Machine learning adaptation of intraocular lens power calculation for a patient group

Abstract Background To examine the effectiveness of the use of machine learning for adapting an intraocular lens (IOL) power calculation for a patient group. Methods In this retrospective study, the clinical records of 1,611 eyes of 1,169 Japanese patients who received a single model of monofocal IOL (SN60WF, Alcon) at Miyata Eye Hospital were reviewed and analyzed. Using biometric metrics and postoperative refractions of 1211 eyes of 769 patients, constants of the SRK/T and Haigis formulas were optimized. The SRK/T formula was adapted using a support vector regressor. Prediction errors in the use of adapted formulas as well as the SRK/T, Haigis, Hill-RBF and Barrett Universal II formulas were evaluated with data from 395 eyes of 395 distinct patients. Mean prediction errors, median absolute errors, and percentages of eyes within ± 0.25 D, ± 0.50 D, and ± 1.00 D, and over + 0.50 D of errors were compared among formulas. Results The mean prediction errors in the use of the SRT/K and adapted formulas were smaller than the use of other formulas (P < 0.001). In the absolute errors, the Hill-RBF and adapted methods were better than others. The performance of the Barrett Universal II was not better than the others for the patient group. There were the least eyes with hyperopic refractive errors (16.5%) in the use of the adapted formula. Conclusions Adapting IOL power calculations using machine learning technology with data from a particular patient group was effective and promising.

Download Full-text

Use of a Machine Learning Method in Predicting Refraction after Cataract Surgery

Journal of Clinical Medicine ◽

10.3390/jcm10051103 ◽

2021 ◽

Vol 10 (5) ◽

pp. 1103

Author(s):

Tomofusa Yamauchi ◽

Hitoshi Tabuchi ◽

Kosuke Takase ◽

Hiroki Masumoto

Keyword(s):

Machine Learning ◽

Cataract Surgery ◽

Test Data ◽

Training Data ◽

Gradient Boosting ◽

Support Vector ◽

Power Calculation ◽

Iol Power Calculation ◽

Significant Difference ◽

Iol Power

The present study aims to describe the use of machine learning (ML) in predicting the occurrence of postoperative refraction after cataract surgery and compares the accuracy of this method to conventional intraocular lens (IOL) power calculation formulas. In total, 3331 eyes from 2010 patients were assessed. The objects were divided into training data and test data. The constants for the IOL power calculation formulas and model training for ML were optimized using training data. Then, the occurrence of postoperative refraction was predicted using conventional formulas, or ML models were calculated using the test data. We evaluated the SRK/T formula, Haigis formula, Holladay 1 formula, Hoffer Q formula, and Barrett Universal II formula (BU-II); similar to ML methods, we assessed support vector regression (SVR), random forest regression (RFR), gradient boosting regression (GBR), and neural network (NN). Among the conventional formulas, BU-II had the lowest mean and median absolute error of prediction. Therefore, we compared the accuracy of our method with that of BU-II. The absolute errors of some ML methods were lower than those of BU-II. However, no statistically significant difference was observed. Thus, the accuracy of our method was not inferior to that of BU-II.

Download Full-text

Assessment of the Influence of Keratometry on Intraocular Lens Calculation Formulas in Long Axial Length Eyes

10.21203/rs.3.rs-639671/v1 ◽

2021 ◽

Author(s):

Shengjie Yin ◽

Chengyao Guo ◽

Kunliang Qiu ◽

Tsz Kin Ng ◽

Yuancun Li ◽

...

Keyword(s):

