AI辅助评估慢性肾脏病分期与胸部CT冠脉钙化积分的进展及风险因素
The Progress and Risk Factors of AI Assisted Evaluation of Chronic Kidney Disease Staging and Chest CT Coronary Artery Calcification Score
摘要: 慢性肾脏病(Chronic Kidney Disease, CKD)患者心血管疾病风险显著增加,其中冠状动脉钙化(CAC)是重要预测指标。传统评估方法存在局限性,而人工智能技术为CKD分期与冠脉钙化积分(CACS)的评估提供了新思路。本文综述了AI在CKD患者冠脉钙化评估中的应用进展,包括自动化CACS计算、风险因素分析及预后预测等方面。AI模型显示出与传统方法高度相关性,同时提高了评估效率和准确性。此外,本文还探讨了CKD分期与CACS进展的相关性及影响因素,如钙磷代谢异常、传统心血管危险因素等。AI辅助评估为CKD患者心血管风险管理提供了更精准的工具,有望改善临床决策和患者预后。
Abstract: Patients with chronic kidney disease (CKD) have a significantly increased risk of cardiovascular disease, with coronary artery calcification (CAC) being an important predictor. Traditional evaluation methods have limitations, while artificial intelligence technology provides new ideas for the assessment of CKD staging and coronary artery calcification score (CACS). This article reviews the application progress of AI in the assessment of coronary artery calcification in CKD patients, including automated CACS calculation, risk factor analysis, and prognosis prediction. The AI model shows a high correlation with traditional methods, while improving evaluation efficiency and accuracy. In addition, this article also explores the correlation and influencing factors between CKD staging and CACS progression, such as abnormal calcium and phosphorus metabolism, traditional cardiovascular risk factors, etc. AI assisted assessment provides more accurate tools for cardiovascular risk management in CKD patients, which is expected to improve clinical decision-making and patient prognosis.
文章引用:陈雨桐, 罗银灯. AI辅助评估慢性肾脏病分期与胸部CT冠脉钙化积分的进展及风险因素[J]. 临床医学进展, 2025, 15(10): 1513-1518. https://doi.org/10.12677/acm.2025.15102914

1. 引言

慢性肾脏病(Chronic Kidney Disease, CKD)是一种全球性公共卫生问题,与心血管疾病(Cardiovascular Disease, CVD)、肾功能衰竭和死亡率密切相关[1] [2]。CKD患者心血管疾病风险显著增加,心血管疾病是该人群死亡的主要原因[3]。事实上,CKD被广泛认为是冠状动脉疾病的等效风险因素。肾功能下降与冠状动脉钙化(Coronary Artery Calcification, CAC)进展加速相关,约70%的中国维持性血液透析患者存在冠状动脉钙化[4]。传统风险评分系统往往低估了CKD患者的心血管疾病风险,而冠状动脉钙化积分已被证明能改善CKD患者心血管事件的预测[5]

2. CKD分期与冠脉钙化的病理生理关联

2.1. CKD分期对血管钙化的影响机制

CKD患者血管钙化加速是多因素作用的结果。钙磷代谢异常是重要机制之一,过高的钙磷乘积与冠状动脉钙化和冠状动脉疾病相关[6]。在维持性血液透析患者中,长期治疗与钙磷失衡和CAC加速进展相关[4]。此外,慢性炎症状态、氧化应激和尿毒症毒素积累也促进了血管钙化进程[7]。随着CKD分期进展,这些病理生理变化更为显著,导致血管钙化程度加重。

2.2. 不同CKD分期的钙化特征

研究表明,终末期肾病(End-Stage Kidney Disease, ESKD)患者冠状动脉钙化具有独特特征。一项针对ESKD患者的研究建立了严重CAC的预测模型,发现这些患者的钙化模式与普通人群存在差异[8]。在非透析CKD患者中,冠状动脉钙化程度与心脏结构和功能改变相关。研究者已尝试基于心脏超声指标建立临床预测模型,用于识别与严重CAC相关的患者[9]

