Vector Quantization for Image Pattern Recognition

描述（由申请人提供）： 该第一阶段SBIR应用解决了病理学家和临床医生在手动检查显微镜载玻片时面临的日益重大的挑战。显微镜检查的缺点是劳动密集型、主观、昂贵并且受限于需要物理接近感兴趣的载玻片样本。自动化显微镜检查的障碍一直是无法有效地在高分辨率下对整个显微镜样品进行扫描。Aperio开发了ScanScope（R），这是一种新型的显微镜载玻片扫描仪，首次实现了以高分辨率快速创建虚拟显微镜载玻片。虚拟载玻片为使用自动模式识别进行自动化显微镜检查奠定了基础。本研究的目的是适应和优化Aperio的现有的和新的矢量量化（VQ）的算法，在虚拟幻灯片的自动模式识别的问题。VQ是使用词汇表对比特流进行编码的通用数学技术。主要目的是证明使用VQ模式识别在一个实际的和良好的特点的应用程序的可行性：自动发现几乎所有的微转移簇细胞学标本。该研究首次尝试使用VQ在虚拟幻灯片中自动进行模式识别。

DESCRIPTION (provided by applicant): This Phase-I SBIR application addresses the increasingly significant challenges faced by pathologists and clinicians in manually inspecting microscope slides. Microscopic inspection suffers from being labor-intensive, subjective, expensive and limited by the need for physical access to the glass slide specimen of interest. The obstacle to automated microscopic inspection has been the inability to efficiently digitize entire microscope specimens at high resolutions. Aperio has developed the ScanScope (R), a novel microscope slide scanner that makes it practical - for the first time - to rapidly create virtual microscope slides at high resolutions. Virtual slides set the stage for automating microscopic inspection using automated pattern recognition. This research aims to adapt and optimize Aperio's existing and novel algorithms for vector quantization (VQ) to the problem of automatic pattern recognition in virtual slides. VQ is a general mathematical technique for encoding bitstreams using a vocabulary. The primary aim is to demonstrate the feasibility of using VQ for pattern recognition in a practical and well-characterized application: automatically finding virtually all micrometastasis clusters in cytology specimens. This proposed research represents a first attempt to automate pattern recognition in virtual slides using VQ.