Domain-Knowledge Informed Deep Learning for Early Detection of Pancreatic Cancer

基于领域知识的深度学习用于胰腺癌的早期检测

• 批准号: 10317236
• 负责人: Chin Hur
• 金额: $ 17.69万
• 依托单位: COLUMBIA UNIVERSITY HEALTH SCIENCES
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-07-28 至 2023-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10317236
• 关键词: Academic Medical Centers; Algorithms; Artificial Intelligence; Attention; Cancer Etiology; Categories; Cessation of life; Characteristics; Chronology; Clinical; Complex; Data; Data Analytics; Data Set; Diagnosis; Early Diagnosis; Electronic Health Record; Goals; Graph; Grouping; Healthcare; Hospitals; Human; Image; Individual; Knowledge; Laboratories; Learning; Life; Literature; Machine Learning; Malignant Neoplasms; Malignant neoplasm of pancreas; Measurement; Medical; Methodology; Methods; Mining; Modality; Modeling; Morbidity - disease rate; Nature; Pancreatic Ductal Adenocarcinoma; Patient Care; Patients; Pharmaceutical Preparations; Physicians; Predictive Factor; Procedures; Protocols documentation; Provider; Records; Reporting; Research; Risk Factors; Savings; Scheme; Series; Signs and Symptoms; Statistical Study; Structure; Survival Rate; Techniques; Testing; Text; Time; Training; Visit; associated symptom; base; cancer diagnosis; clinical decision-making; data format; data quality; deep learning; deep learning algorithm; demographics; design; direct application; effective therapy; electronic structure; feature extraction; health data; improved; innovation; knowledge base; mortality; multimodal data; multimodality; neural network; novel; novel strategies; pancreatic ductal adenocarcinoma model; predictive modeling; programs; relating to nervous system; risk prediction; screening; success

PROJECT SUMMARY The goal of this project is to leverage deep-learning algorithms on Electronic Health Records (EHRs) to improve early detection of pancreatic ductal adenocarcinoma (PDAC), a malignancy with high mortality and morbidity. Although numerous risk factors have been identified, PDAC is most often found in later stages when effective treatments are not feasible or their survival benefit is limited. In this R21, we aim to develop novel structured methodologies for systematically incorporating feature grouping strategy from expert domain knowledge into the training procedure of deep-learning algorithms for improving PDAC diagnosis. The overarching hypothesis for this study is that the groups of highly correlated variables will combine to form superior and interpretable predictors compared to individual clinical variables (current proposal). Furthermore, these new predictors represented by the group of related data will be useful for other downstream tasks such as risk factor identification via causal discovery (future research). The proposed research presents an innovative approach towards unifying human and artificial intelligence, using explainable algorithms to build interpretable prediction models, in contrast to conventional deep-learning algorithms which are non-traceable by humans due to their black-box nature. An optimal strategy for creating composite (grouped) variables should maximize both predictive power as well as human-interpretability. We will thus explore a variety of grouping strategies relying heavily on human-expert knowledge (e.g. clinical workflows) as well as auto-correlation tests. An effective grouping strategy will allow our prediction model to learn the relative importance of both individual measurements as well as interpretable groups of measurements in predicting PDAC. Examples in the literature show that such grouped predictors often have superior predictive power compared to their individual components, which can be attributed to the mutual information shared within the group. Different types of explainable (attention) neural networks may also be applied depending on the group characteristics to further improve interpretability as well as prediction accuracy. We believe that similar methodologies applied to predictive modeling in healthcare data have the potential to fundamentally advance clinical decision making with improved model interpretability. The success of this proposal will be leveraged in a larger ongoing project which aims to establish new causal relationships between various risk factors associated with PDAC. This involves an advanced graph-based approach for building interpretable models. Our direct application of causal discoveries in the future research will be a program for collecting patient-generated health data (PGHD) for PDAC early diagnosis.

项目概要 该项目的目标是利用电子健康记录 (EHR) 的深度学习算法 改善胰腺导管腺癌 (PDAC) 的早期发现，这是一种死亡率很高的恶性肿瘤， 发病率。尽管已经确定了许多风险因素，但 PDAC 最常见于晚期阶段： 有效的治疗不可行或者其生存获益有限。在这个 R21 中，我们的目标是开发新颖的 用于系统地整合专家领域的特征分组策略的结构化方法 将知识融入深度学习算法的训练过程中，以改善 PDAC 诊断。这 这项研究的总体假设是，高度相关的变量组将结合起来形成 与个体临床变量相比，优越且可解释的预测因子（当前建议）。 此外，由相关数据组表示的这些新预测变量将有助于其他 下游任务，例如通过因果发现识别风险因素（未来研究）。 拟议的研究提出了一种统一人类和人工智能的创新方法， 与传统的深度学习相比，使用可解释的算法来构建可解释的预测模型 由于其黑盒性质，人类无法追踪算法。 创建复合（分组）变量的最佳策略应该同时最大化预测能力 作为人类的可解释性。因此，我们将探索严重依赖人类专家的各种分组策略 知识（例如临床工作流程）以及自相关测试。有效的分组策略将允许 我们的预测模型可以了解各个测量值以及可解释性的相对重要性 预测 PDAC 的测量组。文献中的例子表明，这种分组预测变量 与它们的单个组件相比，通常具有卓越的预测能力，这可以归因于 群体内共享的相互信息。不同类型的可解释（注意力）神经网络也可能 根据群体特征进行应用，以进一步提高可解释性和预测 准确性。 我们相信，应用于医疗数据预测建模的类似方法有可能 通过改进模型的可解释性从根本上推进临床决策。此次活动的成功 该提案将在一个更大的正在进行的项目中得到利用，该项目旨在建立新的因果关系 与 PDAC 相关的各种危险因素之间的关系。这涉及到一种先进的基于图形的方法 构建可解释的模型。我们在未来的研究中直接应用因果发现将是 收集患者生成的健康数据 (PGHD) 以进行 PDAC 早期诊断的计划。