Based on normal white matter 18 F-florbetapir retention characteristics: an auxiliary assessment system for Alzheimer's disease.

Abstract

This application relates to the field of medical neuroimaging processing and computer-aided diagnosis technology, and discloses a method based on normal white matter. 18 An Alzheimer's Disease (AD) Auxiliary Assessment System Based on F-florbetapir Retention Features. This system includes a data acquisition module for acquiring multimodal brain imaging data of subjects; an image preprocessing and segmentation module for impurity removal and target localization of the image data; a feature extraction module for extracting image features from the processed images; and a joint assessment and prediction module for AD auxiliary assessment based on the extracted features, incorporating a subject performance characteristic model and a logistic regression system. This Alzheimer's Disease (AD) auxiliary assessment system can significantly improve the accuracy of early AD diagnosis, accurately predict the rate of cognitive decline, and dynamically monitor the effect of immunotherapy.

Description

Technical Field

[0001] This application relates to the field of medical neuroimaging processing and computer-aided diagnostic technology, and more specifically, it relates to a method based on normal white matter... 18 An auxiliary assessment system for Alzheimer's disease based on F-florbetapir retention characteristics. Background Technology

[0002] Amyloid (Aβ) deposition is a core pathological feature of Alzheimer's disease (AD). 18 F-florbetapir (FBP) PET imaging is currently the mainstream method for clinical assessment of Aβ deposition in the brain. By detecting the distribution and burden of Aβ in the brain, it provides important evidence for the diagnosis and assessment of AD.

[0003] However, existing Aβ-PET analysis methods have many limitations, making it difficult to meet the clinical needs for early diagnosis, accurate prognosis, and effective treatment monitoring of Alzheimer's disease (AD), including:

[0004] 1. White matter signal is ignored: Existing Aβ-PET analysis methods (such as SUVR calculation) mainly focus on the cortical region. The signal in the white matter region is usually considered as "non-specific binding" noise and is often removed as background in image processing.

[0005] 2. Insufficient sensitivity in early diagnosis: In the early or preclinical stages of AD, cortical Aβ deposition may not have reached the threshold for visual detection or conventional software assessment, leading to missed diagnoses.

[0006] 3. Limited prognostic dimensions: Relying solely on cortical Aβ load cannot effectively distinguish individuals with the same amyloid levels but different rates of cognitive decline.

[0007] 4. Limitations of monitoring indicators: The efficacy assessment of existing anti-Aβ immunotherapy (such as lencanezumab) mainly relies on the cortical clearance rate, and lacks dynamic monitoring methods for pathological changes in the white matter microenvironment.

[0008] Therefore, how to make full use of the ignored white matter signals in PET images to develop a technical solution that can significantly improve the accuracy of early diagnosis of AD, accurately predict the rate of cognitive decline, and dynamically monitor the effect of immunotherapy has become an urgent technical problem to be solved in this field. Summary of the Invention

[0009] This application aims to address how to utilize the neglected white matter signals in PET images to provide a semi-quantitative analysis technology that can significantly improve the accuracy of early AD diagnosis, accurately predict the rate of cognitive decline, and dynamically monitor the effects of immunotherapy.

[0010] To achieve the above-mentioned objective, this invention provides a method based on normal white matter expression. 18 An auxiliary assessment system for Alzheimer's disease based on F-florbetapir retention characteristics. Specific steps and procedures are as follows: Firstly, the white matter based on normal performance provided in this application 18 An auxiliary assessment system for Alzheimer's disease featuring F-florbetapir retention characteristics includes: The data acquisition module is used to acquire multimodal brain imaging data of the subjects; The image preprocessing and segmentation module is used to remove impurities and locate targets in the image data acquired by the data acquisition module; The feature extraction module is used to extract image features from the image after it has been processed by the image preprocessing and segmentation module; The joint assessment and prediction module is used to perform AD-assisted assessment based on the features extracted by the feature extraction module, and has a built-in subject operating characteristic model and logistic regression system.

[0011] Furthermore, the multimodal brain imaging data includes: Intravenous injection 18 PET images acquired after F-florbetapir; and 3T structural MRI data; The MRI data must include T1-weighted images and T2-FLAIR images.

[0012] Furthermore, the impurity removal and target localization include: The T1-weighted image was segmented into three regions: cortex, white matter, and cerebrospinal fluid using a segmentation tool, and then spatially normalized to the MNI standard space. High-signal voxels of white matter were extracted from T2-FLAIR images and removed from the overall white matter mask. Healthy white matter regions were defined and preserved as the white matter mask for normal expression. Spatial registration was performed between PET images and T1-weighted images, and partial volume correction was performed using a region-based voxel-level algorithm. Combined with subject-specific tissue probability maps, the signal spillover effect caused by the limited spatial resolution of PET was mitigated.

[0013] Furthermore, the extracted image features include: The entire cerebellum was selected as the reference region. SUVR of each voxel is calculated based on the registered and corrected image; In the six bilateral regions defined by the Desikan-Killiany atlas, the AβSUVR values ​​of the whole cortex and the FBP retention SUVR values ​​of the normal white matter were extracted, and the regions of interest in the normal white matter and the cortical regions of interest corresponded precisely vertically in terms of anatomy.

