癌症，即恶性肿瘤，是世界上最常见的死亡原因之一，其中乳腺癌的发病率在全国乃至全球范围内均为首位。肿瘤细胞的恶性增殖及侵袭性是导致患者死亡的主要原因。肿瘤作为一种由海量细胞构成的复杂系统，其组织结构随着肿瘤的恶化发展而变化，展现出异常的物理特征和行为。组织的结构变化是其功能变化的基础，理解肿瘤细胞的群体结构特征不仅有利于我们理解肿瘤异常功能，也是识别乳腺癌肿瘤的基础。然而，过去大多数肿瘤组织结构的研究都基于体外培养皿构建的肿瘤模型，无法完全模拟人体内肿瘤的物理特征，因而我们对人体内肿瘤细胞的组织结构和形成机制尚不清楚。

在本论文中，我们将人体肿瘤组织三维成像实验与系统科学理论相结合，开展了临床人体乳腺肿瘤组织结构的多尺度研究。实验方面，通过与医院合作，我们利用转盘共聚焦显微成像技术对厘米尺度的肿瘤及健康组织分别进行三维荧光显微成像，获取了组织内细胞核的位置及形态，为后续分析建立了数据基础。理论方面，通过对比患者的肿瘤与健康组织，我们由大到小分别从组织、细胞团簇以及单细胞三个不同尺度进行了组织结构的量化分析，获得了乳腺癌组织的多尺度特征。首先，在组织尺度上，肿瘤组织中的细胞密度更大，且更倾向于聚集在一起；通过计算分形维数，研究发现肿瘤组织拥有更复杂的群体细胞构型。其次，在细胞团簇尺度上，我们建立了团簇连通模型，发现了肿瘤组织内部存在更大规模、更紧密的细胞团簇。同时我们进一步的分析结果发现肿瘤组织中的细胞团簇具有更高的信息熵及轮廓分形维数，这表明肿瘤细胞团簇的细胞排列更为无序、且团簇的形貌更加复杂。最后，在单细胞尺度上，我们发现肿瘤细胞核具有更大的体积以及更细长的形态。进一步，我们分析了细胞核形态与局域空间的关联。相比健康组织中位于拥挤区域的细胞核形态会由于较高的局域密度而呈现为球形，而肿瘤组织紧密部分的细胞核形态没有表现出明显的球形特征。基于肿瘤与健康组织的结构特征差异，我们进一步构建了紧密与稀疏组织细胞核形态的差异指标用于识别肿瘤组织。

本研究从真实人体组织切片出发，从系统科学角度对肿瘤多细胞系统展开研究，我们发现在不同尺度下肿瘤组织中细胞的空间分布、细胞间连通、细胞核形态都表现出与健康组织不同的无序特征，展现出较高的复杂性。以此为基础，我们提出了多尺度肿瘤识别参数，为乳腺肿瘤识别提供了新的物理指标。本研究中的肿瘤多尺度结构特征有助于增进对乳腺癌形成及发展机制的理解，为肿瘤防治等相关交叉领域提供了基础理论依据。

Cancer, that is, malignant tumor, is one of the most common causes of death in the world, and the incidence of breast cancer ranks first in the country and even in the world. The malignant proliferation and invasiveness of tumor cells are the main causes of patients’ death. Tumor is a complex system composed of massive cells, and its tissue structure changes with the progression of tumor, showing abnormal physical characteristics and behaviors. Structural changes in tumor tissue leads to its function deteriorates. Understanding the population structural characteristics of tumor cells is not only conducive to our understanding of abnormal tumor functions, but also the basis for identifying breast cancer tumors. However, most studies on tumor tissue structure in the past were based on tumor models constructed in vitro culture dishes, which could not fully simulate the physical characteristics of tumors in the human body. Therefore, we still do not know the tissue structure and formation mechanism of tumor cells in the human body.

In our work, we combined 3D imaging experiments of human tumor tissue with systems science theory to carry out a multi-scale study of clinical human breast tumor tissue structure. In terms of experiments, in cooperation with the hospital, we used spinning disk confocal microscopy imaging technology to perform three-dimensional fluorescence microscopy imaging of centimeter-scale tumors and healthy tissues, obtained the position and shape of the nucleus in the tissue, and established data for subsequent analysis base. In terms of theory, by comparing the patients’ tumor and healthy tissue, we carried out quantitative analysis of the tissue structure from three different scales of tissue, cell clusters, and single cells from large to small, and obtained the multi-scale characteristics of breast cancer tissue. Firstly, at the tissue scale, the cells in tumor tissue are denser and tend to gather. By calculating the fractal dimension, we found that tumor tissue has a more complex population cell configuration. Secondly, at the scale of cell clusters, we established a cluster connectivity model and found that there are larger and tighter cell clusters inside the tumor tissue. At the same time, our further analysis results found that tumor cell clusters have higher information entropy and contour fractal dimension, indicating that the cell arrangement of tumor cell clusters is more disordered, and the shape of clusters is more complex. Finally, at the single-cell scale, we found tumor cells’ nuclear have larger volumes and more elongated morphologies. Further, we analyzed the association of nuclear morphology with its local space. Compared with the cells in the crowded area in the healthy tissue, the nuclear morphology is spherical due to the higher local density, while the nuclear morphology in the compact part of the tumor tissue does not show obvious spherical characteristics. Based on the differences in structural features between tumors and healthy tissues, we further constructed a differential index of compact and sparse tissue nuclear morphology to identify tumor tissues.

Starting from real human tissue slices, our study studied the multicellular system of tumors from the perspective of system science. We found that the spatial distribution, intercellular connectivity, and nuclear morphology of tumor tissues at different scales exhibited different disturbances than healthy tissues, showing high complexity. Based on this, we propose multi-scale tumor identification parameters, which provide new physical indicators for breast tumor identification. The multi-scale structural characteristics of tumors in our study are helpful to enhance the understanding of the formation and development mechanism of breast cancer, and provide a basic theoretical basis for related interdisciplinary fields such as cancer prevention and treatment.