A method and system for assessing the risk of progression of a neurodegenerative disease
By combining multimodal brain imaging and sleep EEG data, a risk prediction model was constructed, which addresses the shortcomings of existing technologies in assessing the risk of early progression of Alzheimer's disease, and enables early and accurate risk assessment and personalized intervention support.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XUANWU HOSPITAL OF CAPITAL UNIV OF MEDICAL SCI
- Filing Date
- 2026-01-06
- Publication Date
- 2026-06-05
Smart Images
Figure CN122158103A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for assessing the risk of progression of neurodegenerative diseases, and also to a corresponding system, belonging to the field of medical auxiliary diagnostic technology. Background Technology
[0002] Alzheimer's disease is a neurodegenerative disease with insidious onset and slow progression, and is the most common cause of dementia in the elderly. Its typical pathological features include amyloid-β (Aβ) plaque deposition in the brain and neurofibrillary tangles composed of hyperphosphorylated tau protein. Numerous studies have shown that abnormal production and impaired clearance of pathological proteins in the brain jointly participate in the development and progression of Alzheimer's disease, with impaired brain waste clearance function considered one of the important mechanisms promoting Aβ and tau accumulation and accelerating cognitive decline.
[0003] Assessing the early progression risk of neurodegenerative diseases such as Alzheimer's disease presents significant challenges. Current technologies suffer from three main deficiencies: (1) Limitations of a single dimension: Using only brain lymphatic system imaging indicators (such as PVS load, DTI-ALPS index, choroid plexus volume, BOLD-CSF coupling) or only sleep EEG oscillation coupling indicators (such as SO-spindle coupling) for assessment cannot reveal the key pathophysiological chain of "sleep oscillation regulating brain clearance function", resulting in weak explanatory power of the mechanism and insufficient sensitivity in early identification.
[0004] (2) Simple superposition of multimodal features: Although some studies have attempted to integrate multiple biomarkers, they have failed to model “brain lymph-sleep coupling” as a predefined functional module with intrinsic physiological connections. The feature combination lacks theoretical guidance and the model has poor interpretability.
[0005] (3) Lack of standardized translation pathways: Existing research processes are mostly exploratory, lacking unified standards in data collection and indicator calculation, making it difficult to translate them into assessment tools that can be stably applied in routine clinical work.
[0006] Therefore, there is an urgent need for a technical solution that can jointly quantify the coupling between brain lymphatic function and sleep oscillations, and build an interpretable predictive model based on their intrinsic relationship, in order to achieve earlier and more accurate risk assessment of disease progression. Summary of the Invention
[0007] The primary technical problem to be solved by this invention is to provide a method for assessing the risk of progression of neurodegenerative diseases.
[0008] Another technical problem to be solved by the present invention is to provide a system for assessing the risk of progression of neurodegenerative diseases.
[0009] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a method for assessing the risk of progression of neurodegenerative diseases is provided, comprising the following steps: S1: Acquire multimodal brain imaging data and overnight sleep EEG data from the same subject; S2: Based on the multimodal brain imaging data, calculate at least one brain lymphatic index that reflects the clearance function of the brain lymphatic system; S3: Based on the overnight sleep EEG data, calculate at least one sleep coupling index that reflects the coupling characteristics between slow wave oscillations during non-rapid eye movement sleep and Theta rhythm or sleep spindle waves. S4: Input the calculated brain lymphatic indicators and sleep coupling indicators into the pre-trained risk prediction model to obtain the risk assessment results of the subject; The pre-trained risk prediction model is based on historical population data. Through mediation analysis, brain lymphatic indicators and sleep coupling indicators that have a significant mediating association with cognitive decline are selected as association features, and the model is trained using these association features.
[0010] Preferably, the pre-trained risk prediction model is constructed in the following manner: Acquire multimodal brain imaging data, overnight sleep EEG data, and longitudinal cognitive assessment data from historical subject populations; Calculate brain lymphocyte markers and sleep coupling markers for each historical subject; Based on the population data of the historical subject group, mediation effect analysis was used to screen out indicator pairs that have a significant mediating relationship with cognitive changes by influencing brain lymphatic indicators. The selected indicator pairs are used as correlation features, and together with the cognitive change corresponding to the correlation features, they form a training dataset. The machine learning model is trained based on the training dataset to obtain the risk prediction model.
[0011] Preferably, the mediation effect analysis includes: Using the sleep coupling index as the independent variable, the brain lymphatic index as the mediating variable, and the cognitive change as the dependent variable, a mediation effect analysis was conducted on the population data of the historical subject group to verify the pathological chain and identify index pairs with significant mediating associations. The aforementioned indicator pair is formed by combining brain lymphatic indicators and sleep coupling indicators of the same subject.