[0014] Furthermore, the subject operating characteristic model determines the classification threshold based on the Youden index. The classification threshold is a global cortical threshold of 1.027, a global normal white matter threshold of 1.342, and a combined threshold of the two of them of 2.368. The combined characteristic value of the subject is compared with the combined threshold. If it is higher than the threshold, it is output as a high risk of amyloid positivity. The logistic regression system stratifies subjects based on baseline normal white matter SUVR values. Under the same cortical Aβ load level, if the normal white matter FBP retention level is higher than the preset normal value, the output is a high risk of rapid cognitive decline. The system receives baseline and follow-up PET data from patients before treatment and calculates the ΔSUVR of the whole brain and each region. If the ΔSUVR of the normal white matter shows a significant negative value, the output indicates a positive response to the treatment. The determination of a significant negative ΔSUVR value is based on statistical analysis, and two-sided p < 0.05 is considered statistically significant.

[0015] Furthermore, the image preprocessing and segmentation module, The segmentation tool is SPM12; The DARTEL algorithm was used to normalize MRI images to the MNI standard space; High-signal voxels of white matter were extracted from T2-FLAIR images using a lesion segmentation toolkit.

[0016] Secondly, this application provides the application of the Alzheimer's disease auxiliary assessment system in the preparation of Alzheimer's disease auxiliary assessment products.

[0017] Thirdly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements all the functions of the Alzheimer's disease auxiliary assessment system.

[0018] Fourthly, this application provides an Alzheimer's disease auxiliary assessment device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements all the functions of the Alzheimer's disease auxiliary assessment system.

[0019] This invention overcomes the bias in the prior art that white matter signals are regarded as useless background, and brings the following significant technical effects: Significantly improved early diagnostic accuracy: Traditional single-layer cortical assessment has limitations in accuracy. This system, by jointly assessing PET features of the cortex and normal-looking white matter, significantly improved the diagnostic accuracy of amyloid-positive classification to 92% in real-world validation.

[0020] Achieving earlier pathological tracking: FBP retention in normal white matter exhibits a unique trajectory, showing a significant increase even in the preclinical stage (asymptomatic period). This system can capture early white matter abnormalities that occur before the formation of extensive cortical plaques, superior to using cortical indicators alone.

[0021] Provides independent prognostic predictive power: Even in patients with the same overall cortical Aβ burden, the baseline normal white matter FBP retention extracted by this system can still serve as a powerful independent predictor of clinical progression and cognitive decline, with predictive power comparable to expensive or invasive liquid biomarkers (such as plasma p-tau217 and cerebrospinal fluid biomarkers).

[0022] This system provides a sensitive means of monitoring treatment: it confirms that the retention of FBP in normal white matter is not only a static phenomenon but also a dynamic signal related to pathology. It can be used to directly quantify the clearance effect of anti-Aβ drugs (such as lencanezumab) in white matter regions, providing a novel neuroimaging endpoint for new drug development and clinical efficacy monitoring. Attached Figure Description

[0023] Figure 1 The figure illustrates the difference between the semi-quantitative values ​​of NAWM and CTX FBP in the Alzheimer's disease continuum.

[0024] Figure 2 The figure illustrates the longitudinal changes in semi-quantitative values ​​of NAWM and CTX FBP in the Alzheimer's disease continuum.

[0025] Figure 3 The illustration shows that lencanezumab treatment reduces CTX Aβ burden and NAWM FBP retention in different brain regions.

[0026] Figure 4 The figure illustrates the ability of CTX Aβ semi-quantitative values ​​and NAWM FBP retention to predict clinical progression.

[0027] Figure 5 The illustrations demonstrate the diagnostic accuracy and mechanistic understanding of the combined assessment of CTX Aβ SUVR and NAWM FBP retention.

[0028] Figure 6 The diagram illustrates the diagnostic efficacy and validation results of the biomarkers. Detailed Implementation

[0029] This application provides a method based on normally expressed white matter. 18An Alzheimer's disease auxiliary assessment system based on F-florbetapir retention characteristics mainly includes: a data acquisition module, an image preprocessing and segmentation module, a feature extraction module, and a joint assessment and prediction module. The specific steps and scheme are as follows: 1. Data Acquisition Module: Acquire multimodal brain imaging data of the subjects, including PET images acquired after intravenous FBP injection and 3T structural magnetic resonance imaging (MRI) data, wherein the MRI data must include T1-weighted images and T2-FLAIR images.

[0030] 2. Image preprocessing and segmentation module (imperfection removal and target localization): Tissue segmentation: The T1-weighted image was segmented into three regions—cortex, white matter, and cerebrospinal fluid—using a segmentation tool (such as SPM12), and then spatially normalized to the MNI standard space.

[0031] Removing white matter high signal (WMH) interference: This is the key to extracting high-purity white matter features in this scheme. The system uses the lesion segmentation toolbox to extract "white matter high signal" voxels from T2-FLAIR images.