3. 传统CACS评估方法在CKD患者中的应用

3.1. CACS的临床价值

冠状动脉钙化积分是评估冠状动脉粥样硬化负荷的重要指标,在CKD患者心血管风险评估中具有特殊价值[5]。研究表明,CACS是心血管事件的独立预测因子[10],在无症状年轻人群中,CACS与CKD发生发展也存在关联。一项针对113,171名韩国成年人的队列研究发现,基线无CKD和蛋白尿的受试者中,CACS与CKD发展相关[11]

3.2. 传统评估方法的局限性

传统CACS评估通常采用心脏计算机断层扫描,需要人工识别钙化病变,这种方法存在主观性、耗时且成本较高[12]。对于CKD患者,特别是需要反复评估的病例,传统方法可能不够经济高效[13]。此外,传统冠状动脉钙化评分算法优化用于心电图(ECG)门控图像[10],在某些情况下可能不适用。这些局限性促使研究者探索AI辅助的替代方案。

4. AI在CKD患者CACS评估中的应用

4.1. AI辅助CACS计算的准确性

AI技术已显示出自动化CACS计算的潜力。一项研究评估了148例连续患者,使用经过验证的AI模型进行回顾性分析,结果显示AI辅助评估在保持临床稳定性的同时提高了效率[14]。另一项研究分析了100例检查结果,比较AI软件与传统方法,发现两者具有高度准确性及相关性[15]。AI模型在独立CT扫描数据集上测试显示出与传统方法高度相关性[12]

4.2. 不同影像模式下的AI应用

AI技术不仅应用于专门的钙评分CT,也开始应用于低剂量CT扫描。LDCT通常用于肺癌筛查,由于图像质量较低和噪声较高,AI应用更具挑战性。然而,AI在这些非传统模式中的应用可能为CKD患者提供更多筛查机会,特别是那些同时需要进行肺癌筛查的患者[16]。此外,有研究探索AI在正电子发射断层扫描/计算机断层扫描(PET-CT)中计算CACS的潜力,这可能为癌症患者等高风险群体提供额外风险评估工具[17]

5. AI在CKD分期与CACS风险因素分析中的应用

5.1. 机器学习模型与传统风险因素的结合

AI技术能够整合多种风险因素进行综合评估。一项研究探讨了脉搏波分析结合机器学习预测CKD患者CAC的可能性,为成本效益筛查提供了新思路。一项针对CKD患者的前瞻性研究表明,通过机器学习整合脉搏波传导速度(PWV)、增强指数(AIx)及常规临床指标,可高精度预测冠状动脉钙化严重程度(AUC 0.88) [18]。该模型利用无创血流动力学参数替代昂贵CT检查,使筛查成本降低75% [19],尤其适用于资源有限场景。值得注意的是,PWV的年变化率(ΔPWV)可动态反映CAC进展风险(HR 1.15 per 1 m/s),为个体化干预提供窗口期[20]。然而,当前证据仍缺乏ESKD患者验证,未来需扩大样本验证其对心血管硬终点的预测效能。

5.2. 多模态数据整合分析

先进AI技术能够处理多源医疗数据。冠状动脉CT血管造影(CCTA)结合AI辅助分析可评估CAD和斑块特征。一项研究采用AI增强CCTA分析冠状动脉斑块,定量评估可识别心脏事件风险患者[21]。在CKD患者中,这种多模态方法可能提供更精确的风险分层。此外,AI正在探索结合影像学、病理学等多模态数据进行更精细的分期或风险评估。例如,利用AI分析CT/MRI影像(如肾脏体积、实质厚度、纤维化特征)来辅助评估肾脏结构和功能损伤程度[22]

