Computational Methods for Capturing and Analyzing Personalized Genomic and Medical Data

我的研究重点是计算机科学方法的发展，以帮助解决生物学和医学中的问题。虽然我的研究应用于医学和基因组学的许多领域，但贯穿我所有工作的统一主线是开发新的计算方法，坚定地在计算机科学领域内，以解决这些问题。在接下来的五年里，我想把重点放在开发三种计算方法上，这些方法可以结合起来，更准确、更快速地捕获和分析医疗数据。在所有这些情况下，主要的努力将是开发新的计算方法和方法，而不是在医疗环境中部署现有技术，所开发的许多方法将在医疗环境之外得到推广。*我的赠款目标1将是开发移动设备和人机界面技术，以从患者访问中捕获数据。病人和医生之间的核心互动包括医生进行观察，(在电脑或纸上)做笔记，最后写(或口述)一份报告。这份报告主要记录了与患者的充分互动以及医生目测询问的无数体征和症状(不是所有的体征和症状都可能被记录在笔记中，特别是如果不是异常的话)。我们建议开发下一代用户界面，使用临床医生(相机和麦克风)上的可穿戴技术，结合移动设备(平板电脑)，集成到临床工作流程中，并能够捕获患者交互的全部视觉和音频频谱。*在赠款的目标2中，我们将致力于开发ML方法，以从捕获的数据中识别概念-音频、视频和文本。我们将利用生物医学本体来帮助提高方法的准确性，训练模型利用文本中的邻近性和生物医学本体空间中的邻近性来帮助训练分类器。我们将致力于将这种方法集成到现有的语音到文本工具中，以帮助改进音频处理，并将这些方法应用于患者检查的音频记录，以测试其准确性。*最后，在提案的目标3中，我们将致力于记录患者数据的数据可视化，目的是为临床医生提供一种易于理解的患者随时间推移的症状摘要，以及允许在患者和其他具有类似疾病(用于鉴别诊断)和具有相同疾病(以了解疾病的变异性和预后)的其他患者之间进行快速比较。

My research centers on the development of Computer Science methods to help solve problems in Biology and Medicine. While my research has application to many areas of medicine and genomics, the unifying thread through all of my work is the development of novel computational methods, firmly within the domain of Computer Science, to solve these problems. In the next five years I want to focus on the development of three computational approaches that can be combined for more accurate and rapid capture and analysis of medical data. In all of these cases the effort will be primarily on developing novel computational methodology and approaches, rather than deployment of existing technologies in the medical setting, and many of the methods developed will be generalizable outside of the medical setting.******Aim 1 of my grant will be on the development of mobile devices and HCI techniques for the capture of data from a patient visit. The core interaction between a patient and their doctor involved the doctor making observations, taking notes (on computer or paper), and finally writing (or dictating) a report. This report serves as the main record of the full interaction with the patient and myriad of signs and symptoms visually interrogated by the doctor (not all of which may be recorded in the notes, especially if not abnormal). We propose to develop a next generation of user interfaces, using wearable technology on the clinician (camera and microphone), combined with mobile devices (tablets), integrated into the clinical workflow, and capable of capturing the full visual and audio spectrum of the patient interaction. ******In Aim 2 of the grant we will work to develop ML methodology to identify concepts from the captured data – audio, video, and text. We will utilize biomedical ontologies to help improve the accuracy of the methods, training models that utilize proximity in text and proximity in biomedical ontology space to help train classifiers. We will work on integrating this approach into existing speech-to-text tools to help improve audio processing, and apply these approaches to audio recording of patient exams to test their accuracy.******Finally, in Aim 3 of the proposal we will work on data visualization of recorded patient data, with the aim of providing the clinician with a easy-to-understand summary of patient symptoms over time, as well as to allow for quick comparison between a patient and others with similar disorders (for differential diagnosis) and same disorder (to understand disease variability and prognosis).