人工智能(artificial intelligence，AI)在白内障手术中的应用越来越广泛，二者结合对于白内障手术的术前诊断和分级管理、术中人工晶状体选择、位置预测及术后管理(视力预测、并发症预测及随访)、手术培训和教学方面均起到巨大的促进作用。诚然，AI在与白内障手术相关的管理、分析和研究中还面临着许多问题，但其广泛的应用前景不可忽视。现对AI在白内障手术治疗和教学中的应用做以总结，并对其未来的发展做出展望。

Artificial intelligence (AI) has been widely used in cataract surgery. The combination of the two can play a great role in improving preoperative diagnosis, grading management of cataract surgery, intraoperative intraocular lens selection and location prediction, postoperative management (vision prediction, complication prediction and follow-up), surgical training and teaching. It is true that AI still faces many problems in the management, analysis and research related to cataract surgery, but its broad application prospects cannot be ignored. This review summarizes the application of AI in cataract surgery and teaching, and the future prospects of AI.

目的：在硅油取出联合白内障手术患者中，使用扫频源光学相干断层扫描生物测量仪OA-2000进行生物测量，比较10种人工晶状体(IOL)屈光力计算公式的准确性。方法：回顾性分析2021年3月—7月于中山大学中山眼科中心接受硅油取出联合白内障手术的患者共62例(62眼)，所有患者均使用扫频源光学相干断层扫描生物测量仪OA-2000进行生物学参数测量。计算并比较新公式[Barrett Universal II (BUII)、Emmetropia Verifying Optical(EVO) 2.0、Hill-Radial Basis Function (Hill-RBF) 3.0、Hoffer QST、Kane、Pearl-DGS]及传统公式(Haigis、Hoffer Q、Holladay 1、SRK/T)的预测准确性，主要评价指标为绝对预测误差中位数(MedAE)及平均绝对预测误差(MAE)。按眼轴长度≤23 mm(组1)，>23 mm且≤26 mm(组2)与>26 mm(组3)进行亚组分析。结果：6个新公式、Haigis、SRK/T公式均出现近视漂移(-0.47 ~-0.27 D，P<0.05)，而HofferQ及Holladay 1公式无系统误差(P>0.05)。Kane公式的MedAE(0.55 D)及MAE(0.81 D)最小，但公式间比较差异无统计学意义(P>0.05)。组1中所有公式均出现近视漂移(-1.46~ -1.25 D，P<0.05)，而其他亚组比较差异无统计学意义(-0.32 ~ 0.41 D，P>0.05)。在组1中，Pearl-DGS公式的MedAE(0.97 D)及MAE(1.26 D)最小，且优于Hill-RBF 3.0(P=0.01)及SRK/T公式(P=0.02)；组2中，Kane公式具有最小的MedAE(0.44 D)及MAE(0.66 D)；组3各个公式屈光预测准确性比较差异无统计学意义(P>0.05)。结论：在使用OA-2000进行术前生物测量时，Kane公式在接受硅油取出联合白内障手术患者中的预测准确性较高；而眼轴长度≤23 mm时，Pearl-DGS公式可能更为准确。