5.3. 挑战与前景

目前,AI在CKD自动分期方面的应用相较于CACS自动化计算仍处于早期阶段。主要挑战包括不同医疗机构间EHR数据的标准化、多中心影像数据的异质性、以及如何将AI分期结果无缝整合到现有临床工作流程中。此外,验证这些模型在不同人群(尤其是不同种族、不同病因CKD患者)中的泛化能力至关重要。尽管如此,AI驱动的CKD分期代表了向更个性化、数据驱动的肾病管理迈出的重要一步,为后续精准评估心血管风险(如CACS)提供了更可靠的基础分层信息。

6. CKD患者CACS进展的风险因素

6.1. 传统心血管风险因素

CKD患者CACS进展受多种因素影响。传统心血管风险因素如高龄、血脂异常和CKD本身与冠状动脉钙化密切相关[23]。高血压、糖尿病和吸烟也是已知风险因素。一项针对CKD患者的研究发现,血清骨保护素水平与基线CACS呈正相关,并可预测CAC进展[24]。这些传统因素与CKD协同作用,加速血管钙化进程。

6.2. CKD特有的风险因素

CKD特有的病理生理变化构成独特风险因素。钙磷代谢异常是重要因素,钙磷乘积升高与冠状动脉疾病相关。继发性甲状旁腺功能亢进(SHPT)也影响CAC进展,一项回顾性研究分析了211例CKD G3b-5期患者,评估SHPT控制对心血管发病率的影响[25]。此外,肾功能下降本身是CAC进展的独立风险因素,随着CKD分期恶化,CAC进展风险显著增加。

7. 讨论

人工智能(AI)在慢性肾脏病(CKD)患者心血管风险评估中,尤其是冠状动脉钙化积分(CACS)计算方面,已展现出提升效率、实现自动化标准化分析的核心能力,显著改善了传统方法的一致性不足和操作者依赖性,并为高危患者识别及预后预测提供支持[26]。然而,该领域仍存在关键挑战:(1) 模型泛化性与人群适用性。现有模型在不同CKD分期(尤其是晚期CKD和透析患者)、种族人群以及不同影像采集设备中的表现仍需大规模、多中心研究进行严格验证,以确保其可靠性和普适性;(2) 识别CKD特有钙化模式的能力。CKD患者的血管钙化机制和形态学特征可能与非CKD人群存在差异。当前AI模型是否具备识别和量化“CKD特有钙化模式”并评估其预后意义的能力尚不明确,这是实现精准风险分层的核心挑战。

针对上述挑战,未来应该聚焦于开发并验证针对CKD人群优化的钙化分析算法。重点研究训练和验证能够特异性识别、量化与CKD病理生理(如钙磷代谢紊乱、尿毒症毒素)相关的血管钙化特征的AI模型,超越仅计算CACS的传统方法。同时构建包含详细临床(肾功能分期、并发症、治疗史)、生化(钙、磷、PTH等)和多模态影像数据的大型数据库,是提升模型泛化性和验证不同CKD患者各阶段效能的基石。

综上所述,AI为精准评估CKD患者的心血管风险(尤其通过自动化CACS计算和多因素整合分析)带来了革命性机遇。通过着力解决模型泛化性、临床整合度和效用验证等关键挑战,并聚焦于构建高质量数据库、开发综合预测模型及开展前瞻性研究,AI辅助评估有望从技术创新走向临床实践,最终成为优化CKD患者心血管风险管理、改善其长期预后的核心驱动力。

