ABI Innovation: Modeling the Drosophila Brain with Single-neuron Resolution using Computer Vision Methods

该项目的资助是为了利用计算机视觉方法对神经元的形态以及它们如何在中枢神经系统(CNS)中发展和形成突触进行建模。通过利用果蝇运动神经元的形态刻板印象和无与伦比的单细胞分辨率，该项目将在果蝇中枢神经系统中产生第一张运动神经元图谱，作为一个案例研究，并朝着建立整个大脑的模型迈出了一步。第一张完整的全脑地图是线虫的。它的优点，到目前为止还没有受到挑战，是分辨率下降到单个视觉识别神经元的水平。由于单个神经元的连接最终决定了大脑的功能，在其他更复杂的模型生物中创建这一水平的神经地图对于促进我们对大脑发育的理解至关重要。这个项目的直接目标是在单个神经元水平上在健康的果蝇大脑中创建一个确定的运动神经元群体的神经图谱。它的目的是为任何想要在活体中“导航”模型大脑的人提供资源。为实现这一目标，本项目包括以下任务。(A)使用基于图像的运动神经元自动分类对单个神经元的形态进行建模。为了达到这一目标，将开发一个新的框架来对神经元进行三维划分？形态分室，即胞体、轴突和树突。(B)从分割的体积中对神经元形态进行分类。该方法包括一种新的可能性半监督聚类框架，该框架足够健壮，能够区分不同的神经元形态，并处理类内差异和类间相似性。(C)通过推断个别神经元建模的知识，并构建一个运动电路原型，识别和追踪部分果蝇中枢神经系统中特定形态的运动神经元。所有发现、方法、开发的算法(开放源码)、出版物和数据一旦可供公众使用和准备就绪，将立即通过该项目的网站传播：http://web.mac.com/gavriil/Gavriil_Tsechpenakis/neurovision.这项工作立即扩展到：(A)通过将所有(约400个)人工识别的运动神经元包括在单个果蝇中枢神经系统半节段中，完成运动电路模型；(B)识别更多类型的神经元，以扩大这项工作的发现，并对大脑中更多类型的神经元电路进行建模。长期目标是创建一个完整的大脑模型，以便对突变后的大脑发育进行全面分析，并评估药物治疗、疾病或衰老和压力导致的大脑连接模式的变化。

The project is funded to model the morphology of neurons and how they develop and form synapses in the Central Nervous System (CNS) using computer vision approaches. By exploiting the morphological stereotypes and unparalleled single-cell resolution of Drosophila motor neurons, this project will yield the first motor neuron map in the Drosophila CNS, serving as a case study and a step towards modeling the entire brain. The first complete map of a whole brain is that of C. elegans. Its virtue, unchallenged thus far, is resolution down to the level of single visually identified neurons. Because the connectivity of individual neurons ultimately determines brain function, the creation of neural maps at this level in other more complex model organisms is critical for advancing our understanding of brain development. The immediate goal of this project is to create a neural map of a defined population of motor neurons in a healthy Drosophila brain at the single-neuron level. It is intended to be a resource for anyone who wants to "navigate" a model brain in vivo. To achieve this goal, this project includes the following tasks. (a) Modeling the individual neuron morphology with automated image-based motor neuron classification. To meet this objective, a new framework will be developed for three-dimensional partitioning of the neurons? morphological compartments, namely soma (cell body), axon and dendrites. (b) Classifying the neuron morphology from the partitioned volumes. The approach consists of a novel possibilistic semi-supervised clustering framework, robust enough to discriminate between different neuron morphologies, and to handle the intra-class variability and inter-class similarities. (c) Identifying and tracing morphology-specific motor neurons in part of the Drosophila CNS by inferring knowledge of individual neuron modeling, and constructing a prototype motor circuit.All findings, methods, developed algorithms (open source codes), publications and data will be disseminated through the project's website, as soon as they become available and ready for the public: http://web.mac.com/gavriil/Gavriil_Tsechpenakis/neurovision. This work has immediate extensions to (a) the completion of a model motor circuit, by including all (approximately 400) manually identified motor neurons within a single Drosophila CNS hemisegment; (b) the identification of more types of neurons, for scaling up the findings of this work and modeling more types of neuron circuitries in the brain. The long-term goal is to create a complete model brain that will allow for comprehensive analysis of brain development after mutation, and the assessment of changes in brain connectivity patterns resulting from drug treatments, disease, or aging and stress.