第一部分：

【目的】 探讨多序列磁共振成像（MRI）及全肿瘤表观扩散系数（ADC）直方图分析在子宫内膜癌（EC）术前分子分型预测中的价值。

【材料与方法】 本研究回顾性纳入2021年1月至2024年5月146例经病理确诊并完成分子分型的EC患者，所有患者术前均接受了MRI检查。分子分型的判定标准与流程遵循2022版欧洲肿瘤内科学会（ESMO）指南。EC的分子分型包括以下四种：POLE超突变型（POLEmut）、错配修复缺陷型（dMMR）、p53突变型（p53abn）以及无特异性分子特征型（NSMP）。本研究针对以下两个方向进行分组探讨：不良预后筛选（p53abn型vs非p53abn型）和高肿瘤突变负荷筛选（高肿瘤突变负荷/ TMB-H组vs低肿瘤突变负荷/ TMB-L组）。MRI特征由两名放射科医生独立评估，包括最大肿瘤直径、生长模式以及与分期相关的因素（如深肌层浸润等）。基于Function tool ADC软件在轴位扩散加权成像（DWI）序列执行肿瘤最大径层面的感兴趣区（ROI）手动分割，获取二维ADC值（2D-ADC）；随后通过3D-Slicer进行多平面肿瘤三维容积重建，在DWI序列连续层面逐层勾画ROI后，将ROI自动映射至ADC图；最终应用radiomics插件从ADC图中提取直方图参数。对于不同类型变量的统计差异分析：连续型数据根据分布特征选用Mann-Whitney U检验或t检验进行组间比较，类别型数据则依据样本量特性选择卡方检验或Fisher精确检验。在诊断效能评估方面，通过构建受试者工作特征（ROC）曲线并计算其对应的曲线下面积（AUC）来量化检测方法的判别能力。

【结果】 TMB-L组EC相比TMB-H组更易发生子宫外扩散（p<0.05）。p53abn型EC相比非p53abn型更易出现子宫外扩散和淋巴结转移（p<0.05）。最小ADC值、平均ADC值、第10、50、90百分位ADC值以及峰度在p53abn型与非p53abn型EC之间的差异具有统计学意义（p<0.05）。在所有参数中，最小ADC值的AUC最高（0.70, 95%置信区间: 0.60-0.80）。

【结论】 多序列MRI和全肿瘤 ADC 直方图分析具有术前预测EC分子分型的潜在价值。

第二部分：

【目的】 探讨多序列磁共振成像（MRI）及全肿瘤表观扩散系数（ADC）直方图分析在子宫内膜癌（EC）术前风险分层预测中的价值。

【材料与方法】 本研究回顾性纳入2021年1月至2024年5月在中国医学科学院肿瘤医院经手术病理确诊并接受了术前MRI检查的146例EC患者。本研究基于2022版欧洲肿瘤内科学会（ESMO）指南，将146例EC患者分为五个风险组（低风险组、中风险组、中高风险组、高风险组、远处转移组），并重点分析低风险组与非低风险组、（低+中）风险组与非（低+中）风险组之间各参数的差异。MRI特征由两名放射科医生独立评估，包括肿瘤最大直径、生长模式以及与分期相关的因素（如深肌层浸润等）。采用Function tool ADC软件及3D-Slicer软件分别获取单层面二维ADC（2D-ADC）值和全肿瘤ADC直方图参数。同时评估血清癌抗原125（CA125）水平。通过整合多序列MRI特征、全肿瘤ADC直方图参数及血清CA125水平，构建联合预测模型，评估其在术前风险分层中的诊断效能。对于不同类型变量的统计差异分析：连续型数据根据分布特征选用Mann-Whitney U检验或t检验进行组间比较，类别型数据则依据样本量特性选择卡方检验或Fisher精确检验。多变量二元逻辑回归分析构建组合模型，模型效能以受试者工作特征（ROC）曲线下面积（AUC）作为评价标准。

【结果】 低风险组及（低+中）风险组EC的肿瘤最大直径显著小于非对应风险组（p＜0.05），且侵袭性特征（如深肌层浸润、淋巴结转移等）的发生率显著较低（p＜0.05）。此外，非（低+中）风险组EC的血清CA125水平显著高于（低+中）风险组（p＜0.05）。最小ADC值、均匀度、峰度和熵值在低风险组与非低风险组EC以及（低+中）风险组与非（低+中）风险组EC之间均存在显著差异（p＜0.05）。此外，2D-ADC值和第10百分位ADC值在低风险与非低风险组EC之间存在显著差异（p＜0.05）。基于肿瘤最大直径、深肌层浸润、淋巴结转移及最小ADC值构建的组合模型对低风险组EC的预测效能最优（AUC=0.84，95%置信区间: 0.78-0.90）。在此基础上进一步纳入血清CA125水平后，组合模型对非（低+中）风险组EC的预测效能最佳（AUC=0.80，95%置信区间: 0.72-0.87）。

【结论】 多序列MRI特征联合全肿瘤ADC直方图分析，可为EC的术前风险分层提供有效预测。

【关键词】 子宫内膜癌；磁共振成像；风险分层；表观扩散系数；直方图分析

第一部分：

[Objective] This study aimed to assess the clinical utility of integrated multi-parametric magnetic resonance imaging (MRI) sequences combined with whole-tumor apparent diffusion coefficient (ADC) histogram profiling for predicting molecular subtypes in endometrial cancer (EC) during preoperative evaluation.

