Investigating the Prediction Accuracy of Recently Updated Intraocular Lens Power Formulas with Artificial Intelligence for High Myopia

Comparing the accuracy of intraocular lens power calculation formulas using artificial intelligence and traditional formulas in highly myopic patients: a meta-analysis

PURPOSE

The accuracy of intraocular lens (IOL) calculations is one of the key indicators for determining the success of cataract surgery. However, in highly myopic patients, the calculation errors are relatively larger than those in general patients. With the continuous development of artificial intelligence (AI) technology, there has also been a constant emergence of AI-related calculation formulas. The purpose of this investigation was to evaluate the accuracy of AI calculation formulas in calculating the power of IOL for highly myopic patients.

METHODS

We searched the relevant literature through August 2023 using three databases: PubMed, EMBASE, and the Cochrane Library. Six IOL calculation formulas were compared: Kane, Hill-RBF, EVO, Barrett II, Haigis, and SRK/T. The included metrics were the mean absolute error (MAE) and percentage of errors within ± 0.25 D, ± 0.50 D, and ± 1.00 D.

RESULTS

The results showed that the MAE of Kane was significantly lower than that of Barrett II (mean difference = - 0.03 D, P = 0.02), SRK/T (MD = - 0.08 D, P = 0.02), and Haigis (MD = - 0.12 D, P < 0.00001). The percentage refractive prediction errors for Kane at ± 0.25 D, ± 0.50 D, and ± 1.00 D were significantly greater than those for SRK/T (P = 0.007, 0.003, and 0.01, respectively) and Haigis (P = 0.009, 0.0001, and 0.001, respectively). No statistically significant differences were noted between Hill-RBF and Barret, but Hill-RBF was significantly better than SRK/T and Haigis.

CONCLUSION

The AI calculation formulas showed more accurate results compared with traditional formulas. Among them, Kane has the best performance in calculating IOL degrees for highly myopic patients.

Investigation of the myopic outcomes of the newer intraocular lens power calculation formulas in Korean patients with long eyes

This study investigated the underlying causes of the myopic outcomes of the optic-based newer formulas (Barrett Universal II, EVO 2.0, Kane, Hoffer-QST and PEARL-DGS) in long Korean eyes with Alcon TFNT intraocular lens (IOL) implantation. Postoperative data from 3100 randomly selected eyes of 3100 patients were analyzed to compare the reference back-calculated effective lens positions (ELPs) based on the Haigis formula using conventional axial length (AL) and Cooke-modified AL (CMAL) with the predicted ELP of each single- and triple-optimized Haigis formula applied to AL- and CMAL. Contrary to the AL-applied Haigis formula, the predicted ELP curve of the CMAL-applied, single-optimized Haigis formula, simulating the methods of the newer formulas, exhibited a significant upward deviation from the back-calculated ELP in long eyes. The relationship between the AL and anterior chamber depth in our long-eyed population differed from that in the base population of the PEARL-DGS formula. The myopic outcomes in long eyes appeared to stem from the substantial overestimation of the postoperative IOL position with AL modification, leading to the implantation of inappropriately higher-powered IOLs. This discrepancy may be attributed to the ethnic differences in ocular biometrics, particularly the relatively smaller anterior segment in East Asian patients with long AL.

How to choose the intraocular lens power calculation formulas in eyes with extremely long axial length? A systematic review and meta-analysis

OBJECTIVE

To evaluate the accuracy of 10 formulas for calculating intraocular lens (IOL) power in cataract eye with an axial length (AL) of more than 28.0 mm.

METHODS

We searched scientific databases including PubMed, EMBASE, Web of Science and Cochrane Library for research published over the past 5 years, up to Sept 2023. The inclusion criteria were case series studies that compared different formulas (Barrett II, EVO, Kane, Hill-RBF, Haigis, Hoffer Q, Holladay 1, SRK/T, Holladay 1 w-k and SRK/T w-k), in patients with extremely long AL undergoing uncomplicated cataract surgery with IOL implantation. The mean difference (MD) of mean absolute error (MAE) and the odds ratio (OR) of both the percentage of eyes within ±0.50D of prediction error (PPE±0.50D) and the percentage of eyes within ±1.00D of prediction error (PPE±1.00D) among different formulas were pooled using meta-analysis.

RESULTS

A total of 11 studies, involving 1376 eyes, were included to evaluate the 10 formulas mentioned above. Among these formulas, Barrett II, EVO, Kane, and Hill-RBF demonstrated significantly lower MAE values compared to SRK/T. Furthermore, Kane and Hill-RBF had lower MAE values than EVO. Additionally, Barrett II and Kane yielded significantly lower MAE values than Haigis while Hill-RBF showed significantly lower MAE values than Holladay 1. Moreover, Hill-RBF showed the highest values for both PPE±0.50D and PPE±1.00D, followed by Kane. Both EVO and Kane had higher values of PPE±0.50D and PPE±1.00D compared to Haigis and SRK/T.