[0032] Define Normal White Matter (NAWM): In subsequent analysis, the system forcibly removes voxels containing high white matter signals from the overall white matter mask, retaining only healthy white matter regions as the "Normal White Matter (NAWM)" mask.

[0033] PET Image Registration and Partial Volume Correction (PVC): Spatial registration of PET images with T1-weighted images is performed. Partial volume correction is performed using a region-based voxel-level algorithm (RBV), combined with subject-specific tissue probability maps to mitigate signal spillover effects caused by the limited spatial resolution of PET.

[0034] 3. Feature Extraction Module: Reference region selection: The whole cerebellum is used as the reference region.

[0035] SUVR calculation: Based on the registered and corrected image, the standardized uptake ratio (SUVR) of each voxel is calculated.

[0036] Region of Interest (ROI) feature extraction: Aβ SUVR values ​​of the whole cerebral cortex (CTX) and FBP retention SUVR values ​​of normal white matter (NAWM) were extracted from six bilateral regions (frontal, parietal, temporal, occipital, cingulate, and limbic lobes) defined by the Desikan-Killiany atlas. The NAWM and NAWM regions of interest in the normal white matter correspond precisely anatomically.

[0037] 4. Joint Assessment and Prediction Module (Core Decision-Making Logic): The system inputs the extracted CTX SUVR features and NAWM SUVR features into a preset evaluation model for calculation: Diagnostic determination logic: The subject operating characteristic (ROC) model has built-in classification thresholds based on the Youden index: the overall cortical threshold is 1.027, the overall normal white matter threshold is 1.342, and the combined threshold is 2.368.

[0038] The system compares the combined feature values ​​of the subjects with a combined threshold; if the values ​​are higher than the threshold, the subjects are identified as amyloid-positive (Aβ+). This combined assessment mode significantly reduces the misdiagnosis rate for patients in the diagnostic gray area.

[0039] The prognostic assessment logic for cognitive decline is as follows: The logistic regression system stratifies subjects based on their baseline NAWM SUVR values. If, at the same cortical Aβ load level, a subject's NAWM FBP retention level is higher than the preset normal value, the system outputs a warning of "high risk of rapid cognitive decline".

[0040] Logic for monitoring the efficacy of anti-Aβ immunotherapy (lencanemab): We collected baseline and follow-up PET data from patients before treatment and calculated ΔSUVR (follow-up SUVR minus baseline SUVR) for the whole brain and each region.

[0041] If the NAWM ΔSUVR shows a significant negative value (decrease), the system determines that there is a positive drug response to the treatment.

[0042] More specifically: Subject data acquisition Medical imaging data, clinical data, and biomarker data were acquired from multiple subjects, including: The subjects were drawn from multiple independent data sources, including: the primary data source: image data from publicly available databases; Second data source: Imaging data collected at the clinical center; Third data source: Follow-up data of patients receiving anti-Aβ therapy; The participants included individuals with normal cognition, patients with mild cognitive impairment, and patients with Alzheimer's disease.

[0043] Furthermore, cognitive assessment data of the subjects were obtained, including: Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA). All participants signed informed consent forms, and the process was approved by the ethics committee.

[0044] Image acquisition and preprocessing Acquire PET and MRI images of the subjects, including: The PET image acquisition process includes: injecting the subject with the Aβ tracer FBP at a dose of 370 MBq ± 10%, and performing a static scan 50 minutes after injection, with the scan lasting for 20 minutes; MRI scans included: T1-weighted structural images and T2-FLAIR sequences; MRI image processing includes: using SPM software for tissue segmentation, dividing brain tissue into: gray matter, white matter, and cerebrospinal fluid; The images were normalized to the MNI space using the DARTEL algorithm; Furthermore, high-signal regions of white matter were segmented from T2-FLAIR images, and corresponding voxels were excluded in subsequent analysis; PET image processing includes: spatial registration with T1-weighted MRI images, partial volume effect correction, spatial standardization, and Gaussian smoothing. Partial volume effect correction employs a region-based voxel-level algorithm (RBV algorithm).

[0045] Region of Interest Determined Multiple regions of interest were identified based on standard brain atlases, including the whole cerebellum region as a reference region for calculating the standard uptake ratio. Target areas include: Cortical regions (CTX): frontal cortex, parietal cortex, temporal cortex, occipital cortex, cingulate gyrus, and limbic lobe; Normal white matter region (NAWM): The white matter region obtained after excluding high-signal voxels in white matter.

[0046] SUVR calculation Calculate the standard uptake ratio of the target region based on the reference region: SUVR = SUV Target / SUV reference , including: SUV target For the target area average SUV, SUV reference The average SUV for the entire cerebellar region; Calculate separately: cortical SUVR, and white matter SUVR with normal presentation.

[0047] Biomarker detection Obtain cerebrospinal fluid and plasma biomarker data from the subjects, including: Cerebrospinal fluid indicators: Aβ42, p-tau, t-tau; Plasma markers: Aβ42, Aβ40, p-tau181, p-tau217, GFAP, NfL; Quantification was performed using electrochemiluminescence immunoassay or single-molecule array detection technology.

