Stopyra Wiktor, Voytsekhivskyy Oleksiy, Grzybowski Andrzej
MW-med Eye Centre, Krakow, Poland; Department of Medicine, University of Applied Sciences, Nowy Targ, Poland.
Kyiv Clinical Ophthalmology Hospital Eye Microsurgery Center, Kyiv, Ukraine.
Can J Ophthalmol. 2025 Aug;60(4):200-207. doi: 10.1016/j.jcjo.2025.01.020. Epub 2025 Feb 26.
To compare accuracy of 7 artificial intelligence (AI)-based intraocular lens (IOL) power calculation formulas in medium-long eyes DESIGN: Retrospective observational study.
The data of patients with eyes with an axial length of 24.5 mm to 25.99 mm, who underwent phacoemulsification between May 2018 and September 2023 were analyzed. Inclusion criteria were complete biometric and refractive data. Exclusion criteria were intraoperative or postoperative complications, previous eye surgery, or corneal diseases, postoperative best-corrected visual acuity less than 0.8 and corneal astigmatism greater than 2.0 D. Prior to cataract surgery, IOL power was calculated using a Zeiss IOLMaster 700 (Carl Zeiss Meditec AG, Jena, Germany). The power of the implanted IOL was randomly selected from the outcomes of one of the following formulas, i.e., SRK/T, Holladay 2, or Barrett Universal II. Three months after phacoemulsification, refraction was measured. Postsurgery, IOL power calculations were performed using the following formulas: Hill-RBF 3.0, Kane, PEARL-DGS, Ladas Super Formula AI (LSF AI), Hoffer QST, Karmona, and Nallasamy. The main outcome measures used were SD, since prediction error (PE) was zeroed, and the percentage of eyes with PE was within ±0.50 D.
One hundred eighty-four eyes with axial lengths of between 24.50 mm and 25.97 mm, were studied. The Karmona formula obtained the lowest SD (0.322) just before Hill-RBF 3.0 (0.324) and Pearl-DGS (0.334), however, without statistical significance (p > 0.05). The highest percentage of eyes with PE within ±0.50 D was achieved by Karmona (89.67%) ahead of Hill-RBF 3.0, LSF AI and Pearl-DGS (all equally 87.50 each) without statistical significance (p > 0.490).
All studied AI-based formulas provided highly accurate outcomes in medium-long eyes.
比较7种基于人工智能（AI）的人工晶状体（IOL）屈光力计算公式在中长眼中的准确性。
回顾性观察研究。
分析2018年5月至2023年9月期间接受白内障超声乳化手术、眼轴长度在24.5 mm至25.99 mm之间的患者数据。纳入标准为完整的生物测量和屈光数据。排除标准为术中或术后并发症、既往眼部手术史、角膜疾病、术后最佳矫正视力低于0.8以及角膜散光大于2.0 D。在白内障手术前，使用蔡司IOLMaster 700（德国耶拿卡尔蔡司医疗技术股份公司）计算IOL屈光力。植入IOL的屈光力从以下公式之一的结果中随机选择，即SRK/T、Holladay 2或Barrett Universal II。白内障超声乳化术后3个月，测量屈光度数。术后，使用以下公式进行IOL屈光力计算：Hill-RBF 3.0、Kane、PEARL-DGS、Ladas Super Formula AI（LSF AI）、Hoffer QST、Karmona和Nallasamy。主要观察指标为标准差（SD），因为预测误差（PE）已归零，且PE在±0.50 D范围内的眼的百分比。
研究了184只眼轴长度在24.50 mm至25.97 mm之间的眼睛。Karmona公式获得的标准差最低（0.322），仅次于Hill-RBF 3.0（0.324）和Pearl-DGS（0.334），但无统计学意义（p>0.05）。PE在±0.50 D范围内的眼的百分比最高的是Karmona（89.67%），领先于Hill-RBF 3.0、LSF AI和Pearl-DGS（均为87.50%），无统计学意义（p>0.490）。
所有研究的基于AI的公式在中长眼中均提供了高度准确的结果。