[0012] Preferably, the brain lymphocyte markers include at least one or more of the following combinations: Normalized choroid plexus volume, perivascular space load index, diffusion tensor imaging index along the perivascular space, and coupling strength index of blood oxygen level dependent signal and cerebrospinal fluid signal.
[0013] Preferably, the sleep coupling index includes at least one or more of the following combinations: The coupling strength between slow wave oscillation and Theta rhythm, the coupling strength between slow wave oscillation and sleep spindle, the preferred phase shift between slow wave oscillation and Theta rhythm, and the preferred phase shift between slow wave oscillation and sleep spindle.
[0014] Preferably, the sleep coupling index is calculated in the following manner: Slow-wave oscillation events, Theta rhythm events, and sleep spindle events were detected from EEG signals during NREM sleep. The instantaneous phase of slow-wave oscillation events and the instantaneous amplitude of Theta rhythm or spindle wave events are extracted using Hilbert transform. Calculate the slow-wave oscillation phase corresponding to the peak amplitude of the Theta rhythm or spindle wave event to obtain the corresponding coupling strength or preferred phase offset.
[0015] Preferably, the machine learning model is a LASSO regression model with L1 regularization, which selects the optimal regularization parameter through cross-validation to screen features with predictive value for disease progression and generate a risk score.
[0016] Preferably, the risk prediction model is configured to output a continuous risk score; The method further includes comparing the continuous risk score with a preset threshold range to determine the discrete risk level corresponding to the subject.
[0017] According to a second aspect of the present invention, a system for assessing the risk of progression of neurodegenerative diseases is provided, comprising: The data acquisition module is used to acquire multimodal brain imaging data and overnight sleep EEG data of the same subject; The first calculation module is used to calculate at least one brain lymphatic index reflecting the clearance function of the brain lymphatic system based on the multimodal brain imaging data. The second calculation module is used to calculate at least one sleep coupling index based on the overnight sleep EEG data, which reflects the coupling characteristics between slow wave oscillations during non-rapid eye movement sleep and Theta rhythm or sleep spindle waves. The risk assessment module is used to receive the subject's brain lymphatic indicators and sleep coupling indicators, and input them into a pre-trained risk prediction model to obtain the subject's risk assessment results. The pre-trained risk prediction model is based on historical population data. Through mediation analysis, brain lymphatic indicators and sleep coupling indicators that have a significant mediating association with cognitive decline are selected as association features, and the model is trained using these association features.
[0018] Preferably, the system further includes: The output module is connected to the risk assessment module to obtain the subject's current risk assessment result and, in combination with historical risk assessment results, output a risk assessment curve.
[0019] Compared with the prior art, the present invention has the following technical effects: (1) Mechanism-driven and highly interpretable: For the first time, this invention establishes a quantitative correlation between functional indicators of the brain lymphatic system and specific sleep oscillations in NREM through mediation analysis, constructing and validating a continuous pathological chain of "sleep coupling mismatch → impaired brain lymphatic clearance → cognitive decline". This enables the risk assessment model to not only provide predictive results but also offer mechanistic explanations, providing a clear biological basis for targeted interventions (such as improving sleep rhythms or enhancing brain clearance).
[0020] (2) Significantly improved predictive efficacy, enabling early warning: By inputting brain-sleep indicators with strong correlation verified by mediation analysis as core features into the machine learning model, the constructed risk prediction model showed significantly better accuracy (AUC) in predicting the progression of Alzheimer's disease than models using only a single modality (brain lymphocytes only or sleep EEG only). This synergistic effect can more sensitively identify high-risk individuals with cognitive decline in the early stages of the disease when brain structural lesions are not yet obvious.
[0021] (3) Non-invasive, objective, standardized, and easy to promote in clinical practice: The core indicators of the embodiments of this invention are all derived from routine non-invasive examinations (multimodal MRI and overnight polysomnography), which do not rely on the active cooperation of the subjects or the performance of complex cognitive tasks, and are especially suitable for patients with impaired executive function. At the same time, the embodiments of this invention provide a standardized and automated analysis process from data collection and indicator calculation to model construction, which greatly reduces the complexity of operation and dependence on professional personnel, and is conducive to large-scale application and longitudinal follow-up in memory clinics, sleep centers and large cohort studies.
[0022] (4) The output results are intuitive and support hierarchical management and decision-making: The final output of the model is a continuous risk score or a discrete risk level (such as low / medium / high risk). The results are intuitive and clear, and can directly serve clinicians' prognosis judgment, patient hierarchical management, clinical trial enrollment screening and individualized intervention plan formulation, realizing the transformation from cutting-edge scientific research results to operable clinical tools. Attached Figure Description
[0023] Figure 1 A flowchart illustrating a method for assessing the risk of progression of neurodegenerative diseases, provided in the first embodiment of the present invention; Figure 2 This is a flowchart illustrating the construction process of the risk prediction model in the first embodiment of the present invention. Figure 3 A system structure diagram for assessing the risk of progression of neurodegenerative diseases is provided in the second embodiment of the present invention; Figure 4 This is a system architecture diagram for assessing the risk of progression of neurodegenerative diseases, provided in the third embodiment of the present invention. Detailed Implementation
[0024] The technical content of the present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.