Intraocular Lens ◽

Axial Length ◽

Case Series ◽

Absolute Error ◽

Power Calculation ◽

Prediction Errors ◽

Retrospective Case ◽

Iol Power Calculation ◽

Iol Power ◽

Intraocular Lens Calculation

Abstract Purpose: Hyperopic surprises tend to occur in axial myopic eyes and other factors including corneal curvature have rarely been analyzed in cataract surgery, especially in eyes with long axial length (≥ 26.0 mm). Thus, the purpose of our study was to evaluate the influence of keratometry on four different formulas (SRK/T, Barrett Universal II, Haigis and Olsen) in intraocular lens (IOL) power calculation for long eyes.Methods: Retrospective case-series. 180 eyes with axial length (AL) ≥ 26.0 mm were divided into 3 keratometry (K) groups: K ≤ 42.0 D (Flat), K ≥ 46.0 D (Steep), 42.0 < K < 46.0 D (Average). Prediction errors (PE) were compared between different formulas. Multiple regression analysis was performed to investigate factors associated with the PE.Results: The mean absolute error was higher for all evaluated formulas in Steep group (ranging from 0.66 D to 1.02 D) than the Flat (0.34 D to 0.67 D) and Average groups (0.40 D to 0.74D). The median absolute errors predicted by Olsen formula were significantly lower than that predicted by Haigis formula (0.42 D versus 0.85 D in Steep and 0.29 D versus 0.69 D in Average) in Steep and Average groups (P = 0.012, P < 0.001, respectively). And the Olsen formula demonstrated equal accuracy to the Barrett II formula in Flat and Average groups. The predictability of the SRK/T formula was affected by the AL and K, while the predictability of Olsen and Haigis formulas was affected by the AL only. Conclusions: Steep cornea has more influence on the accuracy of IOL power calculation than the other corneal shape in long eyes. Overall, both the Olsen and Barrett Universal II formulas are recommended in long eyes with unusual keratometry.

Download Full-text

An ensemble-based approach for estimating personalized intraocular lens power

Scientific Reports ◽

10.1038/s41598-021-02288-x ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Salissou Moutari ◽

Jonathan E. Moore

Keyword(s):

Intraocular Lens ◽

Prediction Accuracy ◽

Ad Hoc ◽

Anterior Chamber Depth ◽

Decisive Role ◽

Power Calculation ◽

Prediction Errors ◽

Iol Power ◽

New Formula ◽

The Impact

AbstractThe fundamental difference between modern formulae for intraocular lens (IOL) power calculation lies on the single ad hoc regression model they use to estimate the effective lens position (ELP). The ELP is very difficult to predict and its estimation is considered critical for an accurate prediction of the required IOL power of the lens to be implanted during cataract surgery. Hence, more advanced prediction techniques, which improve the prediction accuracy of the ELP, could play a decisive role in improving patient refractive outcomes. This study introduced a new approach for the calculation of personalized IOL power, which used an ensemble of regression models to devise a more accurate and robust prediction of the ELP. The concept of cross-validation was used to rigorously assess the performance of the devised formula against the most commonly used and published formulae. The results from this study show that overall, the proposed approach outperforms the most commonly used modern formulae (namely, Haigis, Holladay I, Hoffer Q and SRK/T) in terms of mean absolute prediction errors and prediction accuracy i.e., the percentage of eyes within ± 0.5D and ± 1 D ranges of prediction, for various ranges of axial lengths of the eyes. The new formula proposed in this study exhibited some promising features in terms of robustness. This enables the new formula to cope with variations in the axial length, the pre-operative anterior chamber depth and the keratometry readings of the corneal power; hence mitigating the impact of their measurement accuracy. Furthermore, the new formula performed well for both monofocal and multifocal lenses.

Download Full-text

Accuracy of Intraocular Lens Power Calculation Formulas for Highly Myopic Eyes

Journal of Ophthalmology ◽

10.1155/2016/1917268 ◽

2016 ◽

Vol 2016 ◽

pp. 1-7 ◽

Cited By ~ 9

Author(s):

Yichi Zhang ◽

Xiao Ying Liang ◽

Shu Liu ◽

Jacky W. Y. Lee ◽

Srinivasan Bhaskar ◽

...

Keyword(s):

