Real time colon histopathology by infrared spectroscopic imaging

摘要 结直肠癌(CRC)是美国主要的死亡原因之一。积极筛查和早期干预 高风险癌症的治疗可以带来良好的结果；然而，快速提供适当的患者护理的瓶颈 组织学评估的时间较长，在预测疾病严重程度方面缺乏精确性。 组织学中流行的形态评估是有用的，但资源密集，预测性不够。 补充传统病理学的分子技术正在出现，但往往需要更多的努力和 时间，但与组织学评估不是特别一致。在这里，我们寻求开发一种技术 它测量组织的化学含量，不需要试剂，与临床完全兼容 并利用现代人工智能(AI)技术提供实时组织学评估。 我们方法的基础是一种新的红外光谱成像系统的设计，其速度比 任何报道，提供更高的空间和光谱质量，并使用固定焦距的固体沉浸镜头 镜片的密封面，供受过最低程度训练的人使用。与仪器结合使用， 我们开发了人工智能算法来测量组织的化学成分，并用它来提供(A)常规的 不使用染料(“不锈钢染色”)的病理图像，以及(B)基于 分子数据，可以提供类似于致命癌症图像的补充成分、疾病和风险 传统病理学。该仪器将由实验室技术人员使用，不需要准备薄片 从切除的组织切片，并将在几分钟内提供信息。使用来自人类患者的初步数据 在来自8个组织微阵列(TMA)和30个手术切除的850多个组织微阵列(TMA)样本上，我们验证了 提供完整的组织学和疾病等级评估的技术。将使用统计方法来 严格评估结果，量化里程碑指导整个方法。然后我们将结果翻译成 到新鲜的组织块，在从体内提取组织几分钟后提供组织学。最后，我们使用 可从组织中获得详细的肿瘤和微环境信息，将患者划分为高危人群 和“低风险”组。快速组织学评估的可用性可以帮助防止延迟提供护理， 提供术中评估，并在筛查后的形态评估中添加更多信息， 在结直肠癌和其他癌症病理中得到广泛应用。

Abstract Colorectal cancer (CRC) is one of the leading causes of death in the US. Active screening and early intervention in risky cancers can lead to good outcomes; however, a bottleneck in rapidly delivering appropriate patient care is the long time period for histologic assessment and lack of precision in predicting disease severity. Morphological assessments prevalent in histology are useful but resource intensive and not predictive enough. Molecular techniques to complement traditional pathology are emerging but often require much more effort and time, without being especially compatible with histologic assessments. Here, we seek to develop a technology that measures the chemical content of tissues, does not require reagents, is entirely compatible with clinical workflows and leverages modern artificial intelligence (AI) techniques to provide real-time histologic assessment. The foundation of our approach is a new design for an infrared spectroscopic imaging system that is faster than any reported, offers a higher spatial and spectral quality and uses a solid immersion lens with a fixed focus at the sealed surface of the lens to enable use by a minimally trained person. In conjunction with the instrument, we develop AI algorithms that measure the chemical content of tissue and use it to provide (a) conventional pathology images without the use of dyes (“stainless staining”), and (b) histologic assessment based on molecular data, which can provide complementary composition, disease and risk of lethal cancer images akin to conventional pathology. The instrument will be usable by laboratory technicians, without the need to prepare thin sections from excised tissue and will provide information in minutes. Using preliminary data from human patients on over 850 tissue microarray (TMA) samples from 8 TMAs and 30 surgical resections, we validate the use of technology in providing complete histologic and disease grade assessment. Statistical methods will be used to assess the results rigorously and quantitative milestones guide the entire approach. We then translate the results to fresh tissue chunks, providing histology minutes after tissue is extracted from the body. Finally, we use the detailed tumor and microenvironment information available from the tissue to segment patients into a “high risk” and “low risk” group. The availability of rapid histologic assessment can help prevent delays in providing care, provide intraoperative assessment, and add more information to morphologic assessments following screening, enabling a wide use in CRC and other cancer pathologies.