for fitting to obtain the T2* value at a given time point, which was then concatenated sequentially to form... Time series.

[0055] The resting-state local brain function index characterization module receives the output from the data quantization and separation module. Time series analysis was used to calculate low-frequency amplitude and local consistency indices to characterize resting-state local brain function.

[0056] The abnormal brain activity localization module locates abnormal brain activity by comparing resting-state local brain function indices of different subjects. It directly quantifies the T2* value, the biophysical source, making it more specific for neural activity reflecting changes in blood oxygenation and achieving higher detection sensitivity. It possesses a unique detection capability that specifically detects activity in blind spots where other methods fail, reliably identifying effective neural activities related to pathological states that other techniques cannot detect. Figure 8 As shown, a paired-samples t-test of an eye-opening and closing task revealed that in the right putamen region, which is strongly associated with consciousness arousal at the green circle location, only this application was able to detect a cluster containing 55 active voxels, while neither the SE nor ME-OC methods detected any activity.

Claims

1. A method for depicting a resting-state local brain function index based on a T2* time series, characterized in that: The multi-echo gradient echo sequence is used for magnetic resonance data acquisition, N signals of different echo times are collected after each radio frequency pulse excitation, and a single exponential decay model is used to fit the initial signal intensity and T2* value at each time point and sequentially spliced to obtain time series; For Statistical analysis was performed on the time series to calculate the amplitude of low frequency fluctuation (ALFF) and regional homogeneity (ReHo) to depict the local brain function in resting state and provide data support for precise and reliable positioning of abnormal brain activity.

2. The method for characterizing resting-state local brain function indicators based on T2* time series as described in claim 1, characterized in that: The value of N is adjusted according to the requirements of time resolution and signal-to-noise ratio of actual application scenes.

3. The method for characterizing resting-state local brain function indicators based on T2* time series as described in claim 1, characterized in that: The collected multi-echo data are decomposed and denoised through principal component analysis.

4. The method of claim 3, wherein the method is characterized in that: The signals of 4 different echo times TE are acquired after the tth radio frequency pulse excitation t1 , TE t2 , TE t3 , TE t4 The initial signal intensity at the tth time point is fitted by applying the following mono-exponential decay model and T2* value : ； Wherein, n=1, 2, 3, 4.

5. The method of claim 4, wherein the method is characterized by: The single exponential decay model is fitted by using a weighted nonlinear least square method, a Levenberg-Marquardt algorithm combining gradient descent and Gauss-Newton method, or a Bayesian fitting method based on a physiological prior distribution of T2* values.

6. The method of claim 4, wherein the method is characterized by the following steps: The initial signal intensity and T2* values are independently analyzed by principal component analysis, and are sequentially spliced into time series. 7.A computer readable storage medium having stored thereon a computer program, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-6.

8. An abnormal brain activity positioning device based on T2* time series resting state local brain function index characterization, characterized in that: The abnormal brain activity locating device comprises: A multi-echo data acquisition module acquires magnetic resonance data by using a multi-echo gradient echo sequence, and acquires magnetic resonance signals of N different echo times after each radio frequency pulse excitation; The data quantification and separation module applies a single exponential decay model to fit a set of multi-echo signal data collected after a radio frequency pulse excitation to obtain a T2* value at a time point, and then sequentially splices into time series; The resting state local brain function index description module receives the time series output by the data quantization and separation module calculates the low frequency amplitude and local consistency index, and describes the resting state local brain function; An abnormal brain activity locating module locates abnormal brain activity by comparing resting state local brain function indexes of different subjects.

Images