[0048] Statistical modeling Statistical models were used to assess the relationship between imaging indicators, biomarkers, and cognitive function. Between-group difference analysis includes: analysis of variance (ANOVA) and paired t-test; Correlation analysis includes Spearman correlation analysis.

[0049] Mediation analysis was used to assess the role of biomarkers in the relationship between imaging indicators and cognitive function. Prognostic analysis was performed using the Cox proportional hazards model to calculate the hazard ratio; Survival analysis was performed using the Kaplan-Meier method.

[0050] Example This application provides a general and / or specific description of the materials and test methods used in the experiments. Unless otherwise specified, all raw materials or instruments used are commercially available and readily available.

[0051] Example 1: Brain Image Sample Acquisition and Statistical Analysis The study included 672 participants from three independent data sources: 546 from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database and 126 from two Chinese clinical centers (Huashan Hospital affiliated with Fudan University in Shanghai; and the First Hospital of Jilin University in Changchun). All participants underwent FBP PET, T1-weighted, and T2-weighted magnetic resonance imaging.

[0052] In the Huashan cohort, 12 patients with cognitive impairment received lencanemab treatment (lencanemab treatment group) and underwent baseline and follow-up FBP PET scans at a median treatment duration of one year. As a control, 18 participants with cognitive impairment from the same center who did not receive lencanemab treatment were included in the control group and also underwent FBP PET imaging at baseline and one year follow-up. Patients in the lencanemab treatment group received lencanemab (LEQEMBI). ® This is a humanized IgG1 monoclonal antibody that targets aggregated soluble fibrils Aβ.

[0053] Of the ADNI participants, 205 had longitudinal imaging data (median follow-up: 5 years). Based on amyloid status (SUVR ≥ 1.11, whole cerebellum reference) and clinical diagnosis, ADNI participants were divided into four groups: amyloid-negative cognitively normal subjects (CN...). The study included 126 participants: 112 from Huashan Hospital (Wuzhong East Road Campus); 10 from CN (n=106); 183 from amyloid-positive cognitively normal subjects; 161 from amyloid-positive patients with mild cognitive impairment (MCI+); and 96 from amyloid-positive Alzheimer's disease (AD+). The Chinese cohort included 126 participants: 112 from Huashan Hospital (Wuzhong East Road Campus); and 10 from CN (n=106). Of the 7 patients, 43 were diagnosed with COVID-19 (CN+), 52 with diabetes (AD+), 7 were from Huashan Hospital (Hongqiao Campus; 2 were diagnosed with COVID-19 (MCI+), 5 were diagnosed with AD+), and 7 were from the First Hospital of Jilin University (3 were diagnosed with COVID-19 (MCI+), 4 were diagnosed with AD+).

[0054] The differences in SUVR values ​​between groups for CTX and NAWM were assessed using analysis of variance (ANOVA) with Bonferroni correction, and longitudinal variation was assessed using paired t-tests. The diagnostic performance of cortical, normal white matter, and combined SUVR values ​​was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC), with thresholds derived from ADNI and validated in a Chinese cohort. The associations between imaging biomarkers, cognitive function, and liquid biomarkers were examined using Spearman correlation analysis, and mediation analysis was used to test whether liquid biomarkers explained the relationship between SUVR and cognition (5000-time bootsampling, adjusted for age and sex).

[0055] Prognostic value was assessed using a Cox proportional hazards regression model, with cognitive decline defined as an MMSE or MoCA score below a threshold (≤26), and risk groups stratified according to predicted conversion rates. Kaplan-Meier survival curves and log-rank tests were used to compare conversion risks, and hazard ratios were used to quantify the effects of biomarkers. SUVR and cognitive trajectories were modeled using nonlinear regression. All analyses were performed in R software (v4.1.0), and a two-sided p < 0.05 was considered statistically significant.

[0056] Table 1. Patient demographic information and clinical characteristics

[0057] Continuous variables are expressed as mean (standard deviation) or median (range). Categorical variables are expressed as count (percentage). Significance markers are expressed as p-values ​​for comparison with the corresponding group: a, C, ... Group A, ADNI Queue vs. China Queue; Group B, CN+ Queue vs. China Queue; Group C, MCI+ Queue vs. China Queue; Group D, AD+ Queue vs. China Queue; Group E, ADNI Queue vs. China Queue; Group F, ADNI Queue vs. China Queue; Group A, ADNI Queue / China Queue (CN- vs. CN+); Group B, ADNI Queue / China Queue (CN- vs. MCI+); Group C, ADNI Queue / China Queue (CN- vs. AD+); Group D, ADNI Queue / China Queue (CN+ vs. MCI+); Group E, ADNI Queue / China Queue (CN+ vs. AD+); Group F, ADNI Queue / China Queue (MCI+ vs. AD+). p<0.05; p<0.01; p<0.001; p<0.0001.