[0025] The present invention aims to propose a novel "brain-lymphatic-sleep coupling assessment paradigm" to functionally measure two key capabilities in the brain information processing process related to Alzheimer's disease: first, the clearance of brain metabolites and pathological proteins and the integrity of pathways by the glymphatic system; and second, the rhythmic coupling organization ability of the brain under the dominance of slow waves during NREM sleep, namely, the temporal coordination ability of coupled oscillations such as slow wave-theta and slow wave-spindle waves.
[0026] The brain lymphatic system is the fundamental pathway for clearing waste products from the brain, responsible for introducing cerebrospinal fluid (CSF) into the brain parenchyma along the perivascular space and expelling metabolic products through channels such as AQP4. Coupled oscillations during NREM sleep provide a "rhythmic drive" for CSF-interstitial fluid exchange through large-scale synchronous neuronal firing and hemodynamic changes. In this embodiment of the invention, the "structural / dynamic state of the clearance pathway" and the "temporal organization ability of sleep oscillations" are quantified as a whole functional module. The novel assessment paradigm provided by this embodiment physiologically simulates the real-world scenario of the brain during nighttime sleep, especially NREM sleep. Against a slow-wave-dominated rhythmic background, neuronal groups in the cortex and thalamus replay and consolidate memory traces through the rhythmic appearance of thetaburst and sleep spindle waves. Simultaneously, the periodic fluctuations in blood flow and CSF flow promote the expulsion of metabolic products such as Aβ and tau through the brain lymphatic pathway.
[0027] In this embodiment of the invention, objective quantitative indicators at the two levels mentioned above are obtained through conventional MRI (including DTI and resting-state fMRI) and overnight sleep EEG. At the imaging level, parameters such as PVS load, choroid plexus volume, DTI-ALPS index, and BOLD-CSF signal coupling are used to characterize the structural integrity and dynamic characteristics of brain lymphatic pathways. At the electrophysiological level, slow wave, thetaburst, and sleep spindle event detection and phase-amplitude coupling analysis are used to quantify the coupling strength, phase center, and phase mismatch degree between SO-theta and SO-spindle. By selecting these indicators that do not rely on subjective reports and fully reflect the automated collaborative processing of clearance pathways and sleep oscillations, the interference of higher cognitive factors such as executive function, motivation, and task compliance on the measurement results is minimized.
[0028] In practical implementation, this invention decomposes the "clearance capacity" of brain lymphatic function and the "organizational capacity" of sleep oscillation coupling into two complementary indicator dimensions: the first is the structural and dynamic dimension, including PVS morphology and load, choroid plexus volume changes, DTI-ALPS index along the perivascular pathway, and the temporal correlation between brain parenchyma BOLD signal and cerebrospinal fluid signal, to reflect whether the pathways for cerebrospinal fluid-interstitial fluid exchange and waste removal are unobstructed. The second is the time-phase dimension, including the coupling strength of SO-theta and SO-spindle, peak phase, and cross-cycle consistency, to reflect whether large-scale neuronal activity during sleep can form stable rhythmic coordination within an appropriate time window.
[0029] This invention combines two types of indicators to form a comprehensive parameter such as the "brain-lymphatic-sleep coupling index," which is used to assess whether an individual has already shown early changes in clearance function and sleep coupling disorder before the pathological protein load has progressed to an extreme degree. This enables early identification and follow-up monitoring of the risk of Alzheimer's disease progression.
[0030] First Embodiment like Figure 1 As shown, the first embodiment of the present invention provides a method for assessing the risk of progression of neurodegenerative diseases, which specifically includes the following steps: S1: Acquire multimodal brain imaging data and overnight sleep EEG data from the same subject.
[0031] Specifically, it includes the following steps: S11: Acquire multimodal brain imaging data of the subjects.
[0032] In this embodiment, a 3.0T high-field magnetic resonance scanner is used to acquire multimodal MRI data, and a pipeline based on neuroimaging analysis tools such as MATLAB, SPM12, CAT12, FSL, ANTs, and DPARSF is used to automatically extract brain lymph node-related indicators (refer to step S2).
[0033] In the specific implementation process, T1-weighted 3D DPRAGE, 3D FLAIR, DTI, and resting fMRI sequences were acquired first: T1 sequence TR≈1690ms, TE≈2.4ms, TI≈800ms, slice thickness 1mm; DTI sequence contains 64 b=1000s / mm 2 Diffusion direction and 11 b=0 images; rs-fMRI sequence TR≈2000ms, TE≈30ms; 3DFLAIR sequence TR≈5000ms, TE≈390ms, TI≈1800ms. On T1 images, CAT12 (based on SPM12) was used to automatically segment the whole brain, calculate the total intracranial volume (TIV), and the choroid plexus (CP) was automatically segmented using the 3DU-NET deep learning algorithm to obtain the CP volume; to reduce the influence of individual brain size differences, the CP volume was divided by the TIV to obtain the normalized index CP / TIV.