[Materials and Methods] This retrospective cohort analysis enrolled 146 histologically confirmed EC that underwent comprehensive molecular classification following standardized diagnostic protocols, spanning the period from January 2021 to May 2024. All patients underwent preoperative MRI examinations. The criteria and procedures for molecular classification were conducted in accordance with the 2022 guidelines of the European Society for Medical Oncology (ESMO). The molecular classification of EC consists of four subtypes: POLE ultramutated (POLEmut), mismatch repair deficient (dMMR), p53 abnormal (p53abn), and no specific molecular profile (NSMP). This study was conducted in two aspects of group analysis: screening for poor prognosis (p53abn vs non-p53abn) and screening for high tumor mutational burden (Tumor Mutational Burden-High, TMB-H vs Tumor Mutational Burden-Low, TMB-L). MRI features, including the maximum tumor diameter, growth pattern, and staging-related factors (such as deep myometrial invasion) were independently assessed by two radiologists. The Function tool ADC platform was employed for manual segmentation of region of interest (ROI) on the single slice corresponding to the maximum tumor diameter in the diffusion-weighted imaging (DWI) sequence, thereby obtaining the two-dimensional ADC (2D-ADC) value. Subsequently, the 3D-slicer software was used to delineate the ROI slice by slice on the DWI sequence, with the ROI automatically replicated onto the ADC map. Finally, histogram parameters were extracted from the ADC map using the radiomics plugin. For continuous variables, statistical significance was determined using either the Mann-Whitney U test or the independent samples t-test. In contrast, for categorical variables, statistical significance was evaluated using the chi-square test or Fisher's exact test. The diagnostic accuracy framework incorporated systematic application of receiver operating characteristic (ROC) analytical methodology, wherein quantitative interpretation of area under the curve (AUC) measurements was adopted as the principal validation criterion.

[Results] TMB-L EC exhibited a higher frequency of extrauterine extension compared to TMB-H EC (p<0.05). p53abn EC was more likely to present with extrauterine extension and lymphadenopathy than non-p53abn EC (p<0.05). The differences in the minimum, mean, ADC values at the 10th, 50th, and 90th percentiles, as well as kurtosis, between p53abn and non-p53abn EC were statistically significant (p<0.05). Among these parameters, the minimum ADC value demonstrated the highest area under the curve (AUC=0.70; 95% confidence interval: 0.60-0.80).

[Conclusion] Multi-sequence MRI and whole-tumor ADC histogram analysis demonstrate potential value in preoperative prediction of molecular classification of EC.

[Keywords] endometrial cancer; magnetic resonance imaging; molecular classification; apparent diffusion coefficient; histogram analysis

第二部分：

[Objective] This study aimed to evaluate the clinical utility of integrated multi-parametric magnetic resonance imaging (MRI) sequences combined with whole-tumor apparent diffusion coefficient (ADC) histogram profiling for preoperative risk stratification assessment in endometrial cancer (EC).

[Materials and Methods] This retrospective cohort analysis enrolled 146 histologically confirmed EC that underwent preoperative MRI examinations, spanning the period from January 2021 to May 2024. This study, based on the 2022 European Society for Medical Oncology (ESMO) guidelines, classified 146 EC patients into five risk groups (low-risk, intermediate-risk, intermediate-high-risk, high-risk, and metastatic groups). It further focused on analyzing the differences in various parameters among the low-risk group versus the non-low-risk group, as well as between the combined (low+intermediate)-risk group and the non-(low+intermediate)-risk group. MRI features, including the maximum tumor diameter, growth pattern, and staging-related factors (such as deep myometrial invasion) were independently assessed by two radiologists. The Function tool ADC software and 3D-Slicer software were utilized to extract the two-dimensional ADC (2D-ADC) values of a single slice and the whole-tumor ADC histogram parameters, respectively. Additionally, serum cancer antigen 125 (CA125) levels were measured. By integrating multi-sequence MRI features, whole-tumor ADC histogram parameters, and serum CA125 levels, we developed a combined predictive model and assessed its diagnostic accuracy for preoperative risk stratification. Intergroup differences for continuous variables were evaluated using the Mann-Whitney U test or the independent samples t-test, whereas for categorical variables, statistical significance was determined using the chi-square test or Fisher's exact test. A multivariate binary logistic regression analysis was utilized to construct a combined predictive model. The diagnostic performance of both individual parameters and the combined model was assessed via receiver operating characteristic (ROC) curve analysis, with the area under the curve (AUC) serving as the primary metric.

[Results] In the low-risk and (low+intermediate)-risk groups, the maximum tumor diameter was markedly reduced compared to their respective non-corresponding risk groups (p<0.05). Additionally, the occurrence of invasive features, including deep myometrial invasion and lymphadenopathy, was substantially lower (p<0.05). Furthermore, serum CA125 levels were markedly elevated in the non-(low+intermediate)-risk group when compared to the (low+intermediate)-risk group (p<0.05). Notably, there were significant variations in the minimum, uniformity, kurtosis, and entropy between the low-risk and non-low-risk groups (p<0.05), as well as between the (low+intermediate)-risk and non-(low+intermediate)-risk groups (p<0.05). Additionally, significant differences were observed in the 2D-ADC value and the 10th percentile ADC value between the low-risk and the non-low-risk groups (p<0.05). A combined model based on maximum tumor diameter, deep myometrial invasion, lymphadenopathy, and minADC demonstrated optimal predictive performance for the low-risk group (AUC=0.84, 95% confidence interval: 0.78-0.90). Furthermore, incorporating serum CA125 levels into this model resulted in the best predictive performance for the non-(low+intermediate)-risk group (AUC=0.80, 95% confidence interval: 0.72-0.87).

[Conclusion] The combination of multi-sequence MRI features and whole-tumor ADC histogram analysis offers a reliable approach for preoperative risk stratification in EC.

[Keywords] endometrial cancer; magnetic resonance imaging; risk stratification; apparent diffusion coefficient; histogram analysis