[0058] Abbreviations: AD, Alzheimer's disease; AD+, Aβ-positive Alzheimer's disease; ADNI, Alzheimer's Neuroimaging Initiative; A-, Aβ-negative; A+, Aβ-positive; CN, cognitively normal; CN-, Aβ-negative with normal cognition; CN+, Aβ-positive with normal cognition; MCI+, Aβ-positive with mild cognitive impairment; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment.

[0059] Example 2: Difference between semi-quantitative values ​​of NAWM and CTX FBP in the Alzheimer's disease continuum Two representative CN+ cases with the same cortical Aβ SUVR (1.14) showed different outcomes: the case with normal-presenting, high white matter retention (ΔSUVR = 0.14) exhibited more pronounced cognitive decline over 24 months (MMSE / MoCA Δ = -2), while the other case remained stable (Δ = 0). Figure 1 A, B).

[0060] Throughout the AD continuous spectrum (CN) In the CN+, MCI+, and AD+ groups, there were significant differences in cortical Aβ SUVR and FBP retention in normal white matter (all p < 0.0001). However, there was a significant difference in cortical SUVR between the CN+ and MCI+ groups (p < 0.001), while there was no significant difference in normal white matter. Figure 1C, D). Regional analysis revealed distinct patterns: cortical Aβ SUVR progressively increased in the frontal, parietal, temporal, and cingulate gyrus regions (all p < 0.001), while showing less variation in the occipital and limbic lobes. Normal white matter retention was observed from CN. In the MCI+ stage, the levels are elevated in the frontal, parietal, temporal, and occipital lobes, but no further increase is observed in the normal white matter of the temporal and occipital lobes between CN+ and MCI+.

[0061] Example 3: Longitudinal variations of semi-quantitative values ​​of NAWM and CTX FBP in the Alzheimer's disease continuum. To characterize the temporal evolution of Aβ accumulation, a subset of the entire AD lineage (CN) was used. Four cases (CN+, MCI+, AD+) underwent FBP PET analysis. Cortical Aβ SUVR in the cases (baseline: 0.98; follow-up: 0.99) and FBP retention in normal white matter (baseline: 0.92; follow-up: 0.97) both showed negligible changes during longitudinal follow-up. Figure 2 A1, A2). Conversely, CN+ cases showed significant progression during the observation period in cortical Aβ SUVR (from 1.21 to 1.28) and normal-presenting white matter FBP retention (from 1.55 to 1.60). Figure 2 B1, B2). MCI+ cases also showed a progressive increase in cortical Aβ SUVR (from 1.21 to 1.34) and normal-looking white matter FBP retention (from 1.46 to 1.60). Figure 3 C). Despite elevated baseline values, AD+ cases still showed persistent amyloid accumulation in the cortex (from 1.30 to 1.37) and normal-looking white matter areas (from 1.43 to 1.54). Figure 2 D1, D2).

[0062] Longitudinal between-group analysis showed that among cognitively normal participants (CN... In the CTX group (n = 50, and CN+: n = 84), from baseline to follow-up, cortical Aβ SUVR and normal white matter FBP retention were significantly increased (CTX: p < 0.001; normal white matter: p < 0.01). Figure 2 E). In participants with cognitive impairment (MCI+: n = 53, and AD+: n = 18), comparable longitudinal progression was observed in cortical Aβ SUVR (p < 0.001) and normal-performing white matter FBP retention (p < 0.01). Figure 2 F).

[0063] To examine the association between overall Aβ load and cognitive function, nonlinear regression curves were fitted across the entire AD spectrum. Cortical Aβ SUVR exhibited a triphasic pattern: in CN... Individuals initially stabilize (MMSE > 27), then progressively accumulate (MMSE ≤ 27), eventually reaching a plateau (MMSE ≤ 25), which is consistent with the saturation phenomenon. Figure 2 G; r = 0.504, p < 0.0001). Similar trajectories were also observed in normal white matter FBP retention, characterized by trajectories in the CN. The participants were stable (MMSE>28), progressively accumulating (MMSE≤28), and reaching a plateau when cognitive impairment was severe (MMSE≤26). Figure 2 H; r = 0.367, p < 0.0001).

[0064] Example 4: Lencanemab treatment reduces CTX Aβ burden and positive NAWM FBP retention in different brain regions. Participants underwent a series of Aβ PET imaging studies before and after intravenous lencanemab administration. In a representative case treated with lencanemab, the global cortical Aβ SUVR decreased from 1.199 to 1.105, and the normal-looking white matter FBP retention decreased from 1.391 to 1.278. Figure 3 A1, A2). ΔSUVR values ​​(SUVR follow-up) of CTX and normal white matter in the frontal, parietal, temporal, occipital, cingulate gyrus, and limbic lobes. (SUVR baseline) See Figure 3 A3.

[0065] At baseline, there were no significant differences in SUVR between the lencanezumab treatment group and the control group, either in the cortex or in normal white matter (p>0.05). Compared with the control group, the lencanezumab treatment group showed significantly greater reductions in CTX ΔSUVR in the whole body (p<0.0001), frontal lobe (p=0.0001), parietal lobe (p<0.0001), temporal lobe (p=0.0002), occipital lobe (p=0.0003), cingulate gyrus (p<0.0001), and limbic system (p<0.0001). Figure 4 B). Similarly, the decrease in ΔSUVR in normal white matter was significantly greater in the overall region (p = 0.0484), frontal lobe (p = 0.0357), parietal lobe (p = 0.0116), temporal lobe (p = 0.0492), and occipital lobe region (p = 0.0250). Figure 3 B). The mean ΔSUVR values ​​of the two groups of cortical and normal white matter are summarized in Figure 3 C.

