第一部分：脉络膜血流量化分析在中心性浆液性脉络膜视网膜病变分型中的应用

目的：本研究旨在依据新的分类系统，将中心性浆液性脉络膜视网膜病变（CSC）患者分为急性、持续性、复发性和慢性四种亚型，同时纳入健康人群作为对照，利用扫频源光学相干断层扫描血管成像（SS-OCTA）对各亚型CSC的脉络膜血流特征进行定量分析，以探讨不同亚型CSC的病理生理机制异同。

方法：横断面观察性研究，共纳入103例CSC眼和29例健康对照眼，将所有CSC眼划分为急性CSC、持续性CSC、复发性CSC、慢性CSC四种亚型，记录所有患者的性别、年龄、高血压及糖尿病史、最佳矫正视力（BCVA）、非接触眼压测量（IOP）、裂隙灯检查、荧光素血管造影（FA）、吲哚菁绿血管造影（ICGA）及SS-OCTA检查结果。使用SS-OCTA测量视网膜中央厚度（CRT）、中心凹下脉络膜厚度（SFCT）、脉络膜大中血管体积（CVV）、脉络膜血管指数（CVI）、脉络膜毛细血管层密度（CCVD）以及黄斑区脉络膜大中血管扩张及侧支循环状态，采用ANOVA检验、Kruskal-Wallis H检验、卡方检验、SNK-q检验多种统计学方法进行多组间比较。

结果：在BCVA方面，急性、持续性和复发性CSC的视力显著差于慢性CSC组。在侧支循环状态方面，慢性CSC组中53.6%的患者表现为侧支循环形成型，而健康对照组中79.3%的患者表现为无扩张型。在脉络膜特征方面，四种亚型CSC的SFCT均显著高于健康对照组，且急性和复发性CSC的SFCT显著高于慢性CSC组；四种亚型CSC的CVI均显著高于健康对照组，且急性、持续性和复发性CSC的CVI均显著高于慢性CSC组；四种亚型CSC的CCVD均显著低于健康对照组，且急性、持续性和复发性CSC的CCVD均显著高于慢性CSC组。

结论：和其他亚型CSC相比，慢性CSC的脉络膜大中血管指数和脉络膜毛细血管密度显著更低，黄斑区脉络膜侧支循环形成比例较高。CVI和CCVD是量化脉络膜血管特征的有效影像学生物标志物，有助于揭示不同CSC亚型的病理生理机制。

第二部分：激光治疗中心性浆液性脉络膜视网膜病变的疗效评估及预测因素分析

目的：本研究旨在评估阈值下微脉冲激光（SML）治疗CSC的疗效及其作用机制，并探讨与SML治疗反应不佳相关的基线因素；同时评价对SML治疗反应不佳的CSC患者转换为689nm激光治疗的有效性和安全性，以探索潜在的新疗法。

方法：本研究为回顾性队列研究，纳入就诊于我院并接受统一模式SML治疗的58例CSC患者基线、治疗后1个月、治疗后3个月的临床资料，对于连续两次SML治疗后反应仍较差的32例患者，进一步分析其接受第三次SML治疗或转换为689nm激光治疗的效果。反应较差定义为治疗后3个月内视网膜下液（SRF）未完全消退，主要有效性指标为SRF的完全消退率，次要指标包括患者最佳矫正视力（BCVA）的变化，以及治疗后在SS-OCTA上观察的脉络膜血管指数（CVI）、脉络膜毛细血管密度（CCVD）的变化。

结果：接受SML治疗1个月后，有15例患者（25.9%）的SRF完全消退；接受SML治疗3个月后，有36例患者（62.1%）的SRF完全消退。经过比较，SRF完全消退组的基线CVI显著低于未完全消退组（0.48 ± 0.07 vs 0.53 ± 0.06，P = 0.015），SRF完全消退组的基线CCVD显著高于未完全消退组（52.43 ± 2.18 vs 49.83 ± 2.48，P < 0.01），多元Logistic回归分析显示，基线CCVD值较大与SRF完全消退显著相关（优势比为0.078，95%置信区间为0.032至0.124，P < 0.01）。对于SML疗效反应不佳的CSC患者，在3个月随访时，与SML组相比，转换为689nm激光治疗组的CRT（P = 0.028）、mSRF（P = 0.007）和SFCT（P = 0.014）均显著更低。

