基于视网膜图像自动分析技术的孕早期抑郁风险评估

张达谦; 曲翌敏; 徐仲鍈; 李作为; 马海会; 叶琪; 江宇

doi:10.16462/j.cnki.zhjbkz.2025.04.016

基于视网膜图像自动分析技术的孕早期抑郁风险评估

doi: 10.16462/j.cnki.zhjbkz.2025.04.016

1.
中国医学科学院北京协和医学院群医学及公共卫生学院，北京 100730
2.
香港中文大学赛马会公共卫生及基层医疗学院生物统计学系，香港 999077
3.
北京市通州区妇幼保健院，北京 101101

基金项目:

国家自然科学基金 82404400

详细信息

通讯作者:
江宇，E-mail: jiangyu@pumc.edu.cn

中图分类号: R173;R770.41
计量
- 文章访问数: 14
- HTML全文浏览量: 13
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2025-01-07
- 修回日期: 2025-03-13
- 刊出日期: 2025-04-10

Risk assessment of early pregnancy depression based on automated retinal image analysis technology

1.
School of Population Medicine and Public Health, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing 100730, China
2.
Centre for Clinical Research and Biostatistic, Jockey Club School of Public Health and Primary Care, the Chinese University of Hong Kong, HongKong 999077, China
3.
Tongzhou Maternal & Child Health Hospital, Beijing 101101, China

Funds:

Naional Natural Science Foundation of China 82404400

More Information

Corresponding author: JIANG Yu, E-mail: jiangyu@pumc.edu.cn

摘要

摘要: 目的探索基于机器学习的视网膜图像自动分析技术用于孕早期抑郁的风险评估。方法于2023年5月―2024年5月在北京市通州区妇幼保健院招募了936名孕妇，在孕早期拍摄其双侧视网膜图像，并采用爱丁堡抑郁量表中文版(Edinburgh postnatal depression scale, EPDS)评估孕妇抑郁情况。按年龄±2岁控制在1∶1至1∶4之间进行病例对照匹配，采用视网膜图像自动分析技术计算常规视网膜特征，并采用卷积神经网络模型提取与孕早期抑郁相关的特异图像特征。基于logistic回归分析模型进行建模，采用十折交叉验证法评估模型性能。结果共449例研究对象被纳入分析，其中孕早期抑郁组92例。抑郁组视网膜血管静脉分叉角度(69.36±2.17)和静脉血管分叉系数(0.83±0.01)均低于非抑郁组(69.95±2.00, t=2.47, P=0.014; 0.84±0.01, t=2.56, P=0.011)。基于视网膜特征风险评估模型的受试者工作特征曲线下面积为0.995(95% CI: 0.990~0.999)，模型灵敏度为0.978，特异度为0.958。结论基于视网膜图像自动分析技术可以准确对孕早期抑郁进行识别，为早期干预和治疗提供了有力支持。
- 孕早期抑郁 /
- 视网膜微血管 /
- 人工智能 /
- 风险评估
Abstract: Objective This study aimed to explore the use of machine learning-based retinal image analysis techniques for risk assessment of early pregnancy depression. Methods From May 2023 to May 2024, a total of 936 pregnant women were recruited at Tongzhou District Maternal and Child Health Hospital, Beijing. Bilateral retinal images were captured during their first trimester of pregnancy, and maternal depressive symptoms were assessed using the Chinese version of the Edinburgh postnatal depression scale (EPDS). A case-control matching was performed with an age difference of ±2 years, maintaining a ratio between 1∶1 and 1∶4. Conventional retinal features were calculated using automated retinal image analysis techniques, and specific image features related to early pregnancy depression were extracted using a convolutional neural network model. Modeling was conducted using a logistic regression model, with model performance evaluated via ten-fold cross-validation. Results A total of 449 subjects were included in the analysis, with 92 cases in the early pregnancy depression group. The retinal venous branching angle (69.36±2.17) and venous branching coefficient (0.83±0.01) in the depression group were both lower than those in the non-depression group (69.95±2.00, t=2.47, P=0.014; 0.84±0.01, t=2.56, P=0.011). The area under the receiver operating characteristic curve for the risk assessment model based on retinal features was 0.995 (95% CI: 0.990-0.999), with a sensitivity of 0.978, a specificity of 0.958. Conclusions Automated retinal image analysis techniques can accurately identify early pregnancy depression, providing strong support for early intervention and treatment.
- Early pregnancy depression /
- Retinal images /
- Artificial intelligence /
- Risk assessment

