人工智能技术在宫颈癌筛查项目中细胞学质控的应用研究

Application of artificial intelligence technology in quality control for cytological examination in cervical cancer screening program

摘要

目的 探讨人工智能技术在宫颈癌筛查项目细胞学质控工作中应用的可行性及价值,降低漏诊和误诊,提升筛查质量。方法 收集江苏省、四川省等多中心的宫颈癌检查项目外送第三方实验室的2500例液基细胞玻片样本,经质控排除不满意样本后,最终共入组2365例样本纳入研究,由市级病理专家、人工智能技术对外送第三方司抽检样本分别进行质控阅片,并由省级病理专家对于市级病理专家与人工智能技术诊断不符合的标本随机抽取标本进行质控,比较人工智能技术和市级病理专家人工阅片复核结果的一致性差异,评价前者在宫颈癌筛查项目质控环节应用的可行性。结果 人工智能技术与市级病理专家质量控制结果相比,总体符合率无统计学差异(χ²=3.81,P>0.05);按细胞学结果进行比较,未见上皮内病变的细胞和恶性细胞(NILM)及未明确意义的不典型鳞状上皮细胞(ASC-US)数据比较差异有统计学意义(χ²=55.57和19.02,P<0.001);低度鳞状上皮内病变(LSIL)、不典型鳞状上皮细胞-不除外高度鳞状上皮内病变(ASC-H)、高度鳞状上皮内病变(HSIL)、不典型腺上皮细胞(AGC)、鳞状细胞癌(SCC)数据比较差异无统计学意义(χ²=1.91,2.00和0.12,P>0.05);市级病理专家与人工智能技术诊断不符合的146例样本,经省级病理专家质控,有119例结果为NILM及ASC-US的样本人工智能诊断与省级专家诊断不符合。结论 在缺少细胞病理质控专家的地区,可以考虑使用人工智能技术进行宫颈癌筛查项目中细胞学质控,但现阶段人工智能技术还需要不断优化去进一步提升质控准确性.

Abstract

Objective To explore feasibility and application value of artificial intelligence (AI) technology in quality control for cytological examination in cervical cancer screening program, reduce missed diagnosis and misdiagnosis, so as to improve screening quality. Methods 2500 liquid-based thin prep cytologic test (TCT) slide samples sent to third-party laboratories from multiple centers of cervical cancer screening in Jiangsu province and Sichuan province and so on were collected. After quality control, unsatisfactory samples were excluded, and finally 2365 TCT slide samples were included in the study. Firstly, municipal-level pathologists and AI technology respectively conducted quality control slide reading on the liquid-based TCT slide samples sent to the third-party laboratories and selected in random sampling method. And then the provincial-level pathologists randomly selected those TCT slide samples with discordant diagnoses between the municipal-level pathologists and the AI, and reviewed those TCT slide samples for quality control assessment. The consistency between the AI results and the manual review results by the municipal-level pathologists was compared to evaluate feasibility of applying the AI technology in quality control procedures for cervical cancer screening programs. Results There was no significant difference in the overall coincidence rate between the quality control results of the AI technology and those of the municipal-level pathologists (χ²=3.81, P>0.05). However, according to cytological results, there were statistically significant differences in data of negative for intraepithelial lesion or malignancy (NILM) and atypical squamous cell of undetermined significance (ASC-US) between the quality control results of AI technology and those of municipal pathologists (χ²=55.57 and 19.02 respectively, both P<0.001), while the differences in data of low-grade squamous intraepithelial lesion (LSIL), atypical squamous cells-cannot exclude HSIL (ASC-H), high-grade squamous intraepithelial lesion (HSIL), atypical glandular cell (AGC) and squamous cell carcinoma (SCC) were not statistically significant between the municipal pathologists and the AI technology (χ²=1.91, 2.00 and 0.12 respectively, both P>0.05). Among the 146 slide samples whose diagnosis was inconsistent between the municipal pathologists and AI technology, 119 slide samples whose diagnosis results were NILM and ASC-US were inconsistent between the AI's diagnoses and the provincial-level pathologists' diagnoses after quality control by the provincial-level pathologists. Conclusion In areas lacking cytopathological quality control experts, AI technology can be considered for cytological quality control in cervical cancer screening programs, but at present, AI technology still needs to be continuously optimized to further improve accuracy of quality control.

关键词

宫颈癌筛查 / 人工智能技术 / 细胞学 / 质量控制

Key words

cervical cancer screening / artificial intelligence technology / cytology / quality control

参考文献

[1] Zhu X, Yao Q, Dai W, et al. Cervical cancer screening aided by artificial intelligence, China[J]. Bull World Health Organ, 2023, 101(6): 381-390.

[2] Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2024, 74(3): 229-263.

[3] Singh D, Vignat J, Lorenzoni V, et al. Global estimates of incidence and mortality of cervical cancer in 2020: a baseline analysis of the WHO Global Cervical Cancer Elimination Initiative[J]. Lancet Glob Health, 2023, 11(2): e197-e206.

[4] 张小松, 毕蕙, 董燕, 等. 人工智能辅助宫颈细胞学诊断技术的应用及质量控制专家共识[J]. 中国妇幼健康研究, 2023, 34(12): 7-13.

[5] Xie Y, Tan X, Shao H, et al. VIA/VILI is more suitable for cervical cancer prevention in Chinese poverty-stricken region: a health economic evaluation[J]. BMC Public Health, 2017, 17: 1-9.

[6] 魏丽惠, 赵昀, 沈丹华, 等. 中国子宫颈癌筛查及异常管理相关问题专家共识(一)[J]. 中国妇产科临床杂志, 2017, 18(2): 190-192.

[7] Gul S, Murad S, Javed A. Prevalence of High risk Human Papillomavirus in cervical dysplasia and cancer samples from twin cities in Pakistan[J]. Int J Infect Dis, 2015, 34: 14-19.

[8] 张丹, 张宏秀, 吴文新. TCT和HPV宫颈癌筛查方法的临床价值[J]. 江苏医药, 2017, 43(12): 890-892.

[9] 张安慧, 朱敏, 毛建, 等. 人工智能技术在女性宫颈癌筛查中的应用分析[J]. 中国卫生信息管理杂志, 2024, 21(5): 739-744, 750.

[10] Nakisige C, de Fouw M, Kabukye J, et al. Artificial intelligence and visual inspection in cervical cancer screening[J]. Int J Gynecol Cancer, 2023, 33(10): 1515-1521.

[11] Pan D, Zhang C Q, Liang Q L, et al. An efficient method that combines the ThinPrep cytologic test with E6/E7 mRNA testing for cervical cancer screening[J]. Cancer Manag Res, 2019, 11: 4773-4780.

[12] Brown B H, Tidy J A. The diagnostic accuracy of colposcopy: a review of research methodology and impact on the outcomes of quality assurance[J]. Eur J Obstet Gynecol Reprod Biol, 2019, 240: 182-186.

[13] Kurita Y, Meguro S, Kosugi I, et al. Enhancing cervical cancer cytology screening via artificial intelligence innovation[J]. Sci Rep, 2024, 14(1): 19535.

[14] Bao H, Sun X, Zhang Y, et al. The artificial intelligence-assisted cytology diagnostic system in large-scale cervical cancer screening: a population-based cohort study of 0.7 million women[J]. Cancer Med, 2020, 9(18): 6896-6906.

基金资助

中国妇幼保健协会妇幼健康领域科研创新技术项目(ZGFYBJXHKY2101)

连云港市重点研发计划(社会发展)项目(SF2203)