[0066] Within the lencanezumab treatment group, from baseline to follow-up, significantly reduced cortical Aβ SUVR was observed in the whole body (p = 0.0015), frontal lobe (p = 0.0024), parietal lobe (p = 0.0015), temporal lobe (p = 0.0015), occipital cortex (p = 0.0034), cingulate cortex (p = 0.0010), and limbic system (p = 0.0005). Figure 3 D). Normal white matter FBP retention was also significantly reduced in the overall lobe (p = 0.0425), frontal lobe (p = 0.0425), parietal lobe (p = 0.0269), temporal lobe (p = 0.0425), and limbic system (p = 0.0161). Figure 3 E). In contrast, no significant change in SUVR was detected between baseline and follow-up scans in the lencanezumab control group (p>0.05).

[0067] Example 5: The ability of CTX Aβ semi-quantitative values ​​and NAWM FBP retention to predict clinical progression To identify predictive biomarkers for clinical progression, a Kaplan-Meier survival analysis was performed, and participants were divided into high-risk and low-risk groups based on AβPET imaging parameters and liquid biomarkers. Figure 4 (A~D), and then the log-rank test was used in the ADNI cohort to estimate the hazard ratio (log-rank test, Figure 4 E).

[0068] Elevated overall cortical Aβ SUVR and FBP retention in normal-functioning white matter showed a robust association with accelerated progression (normal-functioning white matter: p < 0.0001). Figure 4 A; CTX: p < 0.0001, Figure 4 B). Among plasma biomarkers, elevated phosphorylated tau-217 concentration was a significant predictor of clinical deterioration over time (p = 0.0002). Figure 4 C). Furthermore, an elevated p-tau / Aβ42 ratio in cerebrospinal fluid was significantly associated with accelerated disease transformation (p = 0.0016, Figure 4 D).

[0069] These findings were confirmed by risk ratio analysis. Figure 4 E), this analysis confirmed that cortical Aβ SUVR and normal white matter FBP retention, cerebrospinal fluid biomarkers, plasma p-tau217, plasma Aβ42 / 40 ratio, age, and APOE ε4 were robust predictors of clinical progression (all p < 0.05), while plasma NfL, GFAP, Aβ42, Aβ40, and p-tau181 lacked significant predictive value (all p > 0.05).

[0070] Example 6: Joint evaluation of the diagnostic accuracy and mechanistic understanding of CTX Aβ SUVR and NAWM FBP retention Mediation analysis, using global cortical Aβ SUVR as the independent variable, showed that plasma p-tau217 significantly mediated the association between cortical amyloid load and MMSE in the ADNI cohort (standardized indirect effect β = -1.8221, 95% confidence interval: -4.0639 to -0.1991). Figure 5 A1). Similarly, for cerebrospinal fluid Aβ42 (β = -1.0328, 95% CI: -1.9248 to -0.2949; Figure 5 A2) and the ratio of p-tau / Aβ42 in cerebrospinal fluid (β = -1.7903, 95% CI: -3.4164 to -0.2594; Figure 5 A3) also showed a significant mediating effect. When MoCA was used as the outcome measure, the cerebrospinal fluid p-tau / Aβ42 ratio remained a significant mediator (β = -2.3569, 95% CI: -4.4905 to -0.4014); Figure 5 A4).

[0071] A similar pattern emerged when normal white matter FBP retention was used as the independent variable. Plasma p-tau217 (β = -0.7206, 95% CI: -1.6521 to -0.0637); Figure 5 B1), cerebrospinal fluid Aβ42 (β = -0.3725, 95% CI: -0.8637 to -0.0283; Figure 5 B2) and the ratio of p-tau / Aβ42 in cerebrospinal fluid (β = -0.8547, 95% CI: -1.7116 to -0.2005; Figure 5 B3) significantly mediated its association with MMSE. For MoCA, the cerebrospinal fluid p-tau / Aβ42 ratio also showed a significant mediating effect (β = -1.1394, 95% CI: -2.3467 to -0.2604; Figure 5 B4).

[0072] In all participant groups (CN) In the MMSE (MoCA) scores (CN+, MCI+), cognitive performance showed a robust correlation with biomarker concentrations. Elevated plasma p-tau217 levels were associated with decreased MMSE scores (r = -0.3142, p = 0.0108). Similar associations were observed for cerebrospinal fluid (CSF) Aβ42 (r = 0.3164, p = 0.0005) and the CSF p-tau / Aβ42 ratio (r = -0.3587, p < 0.0001). Furthermore, the CSF p-tau / Aβ42 ratio was significantly negatively correlated with MoCA scores (r = -0.2651, p = 0.0039).