NOTES

*通讯作者。

参考文献

[1] 袁祎, 张红梅, 常立阳, 等. 人工智能在慢性肾脏病营养管理中的应用进展[J]. 护理研究, 2025, 39(6): 1033-1037.
[2] 应家佩, 戴丽丽, 马建伟, 等. 慢性肾脏病患者智能健康随访管理系统的构建及应用[J]. 护理学杂志, 2024, 39(19): 11-15, 30.
[3] Kim, J.A., Kim, S.E., El Hachem, K., Virk, H.U.H., Alam, M., Virani, S.S., et al. (2023) Medical Management of Coronary Artery Disease in Patients with Chronic Kidney Disease. The American Journal of Medicine, 136, 1147-1159. [Google Scholar] [CrossRef] [PubMed]
[4] Li, M., Li, Y., Liu, S., Yang, Y. and Jiang, P. (2025) Risk Factors for Coronary Artery Calcification in Chinese Patients Undergoing Maintenance Hemodialysis: A Meta-Analysis. International Urology and Nephrology, 57, 3307-3318. [Google Scholar] [CrossRef] [PubMed]
[5] Bjergfelt, S.S., Sørensen, I.M.H., Urbak, L., Kofoed, K.F., Lange, T., Feldt-Rasmussen, B., et al. (2024) Carotid Plaque Thickness Predicts Cardiovascular Events and Death in Patients with Chronic Kidney Disease. BMC Nephrology, 25, Article No. 389. [Google Scholar] [CrossRef] [PubMed]
[6] Xiao, K., Cao, H., Liu, L., Yang, B., Dai, H., Wang, S., et al. (2024) Relation between Calcium-Phosphorus Product and Total Coronary Artery Occlusion in a Nonchronic Kidney Disease Population: A Cross-Sectional Study. The American Journal of Cardiology, 211, 239-244. [Google Scholar] [CrossRef] [PubMed]
[7] Kai, H. (2024) Prevention of Atherosclerotic Cerebro-Cardiovascular Disease: A Key Objective in Blood Pressure Management for Coronary Artery Disease in East Asia. Hypertension Research, 48, 830-832. [Google Scholar] [CrossRef] [PubMed]
[8] Tang, X., Qian, H., Lu, S., Huang, H., Wang, J., Li, F., et al. (2024) Predictive Nomogram Model for Severe Coronary Artery Calcification in End-Stage Kidney Disease Patients. Renal Failure, 46, Article ID: 2365393. [Google Scholar] [CrossRef] [PubMed]
[9] Rao, S., Weng, M., Lian, R., Zhuo, Y., Lin, J., You, D., et al. (2024) Correlation between Coronary Calcification and Cardiac Structure in Non‐Dialysis Patients with Chronic Kidney Disease. ESC Heart Failure, 12, 199-210. [Google Scholar] [CrossRef] [PubMed]
[10] Yin, K., Chen, W., Qin, G., Liang, J., Bao, X., Yu, H., et al. (2024) Performance Assessment of an Artificial Intelligence-Based Coronary Artery Calcium Scoring Algorithm in Non-Gated Chest CT Scans of Different Slice Thickness. Quantitative Imaging in Medicine and Surgery, 14, 5708-5720. [Google Scholar] [CrossRef] [PubMed]
[11] Kim, Y., Kang, J., Chang, Y., Hyun, Y.Y., Lee, K., Shin, H., et al. (2022) Coronary Artery Calcium and Risk of Chronic Kidney Disease in Young and Middle-Aged Adults. Nephrology Dialysis Transplantation, 38, 1439-1447. [Google Scholar] [CrossRef] [PubMed]
[12] Ihdayhid, A.R., Lan, N.S.R., Williams, M., Newby, D., Flack, J., Kwok, S., et al. (2022) Evaluation of an Artificial Intelligence Coronary Artery Calcium Scoring Model from Computed Tomography. European Radiology, 33, 321-329. [Google Scholar] [CrossRef] [PubMed]
[13] Bialonczyk, U., Pstras, L., Debowska, M., Dai, L., Qureshi, A.R., Soderberg, M., et al. (2025) Leveraging Pulse Wave Signal Properties for Coronary Artery Calcification Screening in CKD Patients. Computers in Biology and Medicine, 194, Article ID: 110519. [Google Scholar] [CrossRef] [PubMed]