Intraocular Lens ◽

Absolute Error ◽

Power Calculation ◽

Prediction Errors ◽

Intraocular Lens Power Calculation ◽

Iol Power Calculation ◽

Predictive Error ◽

Intraocular Lens Power ◽

Iol Power ◽

Lens Power

Purpose.To evaluate and compare the accuracy of different intraocular lens (IOL) power calculation formulas for eyes with an axial length (AL) greater than 26.00 mm.Methods.This study reviewed 407 eyes of 219 patients with AL longer than 26.0 mm. The refractive prediction errors of IOL power calculation formulas (SRK/T, Haigis, Holladay, Hoffer Q, and Barrett Universal II) using User Group for Laser Interference Biometry (ULIB) constants were evaluated and compared.Results.One hundred seventy-one eyes were enrolled. The Barrett Universal II formula had the lowest mean absolute error (MAE) and SRK/T and Haigis had similar MAE, and the statistical highest MAE were seen with the Holladay and Hoffer Q formulas. The interquartile range of the Barrett Universal II formula was also the lowest among all the formulas. The Barrett Universal II formulas yielded the highest percentage of eyes within ±1.0 D and ±0.5 D of the target refraction in this study (97.24% and 79.56%, resp.).Conclusions.Barrett Universal II formula produced the lowest predictive error and the least variable predictive error compared with the SRK/T, Haigis, Holladay, and Hoffer Q formulas. For high myopic eyes, the Barrett Universal II formula may be a more suitable choice.

Download Full-text

Improving clinical refractive results of cataract surgery by machine learning

PeerJ ◽

10.7717/peerj.7202 ◽

2019 ◽

Vol 7 ◽

pp. e7202 ◽

Cited By ~ 6

Author(s):

Martin Sramka ◽

Martin Slovak ◽

Jana Tuckova ◽

Pavel Stodulka

Keyword(s):

Clinical Results ◽

Support Vector ◽

Power Calculation ◽

Prediction Errors ◽

Neural Network Ensemble ◽

Clinical Workflow ◽

Clinical Method ◽

Iol Power Calculation ◽

Significant Difference ◽

Iol Power

Aim To evaluate the potential of the Support Vector Machine Regression model (SVM-RM) and Multilayer Neural Network Ensemble model (MLNN-EM) to improve the intraocular lens (IOL) power calculation for clinical workflow. Background Current IOL power calculation methods are limited in their accuracy with the possibility of decreased accuracy especially in eyes with an unusual ocular dimension. In case of an improperly calculated power of the IOL in cataract or refractive lens replacement surgery there is a risk of re-operation or further refractive correction. This may create potential complications and discomfort for the patient. Methods A dataset containing information about 2,194 eyes was obtained using data mining process from the Electronic Health Record (EHR) system database of the Gemini Eye Clinic. The dataset was optimized and split into the selection set (used in the design for models and training), and the verification set (used in the evaluation). The set of mean prediction errors (PEs) and the distribution of predicted refractive errors were evaluated for both models and clinical results (CR). Results Both models performed significantly better for the majority of the evaluated parameters compared with the CR. There was no significant difference between both evaluated models. In the ±0.50 D PE category both SVM-RM and MLNN-EM were slightly better than the Barrett Universal II formula, which is often presented as the most accurate calculation formula. Conclusion In comparison to the current clinical method, both SVM-RM and MLNN-EM have achieved significantly better results in IOL calculations and therefore have a strong potential to improve clinical cataract refractive outcomes.

Download Full-text

NLOS Multipath Classification of GNSS Signal Correlation Output Using Machine Learning

Sensors ◽

10.3390/s21072503 ◽

2021 ◽

Vol 21 (7) ◽

pp. 2503

Author(s):

Taro Suzuki ◽

Yoshiharu Amano

Keyword(s):

Machine Learning ◽

Satellite System ◽

Training Data ◽

Support Vector ◽

Positioning Errors ◽

Automated Method ◽

Global Navigation Satellite ◽

Better Than ◽

Signal Correlation

This paper proposes a method for detecting non-line-of-sight (NLOS) multipath, which causes large positioning errors in a global navigation satellite system (GNSS). We use GNSS signal correlation output, which is the most primitive GNSS signal processing output, to detect NLOS multipath based on machine learning. The shape of the multi-correlator outputs is distorted due to the NLOS multipath. The features of the shape of the multi-correlator are used to discriminate the NLOS multipath. We implement two supervised learning methods, a support vector machine (SVM) and a neural network (NN), and compare their performance. In addition, we also propose an automated method of collecting training data for LOS and NLOS signals of machine learning. The evaluation of the proposed NLOS detection method in an urban environment confirmed that NN was better than SVM, and 97.7% of NLOS signals were correctly discriminated.