[0073] In the ADNI cohort, both global cortical Aβ SUVR and normal-performing white matter FBP retention showed differentiation in CN. Compared to the superior abilities of CN+ individuals, the areas under the receiver operating characteristic (AUC) were 0.975 ± 0.007 (p = 0.03) and 0.976 ± 0.007 (p = 0.89), respectively, while their combination produced better classification performance (AUC = 0.988 ± 0.004, p = 0.001). Figure 6 A). For distinguishing between cognitively unimpaired (CU) and cognitively impaired (CI) individuals, the AUC of cortical SUVR was 0.803 ± 0.019 (p = 0.02), the normal white matter FBP retention was 0.757 ± 0.020 (p < 0.001), and their combined AUC was 0.791 ± 0.019 (p < 0.0001). Figure 6 B). The Youden index was used to determine the differentiation between Aβ-negative (A... , including CN The optimal classification thresholds were determined for participants in the group (Aβ positive, including CN+, MCI+, and AD+) and Aβ positive (A+, including CN+, MCI+, and AD+). The optimal cutoff value for overall cortical Aβ SUVR was 1.027, for overall normal white matter FBP retention it was 1.342, and for their combination it was 2.368.

[0074] Notably, we observed center-specific differences in diagnostic accuracy when performing visual assessments of the overall cortex and normal-functioning white matter. In center 1 of the Huashan cohort, the accuracy of cortical assessment was 82.1%, normal-functioning white matter was 68.8%, and the combined accuracy was 92.0%. In center 2, the individual accuracies of cortical and normal-functioning white matter assessments were 57.1% and 71.4%, respectively, while their combination improved the accuracy to 85.7%. A similar improvement was observed in the Jilin University First Hospital cohort (combined accuracy: 85.7%). Figure 6 C, D).

[0075] Example 7: Cortical Aβ SUVR and normal-behavioral white matter FBP retention are associated with cerebrospinal fluid and plasma biomarkers in preclinical Alzheimer's disease. For liquid biomarker analysis, inclusion was limited to participants in the ADNI cohort who had both baseline and follow-up data, and correlation analysis used baseline liquid biomarker concentrations. In the CN+ group, 59 participants obtained cerebrospinal fluid biomarkers (Aβ42, p-tau, and t-tau), and 34 participants obtained plasma biomarkers (Aβ42, Aβ40, p-tau181, p-tau217, GFAP, and NfL). In the MCI+ group, 38 participants obtained cerebrospinal fluid biomarkers, and 17 participants obtained plasma biomarkers.

[0076] In the CN+ group, overall normal white matter FBP retention was positively correlated with plasma p-tau217 (r = 0.4740, p = 0.0046), negatively correlated with cerebrospinal fluid Aβ42 (r = -0.3551, p = 0.0058), and positively correlated with the cerebrospinal fluid p-tau / Aβ42 ratio (r = 0.5576, p < 0.0001). Significant associations were also observed with plasma GFAP (r = 0.4418, p = 0.0089), cerebrospinal fluid p-tau (r = 0.3873, p = 0.0024), and plasma p-tau181 (r = 0.3599, p = 0.0366), while the associations with plasma Aβ40 and Aβ42 / 40 did not reach statistical significance. In the MCI+ group, overall normal white matter FBP retention showed significant associations with plasma p-tau217 (r = 0.5696, p = 0.0213), cerebrospinal fluid Aβ42 (r = -0.3493, p = 0.0316), and plasma Aβ42 / 40 (r = -0.5124, p = 0.0355), although the associations with other markers, including cerebrospinal fluid p-tau and GFAP, did not reach significance. In the pooled analysis of CN+ and MCI+ participants, significant associations were found in plasma p-tau217 (r = 0.5035, p = 0.0002), cerebrospinal fluid Aβ42 (r = -0.3453, p = 0.0005), cerebrospinal fluid p-tau / Aβ42 (r = 0.3115, p = 0.0019), plasma GFAP (r = 0.3344, p = 0.0165), plasma Aβ40 (r = 0.3236, p = 0.0206), plasma Aβ42 / 40 (r = -0.3423, p = 0.0139), cerebrospinal fluid p-tau (r = 0.2627, p = 0.0125), and plasma p-tau181 (r = 0.3052, p = 0.0294).