结论：基线CCVD较低是CSC患者SML治疗后SRF持续存在的独立危险因素。对于SML治疗反应不佳的CSC患者，689nm激光治疗是一种有前景的新选择，尽管视力改善不显著，但形态学结果是积极的，且SS-OCTA显示治疗后脉络膜血管指数和脉络膜厚度显著降低。

第三部分：基于脉络膜血流量化分析对于息肉状脉络膜血管病变的治疗反应预测模型

目的：通过评估PCV患者基线和在连续三次标准抗VEGF药物治疗后的临床特征变化，探究与治疗后功能学和形态学反应结局相关的影像学生物标志物，并建立准确性较高的疗效预测模型。

方法：回顾性队列研究，纳入94例接受连续三次标准化抗VEGF药物治疗的PCV患者在治疗前后的临床资料，使用SS-OCTA测量脉络膜特征，并应用AngioTool进行新生血管病灶的定量分析。本研究中功能学反应良好定义为治疗后BCVA增加≥ 5个ETDRS字母，不良反应定义为变化< 4个字母；根据第三次注射后视网膜下积液的变化情况，将所有眼划分为形态学反应良好组和不良组；根据治疗后血管面积相对变化值的中位数将所有眼划分为高收缩组和低收缩组。分别采用Logistic回归模型和最小绝对收缩与选择算子算法（LASSO）回归分析方法筛选变量，并构建列线图预测模型，分别绘制训练集和内部测试集的受试者工作特征（ROC）曲线、校准曲线和临床决策曲线评估模型性能。

结果：以功能学反应良好作为结局，多因素Logistic分析显示所有因素均与功能学反应结局无关；以形态学学反应良好作为结局，多因素Logistic分析显示基线时较高的新生血管平均空隙度（P = 0.028）、较高的SFCT（P = 0.018）与较高的CVI（P < 0.01）与形态学反应不良显著相关；以新生血管面积高收缩作为结局，多因素Logistic分析显示基线时较高的新生血管末端点总数（P = 0.022）与新生血管面积高收缩显著相关。基于单因素Logistic分析建立的列线图预测模型在训练集中的AUC为0.968（95% CI 0.955 - 0.980），测试集的AUC为0.917（95% CI 0.873 - 0.961）；基于LASSO回归分析建立的列线图预测模型在训练集中的AUC为0.983（95% CI 0.970 - 0.997），测试集的AUC为0.968（95% CI 0.950 - 0.986）。

结论：在PCV眼中，基线时较高的新生血管平均空隙度、较高的脉络膜厚度和脉络膜大中血管指数与形态学不良应答显著相关，而基线时较高的新生血管末端点总数与治疗后新生血管面积高收缩性显著相关。本研究构建的列线图预测模型在内部测试集中表现出较高的预测准确性，为PCV患者的个体化治疗提供了重要参考。

第四部分：基于马尔科夫模型评估息肉状脉络膜血管病变治疗的成本效益分析

目的：从社会角度构建一个基于马尔科夫模型的PCV患者假设队列，评估抗VEGF单药治疗、抗VEGF联合PDT治疗及基于预测模型导向的治疗策略的成本效益，为我国卫生决策提供经济学依据。

方法：本研究使用TreeAge Pro软件建立了PCV患者转归的马尔科夫模型，模拟中国PCV患者10年周期内的转归，从中国的社会角度评估和比较三种治疗策略的成本效益：抗VEGF单药治疗、抗VEGF联合PDT治疗、以及基于预测模型导向的治疗策略。本研究使用患者的效用值来计算质量调整生命年（QALYs），主要计算的评价结果是QALYs的增量成本效用比（ICURs），使用成本-效益可接受曲线（CEAC）计算在不同支付意愿阈值下，不同治疗策略的成本效益接受度。为了增加分析结果的稳健性，进一步对主要结果进行单向确定性和概率敏感性分析。

结果：在10年分析周期内，成本-效益可接受曲线显示在1倍人均GDP（95700元）的支付阈值下，抗VEGF单药治疗成本效益最优，其次是联合治疗，最后是基于预测模型导向的治疗。而在3倍人均GDP（287100元）的支付阈值下，基于预测模型导向的治疗成本效益最优，其次是联合治疗，最后是抗VEGF单药治疗。多种敏感性分析证实了结果的可靠性。

结论：在3倍人均GDP支付阈值下，抗VEGF联合PDT治疗的成本效益显著优于抗VEGF单药治疗，而基于预测模型导向的治疗策略表现出最佳的成本效益，为我国实际卫生决策提供了重要的经济学依据。

Part I. Application of Quantitative Choroidal Blood Flow Analysis in the Classification of Central Serous Chorioretinopathy

Objectives: This study aimed to classify patients with central serous chorioretinopathy (CSC) into acute, persistent, recurrent, and chronic subtypes based on a novel classification system, while including healthy controls as a reference group. Swept-source optical coherence tomography angiography (SS-OCTA) was employed to quantitatively analyze choroidal blood flow characteristics in each CSC subtype.