HTML全文

图 1 佳能CR2-AF眼底相机拍摄的视网膜图像示例

Figure 1. Example of retinal image captured by the Canon CR2-AF fundus camera

下载: 全尺寸图片幻灯片

图 2 卷积神经网络算法挖掘的与孕早期抑郁相关的视网膜图像特征分布

Figure 2. Distribution of retinal image features associated with early pregnancy depression identified by the convolutional neural network algorithm

下载: 全尺寸图片幻灯片

图 3 基于视网膜特征的孕早期抑郁风险评估模型的受试者工作特征曲线

Figure 3. Receiver perating characteristic curve of the early pregnancy depression risk assessment model based on retinal features

下载: 全尺寸图片幻灯片

表 1 抑郁组与非抑郁组基线资料比较

Table 1. Comparison of baseline data between the depressed group and the non-depressed group

变量Variable	非抑郁组 Non-depressed group^①	抑郁组 Depression group^①	t/x²值 value	P值 value
年龄/岁Age/years	30.07±3.98	30.36±4.08	0.61	0.541
身高Height/cm	162.25±5.28	161.51±5.46	1.19	0.235
孕前体重Pre-pregnancy weight/kg	60.08±10.74	59.90±10.88	0.14	0.888
孕前BMI Pre-pregnancy BMI/(kg·m^-2)	22.80±3.87	22.99±4.25	0.43	0.669
孕次Gravidity			36.82	< 0.001
1	221(61.90)	87(94.57)
2	86(24.10)	5(5.43)
≥3	50(14.00)	0(0.00)
孕早期收缩压Early pregnancy systolic blood pressure/mmHg	108.16±18.11	106.82±16.51	0.59	0.552
孕早期舒张压Pregnancy diastolic blood pressure/mmHg	74.31±16.85	72.89±17.06	0.71	0.474
孕早期体重Early pregnancy weight/kg	62.71±10.65	62.22±10.85	0.39	0.696
文化程度Education level			2.48	0.478
研究生Postgraduate	35(9.80)	7(7.61)
大学本科Bachelor’s degree	143(40.06)	39(42.39)
大学专科Associate degree	104(29.13)	32(34.78)
高中及以下High school and below	75(21.01)	14(15.22)
民族Ethnicity			0.55	0.458
汉族Han	351(98.32)	92(100.00)
其他Other	6(1.68)	0(0.00)
孕前经期是否规律Pre-pregnancy menstrual regularity			0.06	0.809
是Yes	295(82.63)	77(83.70)
否No	62(17.37)	15(16.30)
月经平均持续天数Average menstrual duration/d			7.61	0.032
≤2	5(1.40)	1(1.09)
＞2~7	336(94.12)	80(86.95)
>7	16(4.48)	11(11.96)
经期的平均经血量Average menstrual blood volume/mL			2.11	0.347
< 20	20(5.60)	7(7.61)
＞20~60	307(86.00)	81(88.04)
>60	30(8.40)	4(4.35)
痛经的程度Dysmenorrhea severity			1.83	0.741
从来不疼No pain	54(15.13)	11(11.96)
轻度Mild	263(73.67)	70(76.08)
中度Moderate	36(10.08)	11(11.96)
重度Severe	4(1.12)	0(0.00)
注：①以人数(占比/%)或x±s表示。
Note：① Number of people (proportion/%) or x±s.

下载: 导出CSV

表 2 视网膜特征抑郁组与非抑郁组视网膜特征比较

Table 2. Comparison of retinal characteristics between the depressed group and the non-depressed group

