MitProfiler: A Cloud-based AI Solution for Profiling of Mitotic Figures in Digitised Images of Cancerous Tissue Slides

有丝分裂象是指一个细胞分裂成两个子细胞时的细胞物质。它们构成了组织病理学的一个重要特征，组织病理学是在细胞水平上研究疾病的。除其他用途外，有丝分裂图在许多常见癌症的分级系统中占有突出地位，例如乳腺癌。目前检测和计数有丝分裂数字的“黄金标准”主要是基于专家组织病理学家对粘在玻片上的极薄（只有几微米厚）的可疑组织标本的视觉显微镜评估。几十年来，这种做法或多或少保持不变，不可避免地导致主观和可变的诊断，从而导致不同的患者管理。手工过程由于其非常主观的性质，很容易出错，也很耗时。此外，目前可以手工记录的数据仅限于单位面积的有丝分裂数。最近由诊断组织病理学实验室和组织研究实验室采用的数字切片扫描仪标志着病理学实践的新革命，我们当地的考文垂和沃里克郡NHS信托是英国第一个使用组织切片数字扫描图像进行常规诊断的机构。数字切片扫描仪从组织切片中产生数十亿像素的全切片图像（WSI），为开发智能计算机算法提供了机会，可以处理和分析图像数据，为诊断辅助和改善患者管理提供更客观、可靠和准确的结果，并为研究目的提供精确和可重复的测量。**这个英国-韩国合作研发项目**旨在生产最近屡获殊荣的基于人工智能（AI）的计算机算法，该算法由我们的团队生产，可以自动识别和计数有丝分裂数字，并以许多以前不可能的新颖方式整理数据。我们将利用最近采用的数码切片扫描仪，生成组织切片的数码扫描图像，以供日常使用。本项目在云中开发和部署的产品将使有丝分裂数字的测量更加准确和一致。此外，通过提供每例有丝分裂图的平均计数和有丝分裂热点区域分布等分析数据，我们的工具将改进提供给组织病理学家和肿瘤学家的数据，从而帮助改善癌症患者的管理。它还将通过节省执行手动计数所花费的时间来提高效率。

**Mitotic figures** refer to cellular material seen as a cell divides into two daughter cells. They form an important feature of histopathology, which is the study of disease at the cellular level. Amongst other uses, mitotic figures feature prominently in the grading systems for many common cancers, such as breast cancer for example.The current '**_gold standard_**' of detecting and counting mitotic figures is largely based on an expert histopathologist's visual microscopic assessment of an extremely thin (only a few micrometers thick) section of the suspicious tissue specimen glued to a glass slide. This practice has remained more or less the same for several decades, and inevitably results in subjective and variable diagnosis and consequently variable patient management. The manual process is prone to error due to its very subjective nature and is also time consuming. Moreover, the data which can be recorded manually at present is limited to number of mitoses per unit area.Recent uptake of **digital slide scanners** by diagnostic histopathology laboratories and tissue research laboratories marks a new revolution in pathology practice, with our local Coventry & Warwickshire NHS trust being the first one in the UK to use digitally scanned images of tissue slides for routine diagnostics. Digital slide scanners produce multi-gigapixel whole-slide image (WSI) from tissue slides, offering the opportunity to develop intelligent computer algorithms that can process and analyse the image data, produce more objective, reliable and accurate results for diagnostic assistance and improved patient management and provide precise and reproducible measurements for research purposes.**This UK-Korea collaborative R&D project** aims to productise a recent award-winning Artificial Intelligence (AI) based computer algorithm produced in our group that can recognise and count mitotic figures automatically and collate the data in a number of novel ways which has not been possible before. We will capitalise on the recent uptake of digital slide scanners to generate digitally scanned images of tissue slides for routine use. The product developed and deployed in the cloud in this project will make the measurements of mitotic figures more accurate and consistent. In addition, by providing profiling data such as average count of mitotic figures per case and distribution of mitotic hotspot regions, our tool will improve the data provided to histopathologists and oncologists and thereby help improve the management of cancer patients. It will also improve efficiency by saving the time taken to perform manual counts.