Integrating Clinical Infrared and Raman Spectroscopy with digital pathology and AI: CLIRPath-AI

诊断任何疾病，特别是各种形式的癌症的一个关键特征是通过活检获得的关键信息。活组织检查包括从患者身上取出一小部分组织样本或几个细胞，供病理学家在光学显微镜下进行检查。目前的做法是用混合染料对样本进行染色，以帮助在图像中获得一些对比度，这有助于病理学家看到细胞。通常，基于对样本和其他相关医学信息的这种目视检查，作出诊断。然而，这一过程远不理想，因为它依赖于相关临床医生的专业知识，并且容易受到观察者之间的内部误差的影响。(换句话说，这一过程并不准确，取决于临床医生的意见，而临床医生的意见可能不同)。近年来，数字病理学和人工智能(AI)领域取得了一些进展。在这里，活检切片的高分辨率照片被拍摄下来，并由计算机算法进行检查，该算法帮助病理学家做出诊断。然而，仅从光谱的可见区域分析数据严重限制了所获得的图像的信息含量。最近，一些概念验证研究表明，分子光谱技术，如红外和拉曼光谱技术，能够根据细胞内包含的内在化学物质区分患病和未患病的细胞和组织。(这些光谱区域的带宽是可见光的40倍，因此包含的信息量是可见光的40倍。)英国在与数字病理学和人工智能相关的发展方面走在了前列，后者由产业战略挑战基金资助的五个新技术中心加强。此外，部分得益于EPSRC资助的网络(CLIRSPEC)，英国在生物医学红外和拉曼光谱领域也处于世界领先地位。然而，目前数字病理/人工智能和生物医学红外/拉曼这两个社区是分开的，没有相互作用。因此，在一个领域取得的进展没有转化为另一个领域。在这两个研究领域，如果这项技术要从概念验证阶段进入翻译阶段并进入临床环境，有许多障碍需要克服。学术界认为，如果我们集中资源，引入工业和临床合作伙伴，共同解决这些通用问题，我们就更有可能克服这些障碍。这项申请是为了支持这样一个合作伙伴网络的资金，该网络将在未来四年内在这些独立的社区之间发展动态和协同作用，以实现造福患者的具体目标。

A key feature of the diagnosis of any disease, but particularly various forms of cancer, is the critical information obtained through a biopsy. A biopsy involves the removal of a small sample of tissue, or a few cells, from the patient for examination by a pathologist looking down an optical microscope. In current practice is that the sample is stained with a combination of dyes to help gain some contrast in the image which helps the pathologist see the cells. Generally, based upon this visual inspection of the sample and other relevant medical information, a diagnosis is made. This process, however, is far from ideal since it relies on the expertise of the clinician concerned and is subject to intra in inter observer error. (In other words the process is not exact and depends upon the opinion of the clinicians which may differ). Recently a number of developments have been made in the field of Digital Pathology and Artificial Intelligence (AI). This is where a high resolution photograph of the biopsy slide is taken and examined by a computer algorithm which helps the pathologist make a diagnosis. However analysing the data from just the visible region of the spectrum severely restricts information content of the images obtained. Recently a number of proof of concept studies have shown that molecular spectroscopic techniques such as infrared and Raman are capable of distinguishing diseased from non-diseased cells and tissue based upon the inherent chemistry contained within the cells. (These regions of the spectrum have 40 times the bandwidth of the visible and therefore contain 40 times the amount of information.) The UK is at the forefront in developments associated with both Digital Pathology and AI, the latter augmented by five new technology centres funded by the Industrial Strategy Challenge Fund. In addition, partly due to an EPSRC funded network (CLIRSPEC) the UK is also world leading in the field biomedical infrared and Raman spectroscopy. At present however the Digital pathology/AI and biomedical infrared/Raman these two communities are separate and are not interacting. As a result therefore, the advances made in one area are not being translated to another. In both areas of research there are many hurdles that need to be overcome if this technology is to move from the proof of concept stage through the translational stage and into the clinical setting. It is the belief of the academic community that we are much more likely to overcome these hurdles if we pool our resources, bring in both industrial and clinical partners and work on these generic problems together. This application is for funding to support such a network of partners that will develop dynamic and synergistic interaction between these separate communities for the next four years, for the specific aim of benefiting patients.