Methods: This cross-sectional observational study enrolled 103 CSC eyes and 29 healthy control eyes. CSC eyes were categorized into four subtypes: acute, persistent, recurrent, and chronic. Clinical data were recorded, including gender, age, history of hypertension and diabetes, best-corrected visual acuity (BCVA), intraocular pressure (IOP), slit-lamp examination, fluorescein angiography (FA), indocyanine green angiography (ICGA), and SS-OCTA. SS-OCTA was used to measure central retinal thickness (CRT), subfoveal choroidal thickness (SFCT), choroidal vascular volume (CVV), choroidal vascularity index (CVI), the vascular density of choriocapillaris (CCVD), and the status of macular choroidal large and medium vessel dilation and collateral circulation. ANOVA, Kruskal-Wallis H test, chi-square test, and SNK-q test were employed for intergroup comparisons.

Results: In terms of BCVA, acute, persistent, and recurrent CSC subtypes had significantly worse BCVA than the chronic CSC group. Regarding collateral circulation, 53.6% of chronic CSC eyes exhibited collateral circulation formation, while 79.3% of healthy control eyes showed no dilation. In terms of SFCT, all CSC subtypes had significantly higher values than the control group, with acute and recurrent CSC subtypes showing significantly higher SFCT than the chronic CSC group. Moreover, CVI was significantly higher in all CSC subtypes than in controls, with acute, persistent, and recurrent CSC subtypes showing significantly higher CVI than chronic CSC. CCVD was significantly lower in all CSC subtypes than in controls, with acute, persistent, and recurrent CSC subtypes showing significantly higher CCVD than chronic CSC.

Conclusions: Chronic CSC eyes exhibited significantly lower CVI and CCVD, along with a higher prevalence of macular choroidal collateral circulation. CVI and CCVD are reliable imaging biomarkers for quantifying choroidal vascular structural changes, providing insights into the pathophysiological mechanisms of different CSC subtypes.

Part II. Efficacy Evaluation and Predictive Factor Analysis of Laser Treatment for Central Serous Chorioretinopathy

Objectives: This study aimed to evaluate the efficacy and mechanisms of subthreshold micropulse laser (SML) treatment for CSC, identify baseline risk factors associated with poor treatment response, and assess the effectiveness and safety of switching to 689 nm laser therapy for patients with poor response to SML, thereby exploring potential new therapeutic options.

Methods: This retrospective cohort study included 58 CSC eyes treated with SML, with clinical data collected at baseline, 1 month, and 3 months after the treatment. Among them, 32 eyes with persistent subretinal fluid (SRF) after two SML treatments were further analyzed for their response to a third SML treatment or switching to 689 nm laser therapy. Poor response was defined as incomplete SRF absorption within 3 months after the second SML treatment. Primary efficacy outcomes included SRF absorption rate, while secondary outcomes included changes in BCVA, choroidal vascularity index (CVI), and the vascular density of choriocapillaris (CCVD) observed on SS-OCTA.

Results: At 1 month after the SML treatment, SRF was completely absorbed in 15 eyes (25.9%); at 3 months after the SML treatment, SRF absorption was achieved in 36 eyes (62.1%). The SRF absorption group had significantly lower baseline CVI (0.48 ± 0.07 vs 0.53 ± 0.06, P = 0.015) and higher baseline CCVD (52.43 ± 2.18 vs 49.83 ± 2.48, P < 0.01) than the non-absorption group. Multivariate logistic regression analysis showed that higher baseline CCVD was significantly associated with SRF absorption (OR = 0.078, 95% CI = 0.032 - 0.124, P < 0.01). For eyes with poor response to SML, the 689 nm laser therapy group exhibited significantly lower CRT (P = 0.028), the maximum height of SRF (P = 0.007), and SFCT (P = 0.014) than the SML group.

Conclusions: Lower baseline CCVD is an independent risk factor for persistent SRF after SML treatment. For CSC eyes with poor response to SML, 689 nm laser therapy is a promising alternative, demonstrating favorable morphological outcomes and significant reductions in CVI and SFCT on SS-OCTA, despite limited visual improvement.

Part III. Treatment Response Prediction Model for Polypoidal Choroidal Vasculopathy Based on Quantitative Choroidal Blood Flow Analysis

Objectives: This study aimed to identify imaging biomarkers associated with treatment responses by evaluating baseline and post-treatment characteristics of patients receiving three consecutive anti-VEGF injections, and construct an accurate prediction model.

Methods: This retrospective cohort study included 94 polypoidal choroidal vasculopathy (PCV) eyes treated with three standardized anti-VEGF injections. SS-OCTA was used to measure choroidal features, and AngioTool was employed for quantitative analysis of neovascular lesions. Good functional response was defined as a gain of ≥ 5 ETDRS letters, while poor functional response was defined as less than 4 ETDRS letters. Morphological response was defined based on changes in subretinal fluid observed on OCT. Eyes were further divided into high-shrinkage and low-shrinkage subgroups based on the median relative change in vascular area post-treatment. Logistic regression and least absolute shrinkage and selection operator (LASSO) regression methods were used for variables selection, and nomogram prediction models were further constructed. Receiver operating characteristic (ROC) curves, calibration curves and clinical decision curves were plotted for the training set and the internal test set to evaluate the model performance.

Results: Multivariate Logistic analysis showed that all factors were not related to functional response outcome. Multivariate Logistic analysis showed that higher mean lacunarity (P = 0.028), SFCT (P = 0.018) and CVI (P < 0.01) at baseline were significantly associated with poor morphological response. Multivariate Logistic analysis showed that higher total number of endpoints (P = 0.022) was significantly associated with high shrinkage of vascular area. The nomogram prediction model based on univariate logistic analysis demonstrated AUC values of 0.968 (95% CI: 0.955 - 0.980) and 0.917 (95% CI: 0.873 - 0.961) in the training and test sets, respectively. The nomogram prediction model based on LASSO regression analysis showed AUC values of 0.983 (95% CI: 0.970 - 0.997) and 0.968 (95% CI: 0.950 - 0.986) in the training and test sets, respectively.

Conclusions: In PCV eyes, higher mean lacunarity, higher SFCT and CVI at baseline were significantly associated with poor morphological response to anti-VEGF therapy, while higher total number of endpoints was significantly associated with high shrinkage of vascular area. Our nomogram prediction models exhibited high predictive accuracy in the internal test set, providing valuable guidance for individualized treatment of PCV patients.

Part IV. Cost-Effectiveness Analysis of Treatment Strategies for Polypoidal Choroidal Vasculopathy Based on Markov Model

Objectives: This study aimed to construct a hypothetical cohort of PCV patients from a societal perspective using a Markov model, and evaluate the cost-effectiveness of anti-VEGF monotherapy, anti-VEGF combined with photodynamic therapy (PDT), and prediction model-guided therapy strategy, and provide economic evidence for healthcare decision-making in China.

Methods: A Markov model was developed using TreeAge Pro software to simulate the outcomes of a hypothetical PCV patient cohort over 10-year horizon in China. The cost-effectiveness of the three strategies was evaluated from a societal perspective. Quality-adjusted life-years (QALYs) and incremental cost-utility ratios (ICURs) were calculated as the main results, and cost-effectiveness acceptability curves (CEACs) were generated to assess the acceptability of each strategy at different willingness-to-pay (WTP) thresholds. Sensitivity analyses were performed to validate the robustness of the results.

Results: Over a 10-year analysis period, the CEAC revealed that at a willingness-to-pay threshold of 1 time the per capita GDP (95700 yuan), anti-VEGF monotherapy demonstrated the optimal cost-effectiveness, followed by combination therapy, and finally, prediction model-guided therapy. At a threshold of 3 times the per capita GDP (287100 yuan), prediction model-guided therapy showed the best cost-effectiveness, followed by combination therapy, and then anti-VEGF monotherapy. Various sensitivity analyses confirmed the robustness of the results.

Conclusions: At a willingness-to-pay threshold of 3 times the per capita GDP, the cost-effectiveness of anti-VEGF combined with PDT therapy was significantly superior to anti-VEGF monotherapy, while the prediction model-guided treatment strategy exhibited the best cost-effectiveness. Our results might provide important evidence for practical health decision-making in China.