目的 分析基于自动分类器的子宫内膜癌患者外周血炎症-营养参数与病理特征的关系及短期预后预测价值。 方 法 回顾性纳入新疆医科大学第一附属医院 2020 年 4 月至 2024 年 4 月 156 例接受手术治疗的子宫内膜癌患者(试验组)及年 龄匹配的 110 例健康女性(对照组),比较两组纤维蛋白原-白蛋白比(FAR)、预后营养指数(PNI)、淋巴细胞-单核细胞比值 (LMR)及全身免疫炎症指数( SII)。 进一步分析试验组不同病理特征[国际妇产科联盟( FIGO) 分期、分级等] 下各指标差 异,基于 Modeler 自动分类器筛选子宫内膜癌患者预后影响因素,分析外周血炎症-营养参数与预后的关系并探讨其预测价 值。 结果 试验患者炎症和营养指标与对照组存在显著差异,试验组 FAR[(0. 08±0. 02) 比 (0. 07±0. 01)]和 SII[(610. 04± 90. 86) 比 (425. 64±102. 12)]显著升高( P< 0. 05),而 PNI [( 49. 61 ± 8. 05 ) 比 ( 52. 69 ± 1. 57)] 和 LMR[( 3. 96 ± 0. 78) 比 (4. 35±0. 51)]显著降低(P<0. 05)。 进一步分析发现,脉管浸润或神经侵犯阳性的患者 FAR 和 SII 更高,PNI 和 LMR 更低; FIGO Ⅱ期患者的 FAR 和 SII 显著低于Ⅲ、Ⅳ期患者(P<0. 05)。 通过 Modeler 自动分类器建立的预后预测模型中,类神经网络 (准确率 96. 15%)、贝叶斯网络(94. 87%)和 C5 决策树(94. 23%)均表现出良好的预测效能,FAR 是其最重要的预测因子。 结 论 子宫内膜癌患者外周血炎症-营养参数与病理特征显著相关,基于 Modeler 自动分类器构建的预测模型准确率达 94%以 上,炎症-营养指标联合病理特征可作为有效的预后评估工具。

To analyze the relationship between peripheral blood inflammation-nutritional parameters and pathological characteristics of patients with endometrial cancer based on automatic classifiers and its short - term prognostic predictive value. Method A retrospective study included 156 patients with endometrial cancer who underwent surgical treatment in the First Affiliated Hospital of Xinjiang Medical University from April 2020 to April 2024 the experimental group and 110 age-matched healthy women the control group . The fibrinogen - albumin ratio FAR prognostic nutritional index PNI lymphocyte-monocyte ratio LMR and systemic immune inflammation index SII were compared between the two groups. Further analyze the differences of each index under different pathological characteristics FIGO stage grade etc. in the experimental group. Based on the Modeler automatic classifier screen the prognostic influencing factors of patients with endometrial cancer analyze the relationship between peripheral blood inflammation-nutrition parameters and prognosis and explore its predictive value. Result The research results showed that there were significant differences in inflammation and nutritional indicators between the experimental patients and the control group. The FAR 0. 08±0. 02 vs 0. 07±0. 01 and SII 610. 04±90. 86 vs 425. 64±102. 12 of the experimental group were significantly higher than those of the control group P<0. 05 . However PNI 49. 61±8. 05 vs 52. 69±1. 57 and LMR 3. 96±0. 78 vs 4. 35±0. 51 were significantly lower than those of the control group P<0. 05 . Further analysis revealed that patients with positive vascular invasion or nerve invasion had higher FAR and SII than those with negative results and lower PNI and LMR than those with negative results. The FAR and SII of patients in FIGO stage II were significantly lower than those of patients in stage Ⅲ and IV P<0. 05 . Among the prognosis prediction models established through the Modeler algorithm neural networks with an accuracy rate of 96. 15% Bayesian networks 94. 87% and C5 decision trees 94. 23% all demonstrated excellent prediction performance with FAR being the most important predictor. Conclusion The inflammation-nutrition parameters in the peripheral blood of patients with endometrial cancer are significantly correlated with pathological characteristics. The accuracy rate of the prediction model constructed based on Modeler is over 94%. The combination of inflammation - nutrition indicators and pathological characteristics can be used as an effective prognostic evaluation tool.