DIGITAL VIDEO ANALYSIS SYSTEM FOR HOME CAGE BEHAVIOR

由于分子生物学的最新进展，对突变小鼠模型进行自动表型分析的需求非常迫切。为了满足这一需求，我们建议开发一种系统，根据家庭笼子中未受干扰的老鼠的视频记录自动评分老鼠的行为。第一阶段的资金在识别、跟踪和评分鼠标行为方面取得了出色的进展，在识别五种姿势方面取得了93.5%的准确率。第二阶段的资金要求将该系统扩展到家庭笼子中老鼠的整个行为，并改进用于识别、跟踪和评分老鼠行为的算法，目的是生产可销售的原型产品。将开发用于分割鼠标、跟踪其位置、识别其姿态和行为的算法。结果的有效性将通过与人类行为评分的比较来确定。通过将原型系统对唐氏综合症、戊二酸血症（敲除模型）和嘌呤代谢基因过表达（GART转基因）小鼠模型的分析结果与其他形式的行为分析的现有数据进行比较，将探索该系统的敏感性。提出的系统将在行为分析和研究人员检测基因改变动物的新表型的能力方面取得重大进展。该系统不仅可以提供小鼠日常行为的定量数据，还可以向研究人员提供它无法识别的新行为的视频片段，从而使研究人员能够看到行为的异常。该系统将足够高效，可以用现成的视频和计算机硬件来实现。为这个项目组建的团队由计算机科学家、数学建模专家和鼠标行为专家组成，他们在第一阶段就合作得很好，为这个项目的成功做好了独特的准备。拟议的商业应用：这项拟议的研究不仅将为神经科学和行为研究界创造巨大的利益，包括医学研究机构、大学、制药公司、医院和诊所，他们正在进行行为研究和分析。而且还能在整个生物学领域找到应用将感兴趣的系统，帮助他们进行诸如基因研究、取消研究、HIV研究等领域的研究。它可以应用于人机界面和监视监控领域。

There is a critical need for automated phenotyping of mutant mouse models that have grown out of recent advances in molecular biology. We propose to meet this need by developing a system for automated scoring of mouse behavior from video records of undisturbed mice in the home cage. Phase I funding resulted in excellent progress on identifying, tracking, and scoring mouse behavior, resulting in 93.5% accuracy in identifying five postures. Phase II funding is requested to extend this system to the entire behavioral repertoire of the mouse in the home cage and refine the algorithms used to identify, track and score mouse behavior with the aim of producing a marketable prototype product. Algorithms for segmenting mouse, tracking its position, identify its posture and behaviors will be developed. The validity of the results will be determined using comparison with human scoring of behavior. The sensitivity of the system will be explored by comparing the results of the prototype systems' analysis of mouse models of Down syndrome, glutaric acidemia (knock- out model) and over expression of purine metabolism genes (GART transgemic) with existing data from other forms of behavioral analysis. The proposed system will produce a major advance in the analysis of behavior and in the ability of investigators to detect novel phenotypes in genetically altered animals. The system will not only provide quantitative data on mouse behavior throughout the daily cycle, but will present the investigator with video clips of novel behaviors that it cannot identify, thus allowing the investigator to see abnormalities in behavior. The system will be efficient enough to be realized with readily available video and computer hardware. The team assembled for this project consists of computer scientists, experts in mathematical modeling, and an expert in mouse behavior has worked well together in Phase 1 and is uniquely prepared to succeed with this project. PROPOSED COMMERCIAL APPLICATION: This proposed research will not only create enormous benefits for the neuroscience and behavior research community including medical research institutes, universities, pharmaceutical companies, and hospitals and clinics who are conducting behavior study and analysis. But also find applications in the entire area of biology will be interested in the system to help them conduct research in such areas as genetic research, cancel research, HIV research, etc. It can be applied to human-computer interface and surveillance and monitoring areas.