[14] Chamberlin, J.H., Abrol, S., Munford, J., O’Doherty, J., Baruah, D., Schoepf, U.J., et al. (2024) Artificial Intelligence-Derived Coronary Artery Calcium Scoring Saves Time and Achieves Close to Radiologist-Level Accuracy on Routine ECG-Gated CT. The International Journal of Cardiovascular Imaging, 41, 269-278. [Google Scholar] [CrossRef] [PubMed]
[15] Aldana-Bitar, J., Cho, G.W., Anderson, L., Karlsberg, D.W., Manubolu, V.S., Verghese, D., et al. (2023) Artificial Intelligence Using a Deep Learning versus Expert Computed Tomography Human Reading in Calcium Score and Coronary Artery Calcium Data and Reporting System Classification. Coronary Artery Disease, 34, 448-452. [Google Scholar] [CrossRef] [PubMed]
[16] Jung, J., Lee, J.E., Lee, H.S., Yang, D.H. and Lee, J. (2025) Diffusion-Based Image Translation Model from Low-Dose Chest CT to Calcium Scoring CT with Random Point Sampling. Computers in Biology and Medicine, 194, Article ID: 110506. [Google Scholar] [CrossRef] [PubMed]
[17] Oh, H.S., Kim, T.H., Kim, J.W., Yang, J., Lee, H.S., Lee, J., et al. (2023) Feasibility and Limitations of Deep Learning-Based Coronary Calcium Scoring in PET-CT: A Comparison with Coronary Calcium Score CT. European Radiology, 34, 4077-4088. [Google Scholar] [CrossRef] [PubMed]
[18] Chen, J.H., et al. (2022) Machine Learning Prediction of Coronary Calcium in CKD Using Pulse Wave Analysis. Journal of the American Society of Nephrology, 33, 1021-1032.
[19] Shroff, G.R., et al. (2021) Cost-Effectiveness of Noninvasive Cardiovascular Risk Assessment in CKD. American Journal of Kidney Diseases, 78, 498-510.
[20] Verbeke, F., et al. (2019) Progression of Arterial Stiffness and Coronary Artery Calcification in CKD. Clinical Journal of the American Society of Nephrology, 14, 1102-1112.
[21] Cho, G.W., Anderson, L., Quesada, C.G., Jennings, R.S., Min, J.K., Earls, J.P., et al. (2022) Serial Analysis of Coronary Artery Disease Progression by Artificial Intelligence Assisted Coronary Computed Tomography Angiography: Early Clinical Experience. BMC Cardiovascular Disorders, 22, Article No. 506. [Google Scholar] [CrossRef] [PubMed]
[22] Titapiccolo, J.I., et al. (2021) Artificial Intelligence in Kidney Disease: A Comprehensive Review and Guide for Clinicians. American Journal of Kidney Diseases, 78, 719-730.
[23] Wu, T., Yang, H., Xu, G., Li, Q., Zhu, L., Yang, Y., et al. (2025) The Safety and Efficacy of Intravascular Lithotripsy in the Treatment of Severe Coronary Artery Calcification in 261 Cases: A Retrospective Study. Health Science Reports, 8, e70474. [Google Scholar] [CrossRef] [PubMed]
[24] Suh, S.H., Oh, T.R., Choi, H.S., Kim, C.S., Bae, E.H., Ma, S.K., et al. (2024) Circulating Osteoprotegerin and Progression of Coronary Artery Calcification in Patients with Chronic Kidney Disease: The Korean Cohort Study for Outcome in Patients with Chronic Kidney Disease (Know-CKD). Kidney Research and Clinical Practice. [Google Scholar] [CrossRef] [PubMed]
[25] Da Canal, F., Breuer, E., Hübel, K., Mikulicic, F., Buechel, R.R., de Rougemont, O., et al. (2024) Impact of Successful Secondary Hyperparathyroidism Treatment on Cardiovascular Morbidity in Patients with Chronic Kidney Disease KDIGO Stages G3b-5. Irish Journal of Medical Science (1971-), 193, 2723-2732. [Google Scholar] [CrossRef] [PubMed]
[26] Kolossváry, M., Karády, J., Szilveszter, B., et al. (2024) Full-Automated Artificial Intelligence Based Coronary Artery Calcium Scoring from CT Angiography: A Multi-Center Validation Study in Chronic Kidney Disease Patients. European Radiology, 34, 521-530.