Objective: To compare the accuracy of 10 intraocular lens (IOL) power calculation formulas in patients undergoing combined silicone oil removal and cataract surgery, biometry is performed using the swept-source optical coherence tomography biometer OA-2000. Methods: A retrospective analysis. A total of 62 patients (62 eyes) who underwent combined silicone oil removal and cataract surgery in Zhongshan Ophthalmic Center, Sun Yat-sen University from March to July in 2021 were enrolled. Preoperative biometry was performed by OA-2000 in all patients. New-generation formulas (Barrett Universal II [BUII], Emmetropia Verifying Optical [EVO] 2.0, Hill-Radial Basis Function [Hill-RBF] 3.0, Hoffer QST, Kane and Pearl-DGS) and traditional formulas (Haigis, Hoffer Q, Holladay 1 and SRK/T) were evaluated. The median absolute prediction error (MedAE) and mean absolute prediction error (MAE) were the main parameters used to assess accuracy. Subgroup analyses were performed based on the axial length of 23 mm and 26 mm. Results: Six new-generation formulas, Haigis, and SRK/T showed myopic shift (-0.47 ~ -0.27 D, P<0.05), while no systematic bias was found in Hoffer Q and Holladay 1 displayed (P>0.05). The smallest MedAE (0.55 D) and MAE (0.81 D) were found in Kane formula, but there was no statistically significant difference compared with other formulas (P>0.05). The myopic shift (-1.46 ~ -1.25 D, P<0.05) in eyes shorter than 23 mm were found in all formulas, while there was no significant systematic bias (-0.32 ~ 0.41 D, P>0.05) in other subgroups. In axial length shorter than 23 mm, the Pearl-DGS formula stated the smallest MedAE (0.97 D) and MAE (1.26 D), and was significantly more accurate than Hill-RBF 3.0 (P=0.01) and SRK/T (P=0.02). In eyes with an axial length between 23 mm and 26 mm, the Kane formula had the lowest MedAE (0.44 D) and MAE (0.66 D). No significant difference was found in eyes longer than 26 mm. Conclusion: The Kane formula showed the highest accuracy in patients undergoing combined silicone oil removal and cataract surgery measured by OA-2000, whereas the Pearl-DGS formula could be more accurate in eyes with an axial length shorter than 23 mm.

目的：比较单一EYESi虚拟手术模拟器(Dry-lab)、Wet-lab以及两种方式联合教学在超声乳化白内障吸除显微手术培训中的效果及差异，以期探索更科学高效的教学方式。方法：选取中山大学中山眼科中心接受住院医师规范化培训的1年级住院医师18名，随机分为Dry-lab组、Wet-lab组和联合组，每组各6人，分别接受8次有效Dry-lab训练、8次Wet-lab训练、4次有效Dry-lab训练联合4次Wet-lab训练。培训前后问卷调查评估三种教学方式的模拟效果与学员满意度，并在猪眼模型上考核超声乳化白内障吸除手术的三个主要步骤，以评估学员的培训效果。结果：Dry-lab与Wet-lab训练均能有效帮助学员学习使用显微器械，操作手感较好。但在立体感(P=0.007)、与人眼操作相比近似度(P<0.001)以及对掌握技术的帮助度(P=0.003)上，Wet-lab优于Dry-lab训练；而在缩短培训用时(P<0.001)上，Dry-lab培训更具优势。联合培训模式培训效果优于单一Wet-lab训练(P=0.014)和模拟器培训(P=0.012)，整体满意度高于Wet-lab训练(P=0.042)和Dry-lab培训(P=0.042)。结论：Dry-lab与Wet-lab训练在超声乳化白内障吸除显微手术培训中各有优势，而两者相结合的教学模式更为高效，培训效果更佳，整体满意度高。

Objective: To compare the effects and distinctions among three methods of phacoemulsification training: EYESi simulator (Dry-lab), Wet-lab, and a combined approach, in order to find out more scientific and efficient teaching method. Methods: 18 first-year residents undergoing residency training at Zhongshan Ophthalmic Center, Sun Yatsen University were randomly assigned to three groups: Dry-lab, Wet-lab, and Combined. Each group, consisting of 6 individuals, underwent a specific training regimen—8 sessions of effective simulator training for the Dry-lab group, 8 Wet-lab sessions for the Wet-lab group, and a combination of 4 effective simulator sessions with 4 Wet-lab sessions for the Combined group. The questionnaires were administered to assess simulation effects and student satisfaction before and after each training sessions. Additionally, the training effects were recorded in the three main steps of phacoemulsification cataract extraction surgery on a pig eye model. Results: Both the EYESi simulator and Wet-lab were proved to be effective in facilitating the learning of microscopic instrument use with commendable operating experience. However, Wet-lab is superior to the virtual simulator in terms of stereoscopic sensation (P=0.007), similarity to the human eye (P<0.001), and assistance in mastering techniques(P=0.003). Simulator training is found to be more advantageous in shortening the overall training time (P<0.001). The training effect of the joint training mode is better than that of single Wet-lab training (P=0.014) and simulator training (P=0.012), and the overall satisfaction is higher than that of Wet-lab training (P=0.042) and simulator training (P=0.042). Conclusion: The EYESi virtual surgery simulator and Wet-lab training have their advantages respectively in training for phacoemulsification cataract extraction microsurgery, and the combined teaching mode is more efficient, with better training effects and overall satisfaction.