视网膜变量特征 Retinal variable feature	非抑郁组 Non-depressed group^①	抑郁组 Depression group^①	t/x²值 value	P值 value
左眼视网膜中央动脉直径Left central retinal artery equivalent	13.55±0.62	13.60±0.62	-0.72	0.474
左眼视网膜中央静脉直径Left central retinal vein equivalent	20.69±0.78	20.76±0.82	-0.84	0.400
左眼静脉对称度Left asymmetry index of venules	1.68±0.08	1.68±0.09	0.64	0.519
左眼动脉对称度Left asymmetry index of arterioles	1.37±0.03	1.37±0.02	-0.98	0.329
左眼静脉分支角度Left bifurcation angle of venules	68.45±1.98	68.40±1.90	0.21	0.837
左眼动脉分支角度Left bifurcation angle of arterioles	74.11±2.05	74.03±1.72	0.37	0.709
左眼静脉分支系数Left bifurcation coefficient of venules	0.82±0.02	0.82±0.01	0.58	0.563
左眼动脉分支系数Left bifurcation coefficient of arterioles	0.77±0.01	0.77±0.01	-1.56	0.121
左眼视网膜动静脉比值Left artery/vein ratio	0.65±0.01	0.66±0.01	-0.05	0.958
左眼血管弯曲度Left vascular tortuosity	0.34±0.07	0.35±0.07	-0.83	0.404
左眼动静脉交叉点狭窄Left arteriovenous nipping	0.20±0.07	0.19±0.06	1.05	0.294
左眼视网膜眼底出血Left retinal hemorrhage	0.21±0.07	0.21±0.07	-0.29	0.774
左眼动脉闭塞Left arteriolar occlusion	0.11±0.06	0.11±0.06	0.67	0.504
左眼眼底渗出Left retinal exudates	0.17±0.07	0.16±0.07	1.30	0.195
右眼视网膜中央动脉直径Right central retinal artery equivalent	13.21±0.72	13.25±0.75	-0.49	0.626
右眼视网膜中央静脉直径Right central retinal vein equivalent	20.27±0.82	20.32±0.89	-0.50	0.620
右眼静脉对称度Right asymmetry index of venules	1.68±0.07	1.67±0.06	1.26	0.210
右眼动脉对称度Right asymmetry index of arterioles	1.35±0.03	1.35±0.03	-0.51	0.610
右眼静脉分支角度Right bifurcation angle of venules	69.95±2.00	69.36±2.17	2.47	0.014
右眼动脉分支角度Right bifurcation angle of arterioles	73.73±2.05	73.91±2.32	-0.77	0.443
右眼静脉分支系数Right bifurcation coefficient of venules	0.84±0.01	0.83±0.01	2.56	0.011
右眼小动脉分支系数Right bifurcation coefficient of arterioles	0.78±0.01	0.78±0.01	0.02	0.986
右眼视网膜动静脉比值Right artery/vein ratio	0.65±0.02	0.65±0.02	-0.28	0.778
右眼血管弯曲度Right vascular tortuosity	0.36±0.07	0.36±0.07	0.02	0.988
右眼动静脉交叉点狭窄Right arteriovenous nipping	0.26±0.09	0.26±0.09	0.59	0.557
右眼视网膜眼底出血Right retinal hemorrhage	0.22±0.09	0.22±0.09	0.89	0.373
右眼动脉闭塞Right arteriolar occlusion	0.09±0.05	0.10±0.05	-0.51	0.607
右眼眼底渗出Right retinal exudates	0.12±0.06	0.12±0.06	-0.43	0.668
注：①以x±s表示。
Note：① x±s.

下载: 导出CSV

参考文献(30)