[0034] S12: Acquire the subject's overnight sleep EEG data.
[0035] Sleep EEG data were recorded overnight using a multichannel PSG system according to the internationally standardized 10–20 timescale. Subsequently, open-source toolkits such as Field Trip, EEGLAB, and MNE-Python were used for standardized preprocessing and oscillation coupling analysis. All signal sampling rates, filtering parameters, and threshold settings were standardized to ensure the comparability and repeatability of data among different subjects.
[0036] In practice, scalp electrodes (such as Fp1 / 2, F3 / 4, F7 / 8, Fz, C3 / 4, Cz, T3 / 4, P3 / 4, Pz, T5 / 6, O1 / 2) are placed according to the internationally standardized 10–20 system using a multi-channel PSG system, with the mastoid process as a reference, and the electrode impedance is controlled below 5kΩ. The signal is sampled at 256Hz, recorded by the Grael acquisition system, and monitored in real time.
[0037] S2: Calculate brain lymphatic indicators based on multimodal brain imaging data.
[0038] Based on the multimodal MRI data obtained in step S1, the clearance capacity of the brain lymphatic system at the structural and dynamic levels was quantified, resulting in a set of stable and repeatable brain lymphatic function indicators.
[0039] Specifically, step S2 includes the following steps: S21: Detection and analysis of perivascular spaces (PVS). For PVS analysis, T1 images were first spatially registered and preprocessed based on ANTs and FSL. Then, Frangi filtering was used to enhance slender tubular structures, automatically generating a PVS mask. The PVS volume fractions (PVSVF-WM, PVSSVF-BG, gPVSVF) were calculated in the white matter, basal ganglia, and whole brain, respectively, and normalized to the proportion of TIV to reflect the perivascular space load in different brain regions. The DTI-ALPS index was calculated based on the DTIFIT module in FSL. Tensor fitting was performed on the DTI data to obtain diffusion maps Dxx, Dyy, and Dzz along the x, y, and z directions, and registered to the JHU-ICBM-FA-2mm template. At the lateral ventricle plane (e.g., MNIz≈26), the ROIs of bilateral projection fibers (corona radiata fibers) and commissural fibers (superior longitudinal fasciculus) were extracted with a thickness of approximately 3mm and subdivided into anterior, middle, and posterior segments.
[0040] S22: Calculate multiple brain lymphocyte markers.
[0041] Dxx-proj, Dxx-assoc, Dyy-proj, and Dzz-assoc were calculated for each ROI. The DTI-ALPS values were then obtained in the left and right hemispheres according to the formula: DTI-ALPS index = (Dxx-proj + Dxx-assoc) / (Dyy-proj + Dzz-assoc). The average value was used as the global gDTI-ALPS index to characterize the degree of diffusion abnormality and fluid clearance capacity along the perivascular pathway of the vein.
[0042] BOLD-CSF coupling in resting-state fMRI is used to reflect the dynamic coordination between cortical activity and cerebrospinal fluid flow. DPARSF was used to preprocess rs-fMRI (time slice correction, head movement correction, spatial registration, etc.) to extract global BOLD signals (gBOLD), temporal BOLD, parietal pBOLD, insula iBOLD, and occipital oBOLD from the cortical gray matter regions defined by the AAL2 template, while preserving the original spatial resolution as much as possible. Simultaneously, CSF time series were extracted from the lateral ventricles or CSF masks. Subsequently, cross-correlation calculations were performed on BOLD and CSF signals within a lag range of -10 to +10 seconds. The absolute maximum correlation coefficient at approximately a lag of -4 seconds was defined as the gBOLD-CSF or regional BOLD-CSF coupling strength, and significance was tested by generating 95% confidence intervals through random re-pairing. Weaker coupling strength indicates more significant impairment of brain lymphatic dynamics.
[0043] Therefore, through the above steps S21 to S23, the calculated brain lymphatic indicators include at least one or more of the following combinations: normalized choroid plexus volume, perivascular space load index, diffusion tensor imaging index along the perivascular space, coupling strength between blood oxygen level-dependent signals and cerebrospinal fluid signals, and other multidimensional brain lymphatic function indicators, which provide a basis for subsequent correlation analysis with sleep oscillations and clinical outcomes.
[0044] Understandably, this embodiment is the first to integrate multiple MRI indicators related to the brain lymphatic system (CP volume / TIV, PVS volume fraction, DTI-ALPS index, BOLD-CSF coupling) into a unified "brain lymphatic function quantification task." Within the same subject, this task, through standardized acquisition and automated analysis pipelines, objectively reflects the coupling state of cerebrospinal fluid generation and circulation, perivascular space structural load, and cortical activity with CSF flow, forming a comprehensive indicator of brain clearance function independent of the subject's performance ability and subjective cooperation. This indicator can be used for the early identification of clearance impairments and disease progression assessment in Alzheimer's disease patients.
[0045] S3: Calculate sleep coupling index based on overnight sleep EEG data.
[0046] After data acquisition is completed through step S1 above, it is necessary to measure the rhythm coupling ability under slow wave dominance during NREM sleep, especially SO-theta and SO-spindle coupling and their phase mismatch, so as to quantify the ability of large-scale neuronal synchronization and rhythm organization during sleep.
[0047] Specifically, step S3 includes the following steps: S31: EEG analysis.
[0048] During the EEG analysis phase, tools such as Field Trip, EEGLAB, and MNE-Python were used to apply bandpass filtering of 0.1–40 Hz to the EEG / EOG signals. A 50 Hz notch filter was added to the EMG signals to remove power supply noise, and then the signals were uniformly downsampled to 100 Hz. High-artifact segments were removed by combining manual visual inspection with an automatic algorithm based on median absolute deviation (MAD). Finally, sleep experts completed offline sleep staging in 30-second increments according to the AASM standard.
[0049] Subsequent analysis selected EEG fragments with fewer artifacts from NREM sleep (especially N2 / N3 stages). For slow wave (SO) detection, the central lead (e.g., C3) was used as the main channel, and a bandpass filter of 0.16–1.25 Hz was applied to identify candidate SO events with durations of 0.8–2 s. Events with peak-to-peak amplitudes in the highest 25% of the subject's total were selected as valid SOs. Artifact-free fragments of ±2 s centered on the SO trough were extracted for subsequent coupling analysis. Theta bursts and sleep spindles were detected using filters of 4–8 Hz and 12–16 Hz, respectively. The sliding RMS amplitude was calculated, and high-amplitude events were selected with a threshold of the 75th percentile, limiting the duration to 0.5–3.0 s. Similarly, artifact-free fragments of ±2 s centered on the peak amplitude of the event were extracted.
[0050] S32: Construct phase-power distribution.
[0051] To prevent systematic biases such as memory effects or event superposition, each event is derived from an independent time window and is repeatedly detected in different sleep cycles. This embodiment of the invention uses the Hilbert transform to extract the instantaneous phase from the SO signal and the instantaneous amplitude envelope from the theta and spindle wave signals. For each theta or spindle wave event, the SO phase angle corresponding to the occurrence of its amplitude peak is recorded to obtain the "preferred SO phase" for that event.
[0052] Then, the preferred phases of all events for each subject are summarized, and the average phase and distribution concentration are calculated to characterize the coupling center and coupling strength between SO–theta and SO–spindle. Each SO cycle is then divided into 18 equally divided phase intervals, and the normalized theta or spindle wave power in each phase interval is statistically analyzed to construct the phase-power distribution.
[0053] S33: Quantize the degree of "mismatch" in coupled oscillation.
[0054] By comparing the offset of the preferred phase and the concentration of the phase-power distribution in different individuals or groups, the degree of "mismatch" in coupled oscillations can be quantitatively reflected. The more severe the coupling mismatch, the more likely the theta or spindle wave is no longer stably locked within a specific phase window of the SO, suggesting impaired temporal organization of neural oscillations during sleep. Understandably, this mismatch represents a key abnormal signal in the sleep-lymphatic coordination mechanism.
[0055] Therefore, through the operations of S31 to S33 above, the sleep coupling index determined in step S3 includes at least one or more of the following combinations: the coupling strength between slow wave oscillation and Theta rhythm, the coupling strength between slow wave oscillation and sleep spindle, the preferred phase shift between slow wave oscillation and Theta rhythm, and the preferred phase shift between slow wave oscillation and sleep spindle.
[0056] Understandably, this embodiment of the invention constructs a standardized NREM sleep oscillation coupling measurement task. In overnight PSG data, SO (0.16–1.25 Hz), Theta bursts (4–8 Hz), and sleep spindle events (12–16 Hz) are automatically detected. The phase-amplitude coupling (PAC) of SO–theta and SO–spindle is calculated using Hilbert transform, along with the optimal SO phase and phase-power distribution. Furthermore, two key indicators, "coupling strength" and "coupling mismatch," are extracted. This task quantifies the rhythmic organization of large-scale neuronal activity during NREM sleep without requiring patients to perform any explicit behavioral tasks. It is particularly suitable for patients with cognitive impairments such as Alzheimer's disease, and can sensitively capture sleep rhythm abnormalities and early changes related to cognitive decline.
[0057] S4: Input the calculated brain lymphatic indicators and sleep coupling indicators into the pre-trained risk prediction model to obtain the risk assessment results of the subjects.
[0058] It is understood that in this embodiment, a risk prediction model needs to be built in advance, and then the brain lymphatic indicators and sleep coupling indicators calculated in the above steps S2 to S3 are input into the risk prediction model to obtain the risk assessment results of the subjects.
[0059] Reference Figure 2 As shown below, the construction process of this risk prediction model is explained in detail: S10: Acquire multimodal brain imaging data, overnight sleep EEG data, and longitudinal cognitive assessment data from historical subject populations.
[0060] In this embodiment, the historical subject population includes Alzheimer's disease patients and cognitively normal control individuals. For each subject, based on two consecutive years of follow-up records, multimodal brain imaging data, overnight sleep EEG data, and longitudinal cognitive assessment data were acquired to obtain the population data of the historical subject population.
[0061] S20: Calculate brain lymphatic indicators and sleep coupling indicators for each historical subject.
[0062] It is understandable that the specific calculation process of step S20 is the same as that of steps S2 to S3 above, and will not be repeated here.
[0063] S30: Based on population data from historical subject groups, mediation effect analysis was used to screen out indicator pairs that show a significant mediating relationship between sleep coupling indicators and cognitive changes by influencing brain lymphatic indicators.
[0064] Based on the multidimensional brain lymphatic and sleep coupling indicators obtained in step S20 above, and combined with baseline and follow-up cognitive assessment results, statistical analysis and machine learning modeling are used to quantitatively predict the risk of Alzheimer's disease progression. Specifically, the following steps are included: S301: Assess the correspondence between impaired brain lymphatic function and abnormal oscillatory coupling.
[0065] In this embodiment, correlation and multiple regression analyses were performed on brain lymphatic indicators (such as CP / TIV, gDTI-ALPS, gBOLD-CSF and regional BOLD-CSF coupling) and sleep coupling indicators (such as the coupling strength, preferred phase and coupling mismatch of SO-theta and SO-spindle) to assess the correspondence between impaired brain lymphatic function and abnormal oscillatory coupling.
[0066] For example, in a preferred embodiment of the invention, lower gDTI-ALPS are positively correlated with more pronounced SO–spindle coupling mismatch, suggesting a functional association between decreased brain lymphatic clearance capacity and coupling oscillation disorder.
[0067] S302: Mediation effect analysis based on population data from historical subject groups.
[0068] Specifically, in this embodiment, sleep coupling mismatch is used as the independent variable, brain lymphatic indicators as the mediating variable, and two-year follow-up cognitive changes (such as changes in MMSE or CDR) as the dependent variable to conduct a mediation effect analysis to verify whether brain lymphatic function plays a mediating role in the "sleep coupling abnormality → cognitive decline" pathway. If the mediation effect is significant, it supports the proposed pathological chain of coupling oscillations, which accelerates the accumulation of pathological proteins such as Aβ / tau and cognitive decline by affecting brain lymphatic clearance efficiency.
[0069] Therefore, through the above mediation effect analysis, we identified indicator pairs that significantly mediate cognitive changes by influencing brain lymphatic indicators. Each indicator pair is formed by the brain lymphatic indicators and sleep coupling indicators of the same subject.
[0070] S40: Construct the training dataset.
[0071] After multiple indicator pairs are selected through step S30 above, each selected indicator pair is used as a correlation feature, and together with the cognitive change corresponding to the correlation feature, they constitute the training dataset.
[0072] S50: Train the machine learning model based on the training dataset.
[0073] In terms of predictive model construction, this embodiment of the invention inputs the aforementioned brain lymphatic indicators and sleep coupling indicators, along with covariates such as age, gender, years of education, and baseline cognitive score, into a L1-regularized LASSO regression or other suitable machine learning algorithm. The regularization parameters are optimized through cross-validation, and key features with independent predictive value for disease progression are automatically selected, thereby constructing a risk prediction model for disease progression.
[0074] The output of this risk prediction model can be a continuous risk score or a binary / multi-level risk stratification (such as low risk, medium risk, high risk), and can be used to predict whether clinical stage transition or significant cognitive decline will occur within two years.
[0075] Through repeated measurements on subjects and validation on independent external datasets, the results of this invention demonstrate that, compared to predictive models constructed solely using brain lymphatic function indicators (AUC = 0.813) or those constructed solely using sleep EEG indicators (AUC = 0.816), the model based on a combination of brain lymphatic function indicators and sleep oscillation coupling indicators exhibits higher discriminative ability and better stability in predicting the risk of Alzheimer's disease progression, with an AUC reaching 0.864. Therefore, this invention enables early identification and risk stratification of patients with cognitive impairment during follow-up, providing a mechanism-oriented quantitative tool for subsequent individualized intervention strategy development and efficacy evaluation.
