Augmem: A Novel Digital Cognitive Assessment for the Early Detection of Alzheimer's Disease

Augmem：一种用于早期检测阿尔茨海默病的新型数字认知评估

• 批准号: 10688227
• 负责人: Adele Gilpin
• 金额: $ 96.07万
• 依托单位: AUGNITION LABS LLC
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2024-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10688227
• 关键词: Acceleration; Achievement; Administrator; Aducanumab; Adult; Age; Aging; Alzheimer disease detection; Alzheimer' s Disease; Alzheimer' s Disease Pathway; Alzheimer' s disease care; Alzheimer' s disease diagnosis; Alzheimer' s disease patient; Alzheimer' s disease risk; American; Amyloid; Architecture; Autopsy; Biological; Biological Markers; Brain; Caregiver Burden; Classification; Clinical; Clinical Trials; Clinical Trials Design; Code; Cognitive; Collaborations; Collection; Computer software; Data; Data Analytics; Data Collection; Data Scientist; Dementia; Development; Digital biomarker; Dimensions; Disease; Early Diagnosis; Early Intervention; Elderly; Episodic memory; Evaluation; Goals; Health; Healthcare Systems; Hippocampus; Immunotherapy; Impaired cognition; Impairment; Infrastructure; Intelligence; MRI Scans; Marketing; Measures; Medicaid; Medical Device; Medicare; Memory; Modeling; Monitor; Neurobiology; Neurofibrillary Tangles; Neuropsychology; Participant; Patient-Focused Outcomes; Patients; Pattern; Performance; Pharmaceutical Preparations; Phase; Population; Prediction of Response to Therapy; Prevention; Probability; Process; Property; Psychometrics; Public Health; Qualifying; Regulatory Pathway; Research; Sampling; Secure; Senile Plaques; Small Business Innovation Research Grant; Social Sciences; Stratification; Surveys; Symptoms; Techniques; Technology; Testing; Therapeutic; Training; Treatment Protocols; United States; Work; age group; baby boomer; care costs; clinical care; clinical outcome assessment; cognitive testing; commercialization; data cleaning; diagnostic tool; digital; digital delivery; early detection biomarkers; economic cost; effective therapy; evidence base; feature extraction; feature selection; health application; high resolution imaging; human old age (65+); improved; innovation; large scale data; model building; neural; novel; novel therapeutics; patient stratification; payment; pre-clinical; predictive modeling; prodromal Alzheimer' s disease; prospective; recruit; standard of care; tau Proteins; tool; treatment response

Summary: Definitive diagnosis of Alzheimer’s Disease (AD) is currently conferred upon autopsy. Probable AD diagnosis is based on a combination of clinical/cognitive measures, often corroborated by structural MRI scans. Limitations of current neuropsychological and clinical tools for precise and early indications of cognitive decline in AD provide the impetus for our focus on developing improved cognitive assessments that are easy to use across platforms, age groups, and diverse cultural groups, and provide an earlier and more accurate indication of preclinical disease. Early diagnosis and intervention are critical for therapeutics to be maximally effective despite the dearth of new therapeutic options for AD. Augnition Labs is developing the Augmem™ digital biomarker platform based on work by Dr. Yassa and colleagues that empirically demonstrated, using a pattern separation task, that the chief function of the hippocampus is pattern separation – the ability to discriminate among similar memories by storing them using unique neural codes. We have developed, validated, and demonstrated the utility of a full suite of pattern separation tasks across the three key dimensions of episodic memory, (1) what happened (object), (2) where it happened (spatial), and (3) when it happened (temporal). Prior work has been neurobiologically validated with high resolution imaging as well as clinically validated against traditional clinical memory measures. In this Direct to Phase II SBIR, we incorporate object, spatial, and temporal pattern separation techniques with feature-rich AI models to produce a more effective digital biomarker for the early prediction of cognitive decline and treatment response. Aim 1. Develop and launch secure and scalable Augmem™ platform. We will develop and implement test management architecture and study administration modules in support of data collection, quality checks, and data analytics. A commercially ready front-end interface for digital delivery of assessments will be iteratively developed and tested. Goal: Completion of User Acceptance Testing with recruited user personas (study participant, study administrator, data scientist), and initiation of FDA regulatory pathway for Clinical Outcome Assessment qualification. Aim 2. Develop and train AI models for predicting subtle impairments based on cognitive and biomarker profiles. Data collection, data cleaning, feature extraction and selection, model building, and model evaluation and analysis will incorporate object, spatial, and temporal pattern separation measures from data collected through the Precision Aging Network as well as directly by Augnition. Goal: A representative sample of up to 500,000 participants across the age spectrum of 18-85, AI engine training, and achievement of predictive accuracy for age of 0.85 ROC AUC (classification) and RMSE ≤ 0.3 (regression). Upon successful completion of the proposed development, we will conduct prospective trials in preclinical/prodromal Alzheimer’s disease to fully validate the predictive power of the Augmem™ platform and initiate the Software as a Medical Device FDA regulatory pathway for AD early detection, stratification, and prediction of treatment response.

摘要：阿尔茨海默病(AD)的明确诊断目前是通过尸检来确定的。可能是AD 诊断是基于临床/认知措施的组合，通常得到结构性核磁共振扫描的证实。 当前神经心理学和临床工具对认知功能下降的精确和早期指征的局限性 在AD中为我们提供动力，使我们专注于开发易于使用的改进认知评估 跨平台、跨年龄段、跨文化群体，并提供更早、更准确的指示 临床前疾病。早期诊断和干预是治疗取得最大效果的关键 尽管缺乏治疗阿尔茨海默病的新方法。Augnition Labs正在开发Augmem™数字 基于Yassa博士和他的同事的工作的生物标记物平台，使用一种模式进行了经验演示 分离任务，即海马体的主要功能是模式分离--辨别能力 通过使用独特的神经编码存储它们，从而在相似的记忆中进行识别。我们已经开发、验证和 演示了全套模式分离任务在剧集的三个关键维度上的实用性 记忆，(1)发生了什么(物体)，(2)发生在哪里(空间)，和(3)发生时(时间)。之前 这项工作已经用高分辨率成像进行了神经生物学验证，并在临床上得到了验证 传统的临床记忆测量方法。在这个直接到第二阶段的SBIR中，我们将对象、空间和时间结合在一起 模式分离技术与特征丰富的人工智能模型，以产生更有效的数字生物标志物 早期预测认知功能下降和治疗反应。目标1.开发和发布安全且可扩展的 Augmem™平台。我们将开发和实施测试管理架构和学习管理 支持数据收集、质量检查和数据分析的模块。商业上准备好的前端 将反复开发和测试数字交付评估的界面。目标：完成用户 使用招募的用户角色(研究参与者、研究管理员、数据科学家)进行验收测试，以及 启动FDA临床结果评估资格的调控途径。目标2.发展和培训 基于认知和生物标记物特征预测细微损伤的AI模型。数据收集， 数据清理、特征提取和选择、模型构建以及模型评估和分析将 从通过Precision收集的数据中整合对象、空间和时间模式分离措施 老化网络以及直接点火。目标：多达500,000名参与者的代表性样本 在18-85岁的年龄范围内，接受人工智能引擎培训，并实现对0.85岁的预测准确性 ROC AUC(分类)和RMSE≤0.3(回归)。在成功完成建议的 开发中，我们将对临床前/先兆阿尔茨海默病进行前瞻性试验，以充分验证 奥格梅姆™平台的预测能力和启动该软件作为医疗器械FDA的监管 AD早期检测、分层和治疗反应预测的途径。