目的：了解白内障手术患者的临床分布特点，为基层医院白内障防治工作提供信息参考。方法：选取2019年1月至2019年1 2月在锦州医科大学附属第三医院眼科手术医治的白内障患者892例1 008眼。查阅相关病历资料，收集姓名、性别、年龄、手术眼别、入院时间、有无糖尿病史；术前视力、角膜内皮细胞密度、眼轴、晶状体混浊类型及白内障病因；手术方式、人工晶状体(intraocular lens，IOL)、术中并发症；术后第1天眼压、视力、眼部情况，进行统计学描述及分析。结果：入 选892例1 008眼，其中男357例401眼，女535例607眼，年龄(71.17±10.28)岁，90.7%的患者病因为年龄相关性白内障。入院时间集中在3~6月和9~12月。不同年龄段白内障患者具有各自的眼部特征。白内障类型与年龄差异有统计学意义(P<0.05)，相比皮质性、后囊膜下性，核性白内障患者入院年龄偏大。白内障类型与性别、有无糖尿病史差异无统计学意义(P >0.05)。白内障患者选择最多的手术方式是超声乳化吸除术联合IOL植入术，术中并发症少，术后恢复快。结论：白内障的防治工作任重道远。手术多为年龄相关性白内障患者，入院时多年龄偏大，视力偏低，错失了手术最佳时机。不同年龄段、不同类型白内障患者具有不同的特征。超声乳化吸除术因其优势成为患者首选手术方式。

Objective: To understand the clinical distribution characteristics of cataract surgery patients and provide reference for cataract prevention and treatment in primary hospitals. Methods: From January 2019 to December 2019, 892 cataract patients (1 008 eyes) were selected from the Department of Ophthalmology, Third Affiliated Hospital of Jinzhou Medical University. Medical records were collected, including patient’s name, gender, age, surgical eye, admission time and history of diabetes, preoperative vision, corneal endothelial cell density, eye axis, phacoscotasmus type and cataract etiology, surgery method, intraocular lens, intraoperative complications, intraocular pressure, visual acuity and eye recovery on the first day after operation. These data were statistically described and analyzed. Results: A total of 892 cataract patients (1 008 eyes) were included, including 357 males (401 eyes) and 535 females (607 eyes), with the average age of (71.17±10.28) years old. The 90.7% patients cause of disease was age-related cataract. Admission time of all patients was mainly in March–June and September– December. Cataract patients of different ages had their own ocular characteristics. There was a statistically significant difference between the type of cataract and age (P<0.05). Compared with the cortical patients and posterior subcapsular patients, nuclear cataract patients were older at admission. There was no significant difference in the type of cataract between the gender and the patients with diabetes or not (P>0.05). The most frequently selected surgical method for cataract patients was phacoemulsification and intraocular lens implantation, which has the advantages of less intraoperative complications and rapid postoperative recovery. Conclusion: We still have a long way to go in the prevention and treatment of cataract. Most of the operations were performed on age-related cataract patients, and most of them were older and had low vision when they were admitted to hospital, so the optimal timing of the operation was missed. Patients of different age groups and different types of cataract have different characteristics. Phacoemulsification is the first choice because of its own advantages.