[1]	Yu M, Li H, Xu DR, et al. Trajectories of perinatal depressive symptoms from early pregnancy to six weeks postpartum and their risk factors-a longitudinal study[J]. J Affect Disord, 2020, 275: 149-156. DOI: 10.1016/j.jad.2020.07.005.
[2]	Jahan N, Went TR, Sultan W, et al. Untreated depression during pregnancy and its effect on pregnancy outcomes: a systematic review[J]. Cureus, 2021, 13(8): e17251. DOI: 10.7759/cureus.17251.
[3]	杨业环, 黄星, 孙梦云, 等. 中国孕产妇不同时点抑郁状态转归及持续抑郁状态影响因素分析[J]. 中华流行病学杂志, 2022, 43(1): 58-64. DOI: 10.3760/cma.j.cn112338-20210628-00502. Yang YH, Huang X, Sun MY, et al. Analysis on depression state outcomes and influencing factors of persistent depression in pregnant and perinatal women in China[J]. Chin J Epidemio, 2022, 43(1): 58-64. DOI: 10.3760/cma.j.cn112338-20210628-00502.
[4]	卢永收. 孕早期孕妇抑郁状况及影响因素分析[J]. 妇儿健康导刊, 2024, 3(15): 195-199. Lu YS. Analysis of depression status and influencing factors of pregnant women in the early pregnancy[J]. Journal of Women and Children's Health Guide, 2024, 3(15): 195-199.
[5]	王瑞凤, 王静雯, 吴燕, 等. 不同孕期焦虑或抑郁对不良出生结局和婴儿身心发育影响的前瞻性研究[J]. 中国计划生育学杂志, 2024, 32(6): 1260-1265. DOI: 10.3969/j.issn.1004-8189.2024.06.008. Wang RF, Wang JW, Wu Y, et al. Influence of the anxiety and depression of pregnant women during different stages of pregnancy on their adverse pregnancy outcomes and their infant psychophysical development: a prospective study[J]. Chin J Fam Plann, 2024, 32(6): 1260-1265. DOI: 10.3969/j.issn.1004-8189.2024.06.008.
[6]	Sergi MR, Saggino A, Balsamo M, et al. Risk factors of the antenatal depression in a sample of Italian pregnant women: a preliminary study[J]. BMC Pregnancy Childbirth, 2024, 24(1): 689. DOI: 10.1186/s12884-024-06704-8.
[7]	王明欢, 李玉红, 袁德慧, 等. 妊娠早期女性非稳态负荷水平与抑郁的相关性研究[J]. 中国全科医学, 2023, 26(21): 2609-2613. DOI: 10.12114/j.issn.1007-9572.2023.0017. Wang MH, Li YH, Yuan DH, et al. Correlation between allostatic load level and depression among women in early pregnancy[J]. Chin Gen Pract, 2023, 26(21): 2609-2613. DOI: 10.12114/j.issn.1007-9572.2023.0017.
[8]	Zhang ZW, Deng C, Paulus YM. Advances in structural and functional retinal imaging and biomarkers for early detection of diabetic retinopathy[J]. Biomedicines, 2024, 12(7): 1405. DOI: 10.3390/biomedicines12071405.
[9]	Chaikijurajai T, Ehlers JP, Wilson Tang WH. Retinal microvasculature: a potential window into heart failure prevention[J]. JACC Heart Fail, 2022, 10(11): 785-791. DOI: 10.1016/j.jchf.2022.07.004.
[10]	Liu F, Chen XN, Wang Q, et al. Correlation between retinal vascular geometric parameters and pathologically diagnosed type 2 diabetic nephropathy[J]. Clin Kidney J, 2024, 17(8): sfae204. DOI: 10.1093/ckj/sfae204.
[11]	O'Neill RA, Maxwell AP, Kee F, et al. Association of reduced retinal arteriolar tortuosity with depression in older participants from the Northern Ireland cohort for the longitudinal study of ageing[J]. BMC Geriatr, 2021, 21(1): 62. DOI: 10.1186/s12877-021-02009-z.
[12]	Ha MJ, Han K, Jung Y, et al. Is retinal vein occlusion associated with depression symptoms? : a nationwide cohort study[J]. Medicine, 2021, 100(32): e26937. DOI: 10.1097/MD.0000000000026937.
[13]	Sadykov E, Hosak L, Stepanov A, et al. Retinal microvascular abnormalities in major depression[J]. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub, 2024, 168(2): 147-155. DOI: 10.5507/bp.2023.026.
[14]	Yang JC, Jiang Y, Qu YM, et al. Factor structure and longitudinal measurement invariance of Edinburgh postnatal depression scale during the whole perinatal period: a multicenter cohort study in China[J]. BMC Public Health, 2025, 25(1): 182. DOI: 10.1186/s12889-024-21213-1.
[15]	Huang LL, Shen Q, Fang QY, et al. Effects of Internet-based support program on parenting outcomes for primiparous women: a pilot study[J]. Int J Environ Res Public Health, 2021, 18(9): 4402. DOI: 10.3390/ijerph18094402.
