Dataset Information

Generalizable, Reproducible, and Neuroscientifically Interpretable Imaging Biomarkers for Alzheimer's Disease.

ABSTRACT: Precision medicine for Alzheimer's disease (AD) necessitates the development of personalized, reproducible, and neuroscientifically interpretable biomarkers, yet despite remarkable advances, few such biomarkers are available. Also, a comprehensive evaluation of the neurobiological basis and generalizability of the end-to-end machine learning system should be given the highest priority. For this reason, a deep learning model (3D attention network, 3DAN) that can simultaneously capture candidate imaging biomarkers with an attention mechanism module and advance the diagnosis of AD based on structural magnetic resonance imaging is proposed. The generalizability and reproducibility are evaluated using cross-validation on in-house, multicenter (n = 716), and public (n = 1116) databases with an accuracy up to 92%. Significant associations between the classification output and clinical characteristics of AD and mild cognitive impairment (MCI, a middle stage of dementia) groups provide solid neurobiological support for the 3DAN model. The effectiveness of the 3DAN model is further validated by its good performance in predicting the MCI subjects who progress to AD with an accuracy of 72%. Collectively, the findings highlight the potential for structural brain imaging to provide a generalizable, and neuroscientifically interpretable imaging biomarker that can support clinicians in the early diagnosis of AD.

SUBMITTER: Jin D

PROVIDER: S-EPMC7375255 | biostudies-literature | 2020 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Generalizable, Reproducible, and Neuroscientifically Interpretable Imaging Biomarkers for Alzheimer's Disease.

Jin Dan D Zhou Bo B Han Ying Y Ren Jiaji J Han Tong T Liu Bing B Lu Jie J Song Chengyuan C Wang Pan P Wang Dawei D Xu Jian J Yang Zhengyi Z Yao Hongxiang H Yu Chunshui C Zhao Kun K Wintermark Max M Zuo Nianming N Zhang Xinqing X Zhou Yuying Y Zhang Xi X Jiang Tianzi T Wang Qing Q Liu Yong Y

Advanced science (Weinheim, Baden-Wurttemberg, Germany) 20200609 14

Precision medicine for Alzheimer's disease (AD) necessitates the development of personalized, reproducible, and neuroscientifically interpretable biomarkers, yet despite remarkable advances, few such biomarkers are available. Also, a comprehensive evaluation of the neurobiological basis and generalizability of the end-to-end machine learning system should be given the highest priority. For this reason, a deep learning model (3D attention network, 3DAN) that can simultaneously capture candidate i ...[more]

PMID: 32714766

Similar Datasets

Project description:IntroductionStratification of Alzheimer's disease (AD) patients into risk subgroups using Polygenic Risk Scores (PRS) presents novel opportunities for the development of clinical trials and disease-modifying therapies. However, the heterogeneous nature of AD continues to pose significant challenges for the clinical broadscale use of PRS. PRS remains unfit in demonstrating sufficient accuracy in risk prediction, particularly for individuals with mild cognitive impairment (MCI), and in allowing feasible interpretation of specific genes or SNPs contributing to disease risk. We propose adORS, a novel oligogenic risk score for AD, to better predict risk of disease by using an optimized list of relevant genetic risk factors.MethodsUsing whole genome sequencing data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort (n = 1,545), we selected 20 genes that exhibited the strongest correlations with FDG-PET and AV45-PET, recognized neuroimaging biomarkers that detect functional brain changes in AD. This subset of genes was incorporated into adORS to assess, in comparison to PRS, the prediction accuracy of CN vs. AD classification and MCI conversion prediction, risk stratification of the ADNI cohort, and interpretability of the genetic information included in the scores.ResultsadORS improved AUC scores over PRS in both CN vs. AD classification and MCI conversion prediction. The oligogenic model also refined risk-based stratification, even without the assistance of APOE, thus reflecting the true prevalence rate of the ADNI cohort compared to PRS. Interpretation analysis shows that genes included in adORS, such as ATF6, EFCAB11, ING5, SIK3, and CD46, have been observed in similar neurodegenerative disorders and/or are supported by AD-related literature.DiscussionCompared to conventional PRS, adORS may prove to be a more appropriate choice of differentiating patients into high or low genetic risk of AD in clinical studies or settings. Additionally, the ability to interpret specific genetic information allows the focus to be shifted from general relative risk based on a given population to the information that adORS can provide for a single individual, thus permitting the possibility of personalized treatments for AD.