[0076] Therefore, by inputting the brain lymphatic indicators and sleep coupling indicators calculated in steps S2 and S3 into the trained risk prediction model, the risk assessment result for the subject can be obtained. Specifically, the risk prediction model is configured to output a continuous risk score, and then compare the continuous risk score with a preset threshold range to finally determine the discrete risk level corresponding to the subject.
[0077] It is important to emphasize that this invention, for the first time, integrates "brain lymphatic function and sleep oscillation coupling" as a holistic functional module for predicting Alzheimer's disease progression. On one hand, it quantitatively reveals the chain of action of "sleep coupling mismatch → impaired brain lymphatic function → cognitive decline" through correlation and mediation analyses. On the other hand, using brain lymphatic indicators and sleep coupling indicators as core inputs, combined with covariates such as age, education, and baseline cognition, a progression risk model is constructed using regularization algorithms such as LASSO, outputting an individualized risk score or risk level. This task achieves a transformation from a "mechanism chain" to a "clinical prediction tool," making the "brain lymphatic-sleep coupling index" both interpretable and directly serving early clinical diagnosis and disease follow-up.
[0078] Second Embodiment like Figure 3 As shown, based on the first embodiment described above, the second embodiment of the present invention further provides a system for assessing the risk of progression of neurodegenerative diseases, including a data acquisition module 1, a first calculation module 2, a second calculation module 3, a risk assessment module 4, and an output module 5. Wherein, Data acquisition module 1 is used to acquire multimodal brain imaging data and overnight sleep EEG data of the same subject. The specific data acquisition process is the same as described in the first embodiment above, and will not be repeated here.
[0079] The first calculation module 2 is used to calculate at least one brain lymphatic index reflecting the clearance function of the brain lymphatic system based on multimodal brain imaging data. The specific calculation process of the brain lymphatic index is the same as that in the first embodiment described above, and will not be repeated here.
[0080] The second calculation module 3 is used to calculate at least one sleep coupling index based on overnight sleep EEG data, reflecting the coupling characteristics between slow-wave oscillations during non-rapid eye movement (NREM) sleep and Theta rhythm or sleep spindle waves. The specific calculation process for the sleep coupling index is the same as described in the first embodiment above, and will not be repeated here.
[0081] The risk assessment module 4 receives the subject's brain lymphatic indicators and sleep coupling indicators, and inputs them into a pre-trained risk prediction model to obtain the subject's risk assessment results. The construction process of this risk prediction model is the same as in the first embodiment described above, and will not be repeated here.
[0082] Furthermore, output module 5 is connected to risk assessment module 4 to obtain the subject's current risk assessment results and, in conjunction with historical risk assessment results, output a risk assessment curve. This output helps subjects and physicians understand the disease progression trend and provides a reference for treatment decisions.
[0083] It is understood that the functions and cooperation relationships of each module unit in this embodiment are only one specific implementation method for the method in the first embodiment above. In other embodiments, the functions and cooperation relationships of each module unit can be adaptively adjusted as needed.
[0084] Third Embodiment like Figure 4 As shown, based on the first embodiment described above, the third embodiment of the present invention further provides a system for assessing the risk of progression of neurodegenerative diseases. The system includes one or more processors and a memory. The memory is coupled to the processor and is used to store one or more programs, which, when executed by the processor, cause the processor to implement the method for assessing the risk of progression of neurodegenerative diseases as described in the above embodiments.
[0085] The processor controls the overall operation of the system to complete all or part of the steps in the method for assessing the risk of neurodegenerative disease progression. The processor can be a central processing unit (CPU), graphics processing unit (GPU), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), digital signal processing (DSP) chip, etc. The memory stores various types of data to support the operation of the system. This data may include, for example, instructions for any application or method operating on the system, as well as application-related data. The memory can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, etc.
[0086] In one exemplary embodiment, the system may be implemented by a computer chip or physical entity, or by a product with certain functions, for performing the method described above for assessing the risk of progression of neurodegenerative diseases and achieving the same technical effects as the method described above. A typical embodiment is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, an in-vehicle human-machine interface device, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.
[0087] In another exemplary embodiment, the present invention also provides a computer-readable storage medium including program instructions that, when executed by a processor, implement the steps of the method for assessing the risk of progression of neurodegenerative diseases in any of the above embodiments. For example, the computer-readable storage medium may be the memory including the program instructions described above, which can be executed by a system processor to complete the method for assessing the risk of progression of neurodegenerative diseases described above, and achieve the same technical effects as the method described above.
[0088] It should be noted that the above embodiments are merely illustrative examples. The technical solutions of each embodiment can be combined, and all are within the protection scope of this invention.
[0089] 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 one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0090] The method and system for assessing the risk of progression of neurodegenerative diseases provided by this invention have been described in detail above. Any obvious modifications made to this invention by those skilled in the art without departing from its essential content will constitute an infringement of the patent rights of this invention and will incur corresponding legal liability.