CONCLUSION

The Wang-Koch adjusted formulas and new-generation formulas have shown potential for higher accuracy in predicting IOL power for cataract patients with extremely long AL compared to traditional formulas. Based on the current limited clinical studies, Hill-RBF and Kane formulas seem to be a better choice for eyes with extremely long AL.

Insights into artificial intelligence in myopia management: from a data perspective

Given the high incidence and prevalence of myopia, the current healthcare system is struggling to handle the task of myopia management, which is worsened by home quarantine during the ongoing COVID-19 pandemic. The utilization of artificial intelligence (AI) in ophthalmology is thriving, yet not enough in myopia. AI can serve as a solution for the myopia pandemic, with application potential in early identification, risk stratification, progression prediction, and timely intervention. The datasets used for developing AI models are the foundation and determine the upper limit of performance. Data generated from clinical practice in managing myopia can be categorized into clinical data and imaging data, and different AI methods can be used for analysis. In this review, we comprehensively review the current application status of AI in myopia with an emphasis on data modalities used for developing AI models. We propose that establishing large public datasets with high quality, enhancing the model's capability of handling multimodal input, and exploring novel data modalities could be of great significance for the further application of AI for myopia.

Efficacy of corneal curvature on the accuracy of 8 intraocular lens power calculation formulas in 302 highly myopic eyes

PURPOSE

To investigate the effect of corneal curvature (K) on the accuracy of 8 intraocular lens formulas in highly myopic eyes.

SETTING

Eye Hospital and School of Ophthalmology and Optometry, Wenzhou Medical University, Wenzhou, China.

DESIGN

Retrospective consecutive case series.

METHODS

302 eyes (302 patients) were analyzed in subgroups based on the K value. The mean refractive error, mean absolute error (MAE), median absolute error (MedAE), root-mean-square absolute prediction error (RMSAE) and proportions of eyes within ±0.25 diopter (D), ±0.50 D, ±0.75 D, ±1.00 D were statistical analyzed.

RESULTS

Emmetropia Verifying Optical (EVO) 2.0, Kane, and Radial Basis Function (RBF) 3.0 had the lower MAE (≤0.28) and RMSAE (≤0.348) and highest percentage of eyes within ±0.50 D (≥83.58%) in the flat (K ≤ 43 D) and steep K (K > 45 D) groups. Hoffer QST had the lowest MedAE (0.19), RMSAE (0.351) and the highest percentage of eyes within ±0.50 D (82.98%) in the normal K group (43 < K ≤ 45 D). When axial length (AL) ≤28 mm, all formulas showed close RMSAE values (0.322 to 0.373) in flat K group. When AL >28 mm, RBF 3.0 achieved the lowest MAE (≤0.24), MedAE (≤0.17) and RMSAE (≤0.337) across all subgroups.

CONCLUSIONS

EVO 2.0, Kane, and RBF 3.0 were the most accurate in highly myopic eyes with a flat or steep K. Hoffer QST is recommended for long eyes with normal K values. RBF 3.0 showed the highest accuracy when AL >28 mm, independent of corneal curvature.

Comparison of the accuracy of 9 intraocular lens power calculation formulas after SMILE in Chinese myopic eyes

As of 2021, over 2.8 million small-incision lenticule extraction (SMILE) procedures have been performed in China. However, knowledge regarding the selection of intraocular lens (IOL) power calculation formula for post-SMILE cataract patients remains limited. This study included 52 eyes of 26 myopic patients from northern China who underwent SMILE at Tianjin Eye Hospital from September 2022 to February 2023 to investigate the suitability of multiple IOL calculation formulas in post-SMILE patients using a theoretical surgical model. We compared the postoperative results obtained from three artificial intelligence (AI)-based formulas and six conventional formulas provided by the American Society of Cataract and Refractive Surgery (ASCRS). These formulas were applied to calculate IOL power using both total keratometry (TK) and keratometry (K) values, and the results were compared to the preoperative results obtained from the Barrett Universal II (BUII) formula for the SMILE patients. Among the evaluated formulas, the results obtained from the Emmetropia Verifying Optical 2.0 Formula with TK (EVO-TK) (0.40 ± 0.29 D, range 0-1.23 D), Barrett True K with K formula (BTK-K, 0.41 ± 0.26 D, range 0.01-1.19 D), and Masket with K formula (Masket-K, 0.44 ± 0.33 D, range 0.02-1.39 D) demonstrated the closest proximity to BUII. Notably, the highest proportion of prediction errors within 0.5 D was observed with the BTK-K (71.15%), EVO-TK (69.23%), and Masket-K (67.31%), with the BTK-K showing a significantly higher proportion than the Masket-K (p < 0.001). Our research indicates that in post-SMILE patients, the EVO-TK, BTK-K, and Masket-K may yield more accurate calculation results. At their current stage in development, AI-based formulas do not demonstrate significant advantages over conventional formulas. However, the application of historical data can enhance the performance of these formulas.

