DeconDTN: Deconfounding Deep Transformer Networks for Clinical NLP

DeconDTN：为临床 NLP 解构深度 Transformer 网络

• 批准号: 10711315
• 负责人: Trevor Cohen
• 金额: $ 31.12万
• 依托单位: UNIVERSITY OF WASHINGTON
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-06-01 至 2026-02-28
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10711315
• 关键词: Acceleration; Adverse effects; Affect; African American; African American population; Alzheimer disease detection; Alzheimer' s Disease; Alzheimer' s disease diagnosis; Alzheimer' s disease patient; Award; Caregivers; Caring; Classification; Clinical; Cognitive; Confounding Factors (Epidemiology); Data; Data Provenance; Data Set; Dementia; Detection; Diagnosis; Disease; Early Diagnosis; Elderly; Emotional; Equity; Ethnic Origin; Ethnic Population; Family; Family Relationship; Financial Hardship; Future; Genetic Transcription; Health Care Costs; Healthcare Systems; Hispanic Americans; Individual; Inequity; Institution; Language; Learning; Linguistics; Measures; Mediating; Methods; Modeling; Modification; Monitor; Natural Language Processing; Nature; Neural Network Simulation; Neurobehavioral Manifestations; Outcome; Parents; Participant; Patients; Performance; Personality; Persons; Population; Prognosis; Race; Research; Sampling; Semantics; Series; Site; Social isolation; Societies; Speech; Structure; Symptoms; Text; Time; Training; Transcript; Uncertainty; Underrepresented Populations; Variant; Work; automated speech recognition; cognitive change; cognitive function; cognitive task; cost; cost effective; deep learning; deep learning model; detection method; ethnic disparity; improved; machine learning classifier; machine learning model; marginalization; neural network architecture; neural network classifier; racial disparity; racial population; response; sound; tool

Abstract This proposal relates to ongoing efforts to develop automated methods for the detection of linguistic manifestations of cognitive changes in Alzheimer’s Disease (AD). These methods have the potential to alleviate the personal and societal burden of AD, by reducing time to diagnosis. Delayed AD diagnosis has adverse effects on care planning and family relationships, and has been estimated to cost the healthcare system close to $8 trillion dollars. Language reflects cognitive status, and contemporary neural network models have been shown to discriminate between transcribed speech from patients with AD and that from healthy controls with promising accuracy. However, most of this work has been conducted in the context of recorded responses to picture description tasks, which are not suitable for repeated, continuous or passive monitoring for linguistic indicators of AD-related decline. In contrast, our recent work has identified the task-agnostic linguistic construct of semantic coherence as a sound basis for machine learning models for detection of language from people with AD in casual conversations, with classification performance in this context exceeding that of Bidirectional Encoder Representations from Transformers (BERT) based classifiers, the state-of-the-art for text-based AD detection. Unfortunately, recent work has shown that automated coherence estimates are vulnerable to bias, with lower estimates for speech from people identifying as black irrespective of diagnosis. Interrogation of coherence- and BERT-based classifiers reveals they dramatically underperform in this group also. For the parent award for this supplement proposal (R01 LM0104056), we will develop methods to debias deep transformer networks to mitigate for the confounding variable of provenance in multi-institutional datasets, culminating in the release of a toolkit to Deconfound Deep Transformer Networks, the DeconDTN suite. In this proposal we will apply these methods to deconfound coherence- and BERT-based AD detection models for the confounding variable of race and/or ethnicity, and evaluate the effects on model performance across groups. In addition, we will evaluate the utility of fine-tuning semantic coherence models and models used to generate automated transcripts on text and speech from the recently-released Corpus of Regional African American Language (CORAAL). We hypothesize that both improvements in these underlying models, and explicit deconfounding for race/ethnicity will reduce the performance differential across groups, resulting in equitable models for language-based AD detection.

抽象的 该建议涉及开发自动化方法以检测语言的努力 阿尔茨海默氏病认知变化的表现（AD）。这些方法有可能减轻 AD的个人和社会负担，通过减少诊断时间。延迟的广告诊断有不良影响 关于护理计划和家庭关系，据估计使医疗保健系统接近8美元 万亿美元。语言反映了认知状况，并且已经显示了当代神经网络模型 区分AD患者的转录语音以及有承诺的健康对照 准确性。但是，这项工作的大部分是在记录对图片的响应的背景下进行的 描述任务，不适合重复，连续或被动监视语言指标 与广告有关的下降。相反，我们最近的工作确定了语义的任务不足语言构造 连贯性作为机器学习模型的合理基础，用于从休闲中从广告中检测语言 对话，在此上下文中的分类性能超过双向编码器 基于变形金刚（BERT）分类器的表示形式，基于文本的AD检测的最新图表。 不幸的是，最近的工作表明，自动相干估计很容易受到偏见的影响，较低 不论诊断如何，从识别为黑人的人的言语估计。询问连贯性和 基于BERT的分类器也揭示了他们在该组中的表现明显不足。为此父母奖 补充提案（R01 LM0104056），我们将开发出Debias Deep Transformer网络的方法 减轻多机构数据集中出处的混杂变量，最终发行 Decondtn Suite的Deconfonfungrom Deep Transformer网络的工具包。在此提案中，我们将应用这些 对种族混杂变量的脱致相干和基于BERT的广告检测模型的方法 和/或种族，并评估跨组对模型绩效的影响。此外，我们将评估 微调语义连贯模型和模型用于在文本和 最近发行的非裔美国人语言语料库（Coraal）的演讲。我们假设 这些基本模型的改进以及对种族/种族的明确变形都将减少 跨组的性能差异，导致基于语言的广告检测的公平模型。