Claims
1. A method for assessing the risk of progression of neurodegenerative diseases, characterized in that... Includes the following steps: S1: Acquire multimodal brain imaging data and overnight sleep EEG data from the same subject; S2: Based on the multimodal brain imaging data, calculate at least one brain lymphatic index that reflects the clearance function of the brain lymphatic system; S3: Based on the overnight sleep EEG data, calculate at least one sleep coupling index that reflects the coupling characteristics between slow wave oscillations during non-rapid eye movement sleep and Theta rhythm or sleep spindle waves. S4: Input the calculated brain lymphatic indicators and sleep coupling indicators into the pre-trained risk prediction model to obtain the risk assessment results of the subject; The pre-trained risk prediction model is based on historical population data. Through mediation analysis, brain lymphatic indicators and sleep coupling indicators that have a significant mediating association with cognitive decline are selected as association features, and the model is trained using these association features.
2. The method as described in claim 1, characterized in that, The pre-trained risk prediction model is constructed in the following way: Acquire multimodal brain imaging data, overnight sleep EEG data, and longitudinal cognitive assessment data from historical subject populations; Calculate brain lymphocyte markers and sleep coupling markers for each historical subject; Based on the population data of the historical subject group, mediation effect analysis was used to screen out indicator pairs that have a significant mediating relationship with cognitive changes by influencing brain lymphatic indicators. The selected indicator pairs are used as correlation features, and together with the cognitive change corresponding to the correlation features, they form a training dataset. The machine learning model is trained based on the training dataset to obtain the risk prediction model.
3. The method as described in claim 2, characterized in that... The mediation effect analysis includes: Using the sleep coupling index as the independent variable, the brain lymphatic index as the mediating variable, and the cognitive change as the dependent variable, a mediation effect analysis was conducted on the population data of the historical subject group to verify the pathological chain and identify index pairs with significant mediating associations. The aforementioned indicator pair is formed by combining brain lymphatic indicators and sleep coupling indicators of the same subject.
4. The method as described in claim 1 or 2, characterized in that... The brain lymphocyte markers include at least one or more of the following combinations: Normalized choroid plexus volume, perivascular space load index, diffusion tensor imaging index along the perivascular space, and coupling strength index of blood oxygen level dependent signal and cerebrospinal fluid signal.
5. The method as described in claim 1 or 2, characterized in that... The sleep coupling index includes at least one or more of the following combinations: The coupling strength between slow wave oscillation and Theta rhythm, the coupling strength between slow wave oscillation and sleep spindle, the preferred phase shift between slow wave oscillation and Theta rhythm, and the preferred phase shift between slow wave oscillation and sleep spindle.
6. The method as described in claim 5, characterized in that, The sleep coupling index is calculated in the following way: Slow-wave oscillation events, Theta rhythm events, and sleep spindle events were detected from EEG signals during NREM sleep. The instantaneous phase of slow-wave oscillation events and the instantaneous amplitude of Theta rhythm or spindle wave events are extracted using Hilbert transform. Calculate the slow-wave oscillation phase corresponding to the peak amplitude of the Theta rhythm or spindle wave event to obtain the corresponding coupling strength or preferred phase offset.
7. The method as described in claim 2, characterized in that, The machine learning model is a LASSO regression model with L1 regularization. The optimal regularization parameter is selected through cross-validation to screen features with predictive value for disease progression and generate a risk score.
8. The method as described in claim 7, characterized in that, The risk prediction model is configured to output a continuous risk score; The method further includes comparing the continuous risk score with a preset threshold range to determine the discrete risk level corresponding to the subject.
9. A system for assessing the risk of progression of neurodegenerative diseases, characterized in that... include: The data acquisition module is used to acquire multimodal brain imaging data and overnight sleep EEG data of the same subject; The first calculation module is used to calculate at least one brain lymphatic index reflecting the clearance function of the brain lymphatic system based on the multimodal brain imaging data. The second calculation module is used to calculate at least one sleep coupling index based on the overnight sleep EEG data, which reflects the coupling characteristics between slow wave oscillations during non-rapid eye movement sleep and Theta rhythm or sleep spindle waves. The risk assessment module is used to receive the subject's brain lymphatic indicators and sleep coupling indicators, and input them into a pre-trained risk prediction model to obtain the subject's risk assessment results. The pre-trained risk prediction model is based on historical population data. Through mediation analysis, brain lymphatic indicators and sleep coupling indicators that have a significant mediating association with cognitive decline are selected as association features, and the model is trained using these association features.
10. The system as described in claim 9, characterized in that... Also includes: The output module is connected to the risk assessment module to obtain the subject's current risk assessment result and, in combination with historical risk assessment results, output a risk assessment curve.