Effect of Segmented Optical Axial Length on the Performance of New-Generation Intraocular Lens Power Calculation Formulas in Extremely Long Eyes

PURPOSE

To evaluate the performance of traditional vergence formulas with segmented axial length (AL) compared to traditional composite AL in extremely long eyes, and to determine whether the segmented AL can be extended to the new-generation formulas, including the Barrett Universal II, Emmetropia Verifying Optical 2.0 (EVO2), Hill-RBF 3.0 (Hill3), Kane, and Ladas Super formula (LSF) formulas in extremely long eyes.

SETTING

National Hospital. Organization, Tokyo Medical Center, Japan.

DESIGN

Retrospective case series.

METHODS

Consecutive patients who underwent uncomplicated cataract surgery implanted with a three-piece intraocular lens between December 2015 and March 2021 were retrospectively reviewed. The composite AL was measured with a swept-source optical coherence tomography (SS-OCT) biometer using a mean refractive index. The segmented AL was calculated by summing the geometric lengths of the ocular segments (cornea, aqueous, lens, and vitreous) using multiple specific refractive indices based on the data obtained by the SS-OCT-based biometer. When refraction was measured at three months postoperatively, the median absolute errors (MedAEs) were calculated with two ALs for each formula.

RESULTS

The study included 31 eyes of 22 patients. The segmented AL (30.45 ± 1.23 mm) was significantly shorter than the composite AL (30.71 ± 1.28 mm, p < 0.001). The MedAEs were significantly reduced when using segmented AL for SRK/T, Haigis, Hill3, and LSF, compared to those obtained using composite AL (0.38 vs. 0.62, 0.48 vs. 0.79, 0.50 vs. 0.90, 0.34 vs. 0.61, p < 0.001 for all formulas, respectively). On the contrary, the MedAE obtained by Kane with segmented AL was significantly worse compared to the one with composite AL (0.35 vs. 0.27, p = 0.03).

CONCLUSION

In extremely high myopic eyes, the segmented AL improves the performance of SRK/T, Haigis, Hill3, and LSF formulas compared to the composite AL, while the segmented AL worsens the prediction accuracy of the Kane formula.

Addressing Glaucoma in Myopic Eyes: Diagnostic and Surgical Challenges

Epidemiological and genetic studies provide strong evidence supporting an association between myopia and glaucoma. The accurate detection of glaucoma in myopic eyes, especially those with high myopia, remains clinically challenging due to characteristic morphologic features of the myopic optic nerve in addition to limitations of current optic nerve imaging modalities. Distinguishing glaucoma from myopia is further complicated by overlapping perimetric findings. Therefore, longitudinal follow-up is essential to differentiate progressive structural and functional abnormalities indicative of glaucoma from defects that may result from myopia alone. Highly myopic eyes are at increased risk of complications from traditional incisional glaucoma surgery and may benefit from newer microinvasive glaucoma surgeries in select cases.

Accuracy of Seven Modern Online IOL Formulas in Eyes With Axial Lengths Longer Than 30 mm

PURPOSE

To evaluate the accuracy of newer online intraocular lens (IOL) formulas in extremely elongated eyes (axial length > 30 mm).

METHODS

This retrospective case series study included 236 patients (236 eyes). Postoperative refractive outcomes of the Barrett Universal II (BU II), Cooke K6 (K6), Emmetropia Verifying Optical (EVO) 2.0, Hoffer QST (HQST), Kane, Pearl-DGS, and Radial Basis Function (RBF) 3.0 formulas were compared. Subgroup analysis was performed in the extreme myopia group 1 (30 < axial length ≤ 32 mm), extreme myopia group 2 (32 < axial length ≤ 35 mm), and meniscus IOL group. The root mean square absolute prediction error (RMSAE) and proportions of eyes of prediction errors within ±0.50 diopters (D) were calculated for statistical analysis.

RESULTS

For the extreme myopia group 1, RBF 3.0 achieved the lowest RMSAE (0.361) and EVO 2.0 showed the highest proportion of eyes within ±0.50 diopters (85.06%). For the extreme myopia group 2, the RMSAE of the K6 (0.442) and EVO 2.0 (0.475) was significantly lower than the BU II (0.610), Kane (0.641), and HQST (0.759, P ≤ .016) formulas. In the meniscus IOL group, the K6 formula showed the lowest RMSAE (0.402) and the highest percentage within ±0.50 diopters (84.31%).

CONCLUSIONS

The EVO 2.0 and K6 formulas are recommended for IOL power calculation in eyes with extreme myopia. Modern artificial intelligence-based formulas should be used cautiously when the axial length is longer than 32 mm or meniscus IOLs are implanted. [J Refract Surg. 2023;39(10):705-710.].

Advances in artificial intelligence models and algorithms in the field of optometry

The rapid development of computer science over the past few decades has led to unprecedented progress in the field of artificial intelligence (AI). Its wide application in ophthalmology, especially image processing and data analysis, is particularly extensive and its performance excellent. In recent years, AI has been increasingly applied in optometry with remarkable results. This review is a summary of the application progress of different AI models and algorithms used in optometry (for problems such as myopia, strabismus, amblyopia, keratoconus, and intraocular lens) and includes a discussion of the limitations and challenges associated with its application in this field.