Clear Volume Imaging with Machine Learning: a novel tool to identify brain-wide neuronal ensembles of opioid relapse in rat models

机器学习清晰体积成像：一种识别大鼠模型中阿片类药物复发的全脑神经元群的新工具

• 批准号: 10241671
• 负责人: RONG CHEN
• 金额: $ 22.97万
• 依托单位: UNIVERSITY OF MARYLAND BALTIMORE
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-05-15 至 2023-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10241671
• 关键词: 3-Dimensional; Abstinence; Address; Atlases; Award; Basic Science; Brain; Brain imaging; Buprenorphine; Cell Count; Cells; Computer software; Custom; Data; Data Set; FDA approved; Fluorescence Microscopy; Goals; Image; Immediate-Early Genes; Immunohistochemistry; Label; Light; Machine Learning; Maps; Methadone; Methods; Modeling; Mus; Naltrexone; Neurons; Opioid; Outcome; Oxycodone; Patients; Positioning Attribute; Procedures; Rattus; Relapse; Research; Research Personnel; Resolution; Risk; Role; Self Administration; Signal Transduction; System; Systems Analysis; Techniques; Transcranial magnetic stimulation; activity marker; addiction; adverse outcome; analytical method; base; bioimaging; brain tissue; computer framework; data analysis pipeline; drug relapse; experience; experimental group; graphical user interface; high throughput analysis; improved; machine learning method; microscopic imaging; motivated behavior; neuroimaging; neuronal patterning; novel; open data; open source; opioid epidemic; opioid mortality; opioid user; overdose risk; programs; relating to nervous system; response; tool; usability

This project is in response to PA-18-437 “Cutting-Edge Basic Research Awards (CEBRA)”. Over the two past decades, there has been a large increase in the abuse of prescription and illegal opioids; this increase coincides with increases in opioid-related deaths. A critical challenge is the occurrence of relapse in treated patients, especially given that relapse episodes carry a risk of overdose. There is a need to improve our understanding of the brain mechanisms of opioid relapse, which hopefully will result in the identification of targeted circuitry-based treatments. We propose to develop a high-throughput computation system termed Clear Volume Analysis with Machine Learning (CVA-ML). We will combine CVA-ML with a rat-optimized version of the whole brain immunostaining and clearing method iDISCO+ and a new rat model of opioid relapse after voluntary abstinence to identify brain- wide neuronal ensembles of opioid relapse. We recently adapted the iDISCO+ method to intact rat brains and developed experimental methods for Fos immunostaining, brain clearing, and light sheet fluorescence microscopy imaging. However, incorporation of the iDISCO+ method to large scale rat studies is currently limited by (1) lack of ABA-CCF-comparable high-resolution 3D rat brain atlas that allows for high-resolution registration of the activity signal in the 3D space, and (2) lack of an automated data analysis pipeline. In Aim 1, we propose to develop a data analysis pipeline that will take light sheet fluorescence microscopy- generated rat brain images and automatically register them into a custom-made 3D rat brain atlas encompassing a converted Paxinos and Watson rat’s brain atlas. As part of Aim 1, we also propose to develop machine-learning methods to identify and analyze the whole brain Fos signals in 3D space. In Aim 2, we propose to use the methods we developed in Aim 1 to identify brain-wide patterns of neuronal activity (‘neural ensembles’) that encode opioid relapse after voluntary abstinence induced by imposing adverse consequences (electric barrier) that results in long-term cessation of opioid (oxycodone) self-administration. Our proposal addresses the goal of PA-18-437: “to develop, and/or adapt, revolutionary techniques or methods for addiction research.” The anticipated outcomes of our proposal are an open-source software package to automatically analyze iDISCO+ data of rat brains, and a rat whole brain activity map for opioid relapse, assessed using a new rat model. The publicly available software will be easy to modify and can be used by investigators to identify brain-wide neuronal ensembles underlying drug relapse and other motivated behaviors in rats.

该项目是为了响应 PA-18-437“尖端基础研究奖（CEBRA）”。过去两届 几十年来，处方药和非法阿片类药物的滥用大幅增加；这一增长恰逢 随着阿片类药物相关死亡人数的增加。一个关键的挑战是接受治疗的患者会出现复发， 特别是考虑到复发发作有服用过量的风险。我们需要提高认识 阿片类药物复发的大脑机制，有望导致基于靶向电路的识别 治疗。 我们建议开发一种高吞吐量计算系统，称为 Clear Volume Analysis with Machine 学习（CVA-ML）。我们将 CVA-ML 与大鼠优化版本的全脑免疫染色相结合 和清除方法 iDISCO+ 和自愿戒断后阿片类药物复发的新型大鼠模型，以识别脑- 阿片类药物复发的广泛神经元群。我们最近将 iDISCO+ 方法应用于完整的大鼠大脑并 开发了 Fos 免疫染色、脑透明和光片荧光的实验方法 显微镜成像。然而，将 iDISCO+ 方法纳入大规模大鼠研究目前还很有限 (1) 缺乏可与 ABA-CCF 相媲美的高分辨率 3D 大鼠脑图谱来进行高分辨率配准 3D 空间中的活动信号，以及 (2) 缺乏自动化数据分析管道。 在目标 1 中，我们建议开发一种采用光片荧光显微镜的数据分析流程 - 生成大鼠大脑图像并自动将其注册到定制的 3D 大鼠大脑图谱中，其中包括 转换后的 Paxinos 和 Watson 大鼠大脑图谱。作为目标 1 的一部分，我们还建议开发机器学习 在 3D 空间中识别和分析全脑 Fos 信号的方法。在目标 2 中，我们建议使用 我们在目标 1 中开发的方法来识别全脑神经元活动模式（“神经集合”）， 编码因施加不良后果（电屏障）而自愿戒断后阿片类药物复发 这导致长期停止阿片类药物（羟考酮）的自我给药。 我们的提案解决了 PA-18-437 的目标：“开发和/或采用革命性技术或方法 用于成瘾研究。”我们提案的预期成果是一个开源软件包 自动分析大鼠大脑的 iDISCO+ 数据，并评估阿片类药物复发的大鼠全脑活动图 使用新的大鼠模型。公开可用的软件将易于修改并且可供调查人员使用 识别大鼠药物复发和其他动机行为背后的全脑神经元群。