Digital image analysis in pathology: Benefits and obligation
Digital image analysis in pathology: Benefits and obligation
复制标题
DOI：
10.3233/acp-2011-0033
复制
发表时间：
2012-01-01
影响因子：
3.2
通讯作者：
Herlin, Paulette
中科院分区：
文献类型：
--
作者：
Laurinavicius, Arvydas;Laurinaviciene, Aida;Herlin, Paulette
病理学最近进入了个性化医疗时代。这为基于组织的诊断的准确性和精确性带来了新的期望，特别是当需要定量组织学特征和生物标志物表达时。虽然多年来传统的病理诊断一直被视为基础事实，但这一概念在当代基于组织的生物标志物研究和临床应用中已不再充分。病理学的另一个重大变化是虚拟显微镜技术的进步，使显微镜载玻片数字化，并提出了新的机会，数字图像分析。计算机视觉提供了增加容量（自动化）和精度（再现性）的直接好处，但不一定是分析的准确性。为了实现准确性的好处，病理学家将不得不承担图像分析算法的验证和质量保证的义务。需要参考值来测量和控制精度。虽然病理学家的共识值通常用于验证这些工具，我们认为，地面真相可以最好地实现体视学方法，估计相同的变量作为一个算法是打算做的。正确采用新技术将需要一种新的病理学定量思维。为了看到疾病的完整和清晰的图像，病理学家需要学习使用他们的模拟眼和数字眼。
Pathology has recently entered the era of personalized medicine. This brings new expectations for the accuracy and precision of tissue-based diagnosis, in particular, when quantification of histologic features and biomarker expression is required. While for many years traditional pathologic diagnosis has been regarded as ground truth, this concept is no longer sufficient in contemporary tissue-based biomarker research and clinical use. Another major change in pathology is brought by the advancement of virtual microscopy technology enabling digitization of microscopy slides and presenting new opportunities for digital image analysis. Computerized vision provides an immediate benefit of increased capacity (automation) and precision (reproducibility), but not necessarily the accuracy of the analysis. To achieve the benefit of accuracy, pathologists will have to assume an obligation of validation and quality assurance of the image analysis algorithms. Reference values are needed to measure and control the accuracy. Although pathologists' consensus values are commonly used to validate these tools, we argue that the ground truth can be best achieved by stereology methods, estimating the same variable as an algorithm is intended to do. Proper adoption of the new technology will require a new quantitative mentality in pathology. In order to see a complete and sharp picture of a disease, pathologists will need to learn to use both their analogue and digital eyes.