[16]	Stén G, Ayers S, Malmquist A, et al. Assessment of maternal posttraumatic stress disorder following childbirth: Psychometric properties of the Swedish version of city birth trauma scale[J]. Psychol Trauma, 2023, 15(7): 1153-1163. DOI: 10.1037/tra0001465.
[17]	中华医学会妇产科学分会产科学组. 围产期抑郁症筛查与诊治专家共识[J]. 中华妇产科杂志, 2021, 56(8): 521-527. DOI: 10.3760/cma.j.cn112141-20210115-00022. Obstetrics Subgroup, Chinese Society of Obstetrics and Gynecology, Chinese Medical Association. Experts consensus on screening and diagnosis of perinatal depression[J]. Chin J Obstet Gynecol, 2021, 56(8): 521-527. DOI: 10.3760/cma.j.cn112141-20210115-00022.
[18]	Shi CY, Lee J, Wang GC, et al. Assessment of image quality on color fundus retinal images using the automatic retinal image analysis[J]. Sci Rep, 2022, 12(1): 10455. DOI: 10.1038/s41598-022-13919-2.
[19]	Gao Y, Xu LJ, He N, et al. A narrative review of retinal vascular parameters and the applications (part I): Measuring methods[J]. Brain Circ, 2023, 9(3): 121-128. DOI: 10.4103/bc.bc_8_23.
[20]	Matei N, Leahy S, Blair NP, et al. Retinal vascular physiology biomarkers in a 5XFAD mouse model of Alzheimer's disease[J]. Cells, 2022, 11(15): 2413. DOI: 10.3390/cells11152413.
[21]	Shi XH, Dong L, Zhang RH, et al. Relationships between quantitative retinal microvascular characteristics and cognitive function based on automated artificial intelligence measurements[J]. Front Cell Dev Biol, 2023, 11: 1174984. DOI: 10.3389/fcell.2023.1174984.
[22]	Karna S. Commentary: eye as a window to the brain[J]. Indian J Ophthalmol, 2020, 68(4): 563-564. DOI: 10.4103/ijo.IJO_2069_19.
[23]	Yamada R, Fujii T, Hattori K, et al. Discrepancy between clinician-rated and self-reported depression severity is associated with adverse childhood experience, autistic-like traits, and coping styles in mood disorders[J]. Clin Psychopharmacol Neurosci, 2023, 21(2): 296-303. DOI: 10.9758/cpn.2023.21.2.296.
[24]	Bai JH, Wan ZQ, Li P, et al. Accuracy and feasibility with AI-assisted OCT in retinal disorder community screening[J]. Front Cell Dev Biol, 2022, 10: 1053483. DOI: 10.3389/fcell.2022.1053483.
[25]	Wang Y, Li C, Liu L, et al. Association of retinal neurovascular impairment with disease severity in patients with major depressive disorder: an optical coherence tomography angiography study[J]. Psychol Res Behav Manag, 2024, 17: 1573-1585. DOI: 10.2147/PRBM.S443146.
[26]	Menon NJ, Sun C, Chhina J, et al. Cerebrovascular dysfunction and depressive symptoms in preclinical models: insights from a scoping review[J]. J Appl Physiol, 2024, 136(6): 1352-1363. DOI: 10.1152/japplphysiol.00031.2024.
[27]	陈淼, 孙静. 孕产期抑郁的识别与治疗进展[J]. 临床精神医学杂志, 2021, 31(2): 159-162. DOI: 10.3969/j.issn.1005-3220.2021.02.022. Chen M, Sun J. Recognition and treatment progress for depression during pregnant period[J]. J Clin Psychiatry, 2021, 31(2): 159-162. DOI: 10.3969/j.issn.1005-3220.2021.02.022.
[28]	Al Rawahi A, Al Kiyumi MH, Al Kimyani R, et al. The effect of antepartum depression on the outcomes of pregnancy and development of postpartum depression: a prospective cohort study of Omani women[J]. Sultan Qaboos Univ Med J, 2020, 20(2): e179-e186. DOI: 10.18295/squmj.2020.20.02.008.
[29]	罗鑫宇, 焦国硕, 修莉芸, 等. 孕期生活事件, 社会支持与孕妇心理健康和幸福感的关系[J]. 中国健康心理学杂志, 2020, 28(12): 1761-1766. DOI: 10.13342/j.cnki.cjhp.2020.12.001. Luo XY, Jiao GS, Xiu LY, et al. Effects of pregnancy life events and social support on the mental health and well-being of pregnant women[J]. China Journal of Health Psychology, 2020, 28(12): 1761-1766. DOI: 10.13342/j.cnki.cjhp.2020.12.001.
[30]	Kjeldsen MZ, Bricca A, Liu XQ, et al. Family history of psychiatric disorders as a risk factor for maternal postpartum depression: a systematic review protocol[J]. Syst Rev, 2022, 11(1): 68. DOI: 10.1186/s13643-022-01952-1.