[0077] In the CN+ group, global cortical Aβ SUVR showed robust associations with plasma p-tau217 (r = 0.6826, p<0.0001), cerebrospinal fluid Aβ42 (r = -0.5113, p<0.0001), and cerebrospinal fluid p-tau / Aβ42 (r = 0.6791, p<0.0001). Significant associations were also observed with plasma GFAP (r = 0.5529, p = 0.0007), NfL (r = 0.4260, p = 0.0120), p-tau181 (r = 0.4899, p = 0.0033), cerebrospinal fluid p-tau (r = 0.3700, p = 0.0039), cerebrospinal fluid t-tau (r = 0.2641, p = 0.0433), and plasma Aβ42 / 40 (r = -0.3977, p = 0.0198), as well as a trending towards significant association with plasma Aβ40 (r = 0.3258, p = 0.0600). In the MCI+ group, global cortical Aβ SUVR showed significant associations with plasma p-tau217 (r = 0.7531, p = 0.0008), cerebrospinal fluid Aβ42 (r = -0.5324, p = 0.0006), cerebrospinal fluid p-tau (r = 0.3711, p = 0.0218), and cerebrospinal fluid p-tau / Aβ42 (r = 0.4895, p = 0.0018), while the associations with plasma Aβ40, Aβ42 / 40, GFAP, NfL, p-tau181, and cerebrospinal fluid t-tau did not reach significance. In the pooled cohort, strong associations persisted across multiple biomarkers, including plasma p-tau217 (r = 0.7008, p < 0.0001), cerebrospinal fluid Aβ42 (r = -0.5187, p < 0.0001), cerebrospinal fluid p-tau / Aβ42 (r = 0.5056, p < 0.0001), plasma GFAP (r = 0.4662, p = 0.0006), NfL (r = 0.2925, p = 0.0373), p-tau181 (r = 0.4401, p = 0.0012), cerebrospinal fluid p-tau (r = 0.3460, p = 0.0005), cerebrospinal fluid t-tau (r = 0.2554, p = 0.0116), and plasma Aβ40 (r = 0.3332, p = 0.0169) and plasma Aβ42 / 40 (r = -0.4212, p = 0.0021).

[0078] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

Claims

1. A method based on normal white matter expression 18 An Alzheimer's disease auxiliary assessment system based on F-florbetapir retention characteristics, characterized by... include: The data acquisition module is used to acquire multimodal brain imaging data of the subjects; The image preprocessing and segmentation module is used to remove impurities and locate targets in the image data acquired by the data acquisition module; The feature extraction module is used to extract image features from the image after it has been processed by the image preprocessing and segmentation module; The joint assessment and prediction module is used to perform AD-assisted assessment based on the features extracted by the feature extraction module, and has a built-in subject operating characteristic model and logistic regression system.

2. The Alzheimer's disease auxiliary assessment system according to claim 1, characterized in that, The multimodal brain imaging data includes: Intravenous injection 18 PET images acquired after F-florbetapir; and 3T structural MRI data; The MRI data must include T1-weighted images and T2-FLAIR images.

3. The Alzheimer's disease auxiliary assessment system according to claim 2, characterized in that, The impurity removal and target localization include: The T1-weighted image was segmented into three regions: cortex, white matter, and cerebrospinal fluid using a segmentation tool, and then spatially normalized to the MNI standard space. High-signal voxels of white matter were extracted from T2-FLAIR images and removed from the overall white matter mask. Healthy white matter regions were defined and preserved as the white matter mask for normal expression. Spatial registration was performed between PET images and T1-weighted images, and partial volume correction was performed using a region-based voxel-level algorithm. Combined with subject-specific tissue probability maps, the signal spillover effect caused by the limited spatial resolution of PET was mitigated.

4. The Alzheimer's disease auxiliary assessment system according to claim 3, characterized in that, The extracted image features include: The entire cerebellum was selected as the reference region. SUVR of each voxel is calculated based on the registered and corrected image; In the six bilateral regions defined by the Desikan-Killiany atlas, the Aβ SUVR values ​​of the whole cortex and the FBP retention SUVR values ​​of the normal white matter were extracted, and the regions of interest in the normal white matter and the cortical regions of interest corresponded precisely vertically in terms of anatomy.

5. The Alzheimer's disease auxiliary assessment system according to claim 4, characterized in that, The subject operating characteristic model determines the classification threshold based on the Youden index. The classification threshold is 1.027 for the overall cortex, 1.342 for the overall normal white matter, and 2.368 for the combination of the two. The subject's combined feature value is compared with the combined threshold. If it is higher than the threshold, it is output as a high risk of amyloid positivity. The logistic regression system stratifies subjects based on baseline normal white matter SUVR values. Under the same cortical Aβ load level, if the normal white matter FBP retention level is higher than the preset normal value, the output is a high risk of rapid cognitive decline. The system receives baseline and follow-up PET data from patients before treatment and calculates the ΔSUVR of the whole brain and each region. If the ΔSUVR of the normal white matter shows a significant negative value, the output indicates a positive response to the treatment. The determination of a significant negative ΔSUVR value is based on statistical analysis, and two-sided p < 0.05 is considered statistically significant.

6. The Alzheimer's disease auxiliary assessment system according to claim 3, characterized in that, The image preprocessing and segmentation module The segmentation tool is SPM12; The DARTEL algorithm was used to normalize MRI images to the MNI standard space; High-signal voxels of white matter were extracted from T2-FLAIR images using a lesion segmentation toolkit.

7. The use of the Alzheimer's disease auxiliary assessment system according to any one of claims 1-6 in the preparation of Alzheimer's disease auxiliary assessment products.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements all the functions of the Alzheimer's disease auxiliary assessment system according to any one of claims 1-6.

9. An Alzheimer's disease auxiliary assessment device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements all the functions of the Alzheimer's disease auxiliary assessment system according to any one of claims 1-6.

Images