Project description:Alzheimer's disease (AD) is neuropathologically characterized by the extracellular deposition of the amyloid-β peptide (Aβ) and the intraneuronal accumulation of abnormal phosphorylated tau (τ)-protein (p-τ). Most frequently, these hallmark lesions are accompanied by other co-pathologies in the brain that may contribute to cognitive impairment, such as vascular lesions, intraneuronal accumulation of phosphorylated transactive-response DNA-binding protein 43 (TDP-43), and/or α-synuclein (αSyn) aggregates. To estimate the extent of these AD and co-pathologies in patients, several biomarkers have been developed. Specific tracers target and visualize Aβ plaques, p-τ and αSyn pathology or inflammation by positron emission tomography. In addition to these imaging biomarkers, cerebrospinal fluid, and blood-based biomarker assays reflecting AD-specific or non-specific processes are either already in clinical use or in development. In this review, we will introduce the pathological lesions of the AD brain, the related biomarkers, and discuss to what extent the respective biomarkers estimate the pathology determined at post-mortem histopathological analysis. It became evident that initial stages of Aβ plaque and p-τ pathology are not detected with the currently available biomarkers. Interestingly, p-τ pathology precedes Aβ deposition, especially in the beginning of the disease when biomarkers are unable to detect it. Later, Aβ takes the lead and accelerates p-τ pathology, fitting well with the known evolution of biomarker measures over time. Some co-pathologies still lack clinically established biomarkers today, such as TDP-43 pathology or cortical microinfarcts. In summary, specific biomarkers for AD-related pathologies allow accurate clinical diagnosis of AD based on pathobiological parameters. Although current biomarkers are excellent measures for the respective pathologies, they fail to detect initial stages of the disease for which post-mortem analysis of the brain is still required. Accordingly, neuropathological studies remain essential to understand disease development especially in early stages. Moreover, there is an urgent need for biomarkers reflecting co-pathologies, such as limbic predominant, age-related TDP-43 encephalopathy-related pathology, which is known to modify the disease by interacting with p-τ. Novel biomarker approaches such as extracellular vesicle-based assays and cryptic RNA/peptides may help to better detect these co-pathologies in the future.

Project description:BackgroundDue to the class imbalance issue faced when Alzheimer's disease (AD) develops from normal cognition (NC) to mild cognitive impairment (MCI), present clinical practice is met with challenges regarding the auxiliary diagnosis of AD using machine learning (ML). This leads to low diagnosis performance. We aimed to construct an interpretable framework, extreme gradient boosting-Shapley additive explanations (XGBoost-SHAP), to handle the imbalance among different AD progression statuses at the algorithmic level. We also sought to achieve multiclassification of NC, MCI, and AD.MethodsWe obtained patient data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, including clinical information, neuropsychological test results, neuroimaging-derived biomarkers, and APOE-ε4 gene statuses. First, three feature selection algorithms were applied, and they were then included in the XGBoost algorithm. Due to the imbalance among the three classes, we changed the sample weight distribution to achieve multiclassification of NC, MCI, and AD. Then, the SHAP method was linked to XGBoost to form an interpretable framework. This framework utilized attribution ideas that quantified the impacts of model predictions into numerical values and analysed them based on their directions and sizes. Subsequently, the top 10 features (optimal subset) were used to simplify the clinical decision-making process, and their performance was compared with that of a random forest (RF), Bagging, AdaBoost, and a naive Bayes (NB) classifier. Finally, the National Alzheimer's Coordinating Center (NACC) dataset was employed to assess the impact path consistency of the features within the optimal subset.ResultsCompared to the RF, Bagging, AdaBoost, NB and XGBoost (unweighted), the interpretable framework had higher classification performance with accuracy improvements of 0.74%, 0.74%, 1.46%, 13.18%, and 0.83%, respectively. The framework achieved high sensitivity (81.21%/74.85%), specificity (92.18%/89.86%), accuracy (87.57%/80.52%), area under the receiver operating characteristic curve (AUC) (0.91/0.88), positive clinical utility index (0.71/0.56), and negative clinical utility index (0.75/0.68) on the ADNI and NACC datasets, respectively. In the ADNI dataset, the top 10 features were found to have varying associations with the risk of AD onset based on their SHAP values. Specifically, the higher SHAP values of CDRSB, ADAS13, ADAS11, ventricle volume, ADASQ4, and FAQ were associated with higher risks of AD onset. Conversely, the higher SHAP values of LDELTOTAL, mPACCdigit, RAVLT_immediate, and MMSE were associated with lower risks of AD onset. Similar results were found for the NACC dataset.ConclusionsThe proposed interpretable framework contributes to achieving excellent performance in imbalanced AD multiclassification tasks and provides scientific guidance (optimal subset) for clinical decision-making, thereby facilitating disease management and offering new research ideas for optimizing AD prevention and treatment programs.

Dataset Information

Generalizable, Reproducible, and Neuroscientifically Interpretable Imaging Biomarkers for Alzheimer's Disease.

Publications

Generalizable, Reproducible, and Neuroscientifically Interpretable Imaging Biomarkers for Alzheimer's Disease.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets