Polycystic ovary syndrome (PCOS) is the most common cause of anovulatory infertility and endocrinopathy among women of reproductive age, with a reported prevalence of 10%–13% (1). Polycystic ovary syndrome is diagnosed using the latest international evidence–based guideline criteria (1, 2), necessitating the presence of two out of the following: clinical or biochemical hyperandrogenism; ovulatory dysfunction; and polycystic ovaries on ultrasound or specific antimüllerian hormone (AMH) levels. This approach has identified at least four distinct phenotypes: phenotype A-androgen excess + ovulatory dysfunction + polycystic ovary morphology; phenotype B-androgen excess + ovulatory dysfunction; phenotype C-androgen excess + polycystic ovary morphology; and phenotype D-ovulatory dysfunction + polycystic ovary morphology (1, 2).

多囊卵巢综合征（PCOS）是导致育龄女性最常见的无排卵性不孕和内分泌疾病最常见原因，据报道发病率为10%—13%（1）。多囊卵巢综合征的诊断是根据最新的国际循证指南标准（1，2）进行的，需要满足以下两个条件之一：临床或生化高雄激素血症；排卵功能障碍；以及超声或特定抗米勒尔素（AMH）水平显示多囊卵巢。这种方法已经确定了至少四种不同的表型：表型A－雄激素过多+排卵功能障碍+多囊卵巢形态；表型B－雄激素过多+排卵功能障碍；表型C－雄激素过多+多囊卵巢形态；以及表型D-排卵功能障碍+多囊卵巢形态（1, 2）。

This definition, however, does not account for the severity of each of the aforementioned criteria, such as Ferrimen Galleway score, follicle count, or the extent of oligoanovulation. Furthermore, cofeatures such as ethnic differences, psychological status, quality of life, elevated basal luteinizing hormone levels, obesity, or metabolic syndrome (defined by three out of the five criteria: centripetal obesity; hypertension; fasting hyperglycemia; hypertriglyceridemia; and low high-density lipoprotein cholesterol levels) were not included in the PCOS diagnostic criteria.

然而，这一定义并未考虑到上述各项标准的严重程度，如Ferriman-Gallwey评分、卵泡计数或不排卵的程度。此外，种族差异、心理状态、生活质量、升高的基础黄体生成素水平、肥胖或代谢综合征（根据五个标准中的三个定义：中心性肥胖、高血压、空腹高血糖、高三酰甘油血症和低高密度脂蛋白胆固醇水平）等共同特征也未被纳入PCOS的诊断标准中。

We might, therefore, conclude that the clinical presentation of patients with PCOS is heterogeneous, ranging from eumenorrhea and sonographic evidence of polycystic ovaries without phenotypic abnormalities or signs of hyperandrogenism to advanced Stein and Leventhal syndrome and its associated long-term sequelae, namely, endometrial carcinoma, hypertension, diabetes mellitus, and cardiovascular disease. All of these make it challenging to define the specific individual components that might aid in tailoring an optimal treatment to a specific patient with PCOS (3).

因此，我们可能会得出结论，PCOS患者的临床表现是多种多样，从月经正常和超声检查显示多囊卵巢而无表型异常或高雄激素血症，到晚期Stein和Leventhal综合征及其相关的长期后遗症，即子宫内膜癌、高血压、糖尿病和心血管疾病。所有这些使得确定具体的个体组分以帮助定制针对特定PCOS患者的最佳治疗方案变得具有挑战性（3）。

The prediction of live birth in in vitro fertilization (IVF) cycles has been a focus of several models, which assess both naïve patients and those with previous failed IVF treatment attempts. Despite incorporating various predictive factors with different weights, only a few models have been validated externally (4). The retrospective study by Cooney et al. (5) delves into the crucial clinical parameters that should be studied to improve IVF success rates in patients with PCOS (5). They included 207 women with PCOS according to Rotterdam criteria from four academic reproductive endocrinology clinics who were undergoing their first fresh IVF cycles. Their objectives were to identify parameters associated with live birth after IVF in a cohort of PCOS women with infertility and then to employ those parameters to derive and validate internally a clinical prediction model for live birth in PCOS women undergoing IVF. This model, derived using multivariable logistic regression, included covariates on the basis of maximization of the area under the receiver operating characteristic curve. Their primary outcome was cumulative live birth rate per IVF cycle started, including FET cycles that were completed within 1 year of index oocyte retrieval. The final model included age < 35 years, White race, presence of polycystic ovaries on ultrasound (PCOM), normal BMI (<25 kg/m2), being healthy metabolically (free from metabolic risk factors), and being nonresponsive to ovulation induction agents including letrozole and clomiphene citrate. They believe that this model will better inform both physicians and their patients with PCOS about the optimal pathway to achieve a live birth.

体外受精（IVF）周期中活产的预测已成为几个模型的焦点，这些模型评估了未经治疗的患者以及之前IVF治疗失败的患者。尽管结合了各种预测因素的不同权重，但只有少数模型经过了外部验证（4）。Cooney等人的回顾性研究（5）深入研究了应研究的关键临床参数，以提高PCOS患者的IVF成功率（5）。根据鹿特丹标准，他们纳入了来自四个生殖内分泌诊所的207名接受首次新鲜IVF周期的PCOS女性。他们的目标是确定与IVF后活产相关的参数，并使用这些参数在PCOS女性不孕症患者队列中推导并内部验证一个临床预测模型。这个模型使用多变量逻辑回归，根据基于最大化接收者操作特征曲线下面积的协变量。他们的主要结果是每个IVF周期的累积活产率，包括在取卵后的一年内完成的冻胚移植（FET）周期。最终模型包括年龄小于35岁、白人种族、超声显示卵巢多囊样改变（PCOM）、正常BMI（小于25 kg/m²）、代谢健康（无代谢风险因素）以及对包括来曲唑和氯米芬在内的促排卵药物无反应。他们相信，这个模型将更好地告知医生和他们的PCOS患者关于实现活产的最佳途径。

IVF success rates vary significantly and are influenced by a wide range of factors, including patient characteristics, genetic considerations, and treatment protocols. This underscores the critical need for personalized approaches in predicting IVF outcomes. We applaud the efforts of Cooney et al. (5) for addressing this need, although it is clear that the clinical and comorbid features included in their analysis represent just a fraction of the complexity involved while dealing with patients with PCOS, with many more parameters of varying severity yet to be considered.

IVF成功率差异很大，受多种因素的影响，包括患者特征、遗传考虑和治疗方案。这强调了在预测IVF结果时采取个性化方法的重要。我们赞赏Cooney等人（5）为满足这一需求所做的努力。尽管很明显，他们分析中包括的临床和合并症特征，仅代表了处理PCOS患者时涉及的复杂性的一小部分，还有许多严重程度不同的参数需要考虑。

In this context, artificial intelligence (AI) has emerged as a game changer, revolutionizing the prediction of IVF success by leveraging patient characteristics in unprecedented ways. Artificial intelligence, particularly through machine learning algorithms and deep learning networks, processes vast amounts of data, identifying patterns and correlations that might elude human analysis. As AI technology continues to evolve, its potential to revolutionize reproductive medicine is immense. Moreover, the integration of AI models in the IVF process for patients with PCOS represents a paradigm shift toward more personalized, efficient, and outcome-focused fertility treatments, overcoming the heterogeneity that characterizes patients with PCOS. The future implementation of AI models might, therefore, provide clinicians and patients with evidence-based predictions, allowing for more informed discussions and offering personalized recommendations tailored to each individual’s unique profile.

在这种情况下，人工智能（AI）已成为改变游戏规则者，通过以前所未有的方式利用患者特征，彻底改变了IVF成功的预测。人工智能，特别是通过机器学习和深度学习网络，处理海量数据，识别可能被人类分析忽视的模式和相关性。随着AI技术的不断发展，其彻底改变生殖医学的潜力是巨大的。此外，将AI模型整合到PCOS患者的IVF过程中，代表了向更个性化、高效和更注重结果的生育治疗的模式转变，克服了PCOS患者的异质性特征。因此，AI模型的未来实施可能会为医生和患者提供基于证据的预测，允许进行更明智的讨论，并提供针对每个人独特的个性化建议。

参考文献：

Challenges encountered in predicting live birth after in vitro fertilization in the different phenotypes of patients with polycystic ovary syndrome Orvieto, Raoul Fertility and Sterility, Volume 121, Issue 6, 966 - 967