Download Full-text

Prediction accuracy of conventional and total keratometry for intraocular lens power calculation in femtosecond laser-assisted cataract surgery

Scientific Reports ◽

10.1038/s41598-021-92354-1 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Soyoung Ryu ◽

Ikhyun Jun ◽

Tae-im Kim ◽

Kyoung Yul Seo ◽

Eung Kweon Kim

Keyword(s):

Femtosecond Laser ◽

Cataract Surgery ◽

Intraocular Lens ◽

Intraclass Correlation ◽

Absolute Error ◽

Power Calculation ◽

Prediction Errors ◽

Intraocular Lens Power Calculation ◽

Iol Power Calculation ◽

Severance Hospital

AbstractThis study evaluated the accuracy of total keratometry (TK) and standard keratometry (K) for intraocular lens (IOL) power calculation in eyes treated with femtosecond laser-assisted cataract surgery. The retrospective study included a retrospective analysis of data from 62 patients (91 eyes) who underwent uneventful femtosecond laser-assisted cataract surgery with Artis PL E (Cristalens Industrie, Lannion, France) IOL implantation by a single surgeon between May 2020 and December 2020 in Severance Hospital, Seoul, South Korea. The new IOLMaster 700 biometry device (Carl Zeiss Meditec, Jena, Germany) was used to calculate TK and K. The mean absolute error (MAE), median absolute error (MedAE), and the percentages of eyes within prediction errors of ± 0.25 D, ± 0.50 D, and ± 1.00 D were calculated for all IOL formulas (SRK/T, Hoffer-Q, Haigis, Holladay 1, Holladay 2, and Barrett Universal II). There was strong agreement between K and TK (intraclass correlation coefficient = 0.99), with a mean difference of 0.04 D. For all formulas, MAE tended to be lower for TK than for K, and relatively lower MAE and MedAE values were observed for SRK/T and Holladay 1. Furthermore, for all formulas, a greater proportion of eyes fell within ± 0.25 D of the predicted postoperative spherical equivalent range in the TK group than in the K group. However, differences in MAEs, MedAEs, and percentages of eyes within the above prediction errors were not statistically significant. In conclusion, TK and K exhibit comparable performance for refractive prediction in eyes undergoing femtosecond laser-assisted cataract surgery.

Download Full-text

Civil Aeroengine Fault Diagnosis Based on Fuzzy Least Square Support Vector Machine

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.130-134.2047 ◽

2011 ◽

Vol 130-134 ◽

pp. 2047-2050 ◽

Cited By ~ 1

Author(s):

Hong Chun Qu ◽

Xie Bin Ding

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Support Vector Machine ◽

Fault Diagnosis ◽

Coefficient Matrix ◽

Least Square ◽

Support Vector ◽

Influence Coefficient ◽

Structural Risk ◽

Better Than

SVM(Support Vector Machine) is a new artificial intelligence methodolgy, basing on structural risk mininization principle, which has better generalization than the traditional machine learning and SVM shows powerfulability in learning with limited samples. To solve the problem of lack of engine fault samples, FLS-SVM theory, an improved SVM, which is a method is applied. 10 common engine faults are trained and recognized in the paper.The simulated datas are generated from PW4000-94 engine influence coefficient matrix at cruise, and the results show that the diagnostic accuracy of FLS-SVM is better than LS-SVM.

Download Full-text

Improving Machine Learning Identification of Unsafe Driver Behavior by Means of Sensor Fusion

Applied Sciences ◽

10.3390/app10186417 ◽

2020 ◽

Vol 10 (18) ◽

pp. 6417 ◽

Cited By ~ 1

Author(s):

Emanuele Lattanzi ◽

Giacomo Castellucci ◽

Valerio Freschi

Keyword(s):

Neural Network ◽

Machine Learning ◽

Driver Behavior ◽

Ground Truth ◽

Support Vector ◽

Svm Classifier ◽

Learning Technology ◽

Average Accuracy ◽

Unsafe Behaviors ◽

Vehicle Sensors

Most road accidents occur due to human fatigue, inattention, or drowsiness. Recently, machine learning technology has been successfully applied to identifying driving styles and recognizing unsafe behaviors starting from in-vehicle sensors signals such as vehicle and engine speed, throttle position, and engine load. In this work, we investigated the fusion of different external sensors, such as a gyroscope and a magnetometer, with in-vehicle sensors, to increase machine learning identification of unsafe driver behavior. Starting from those signals, we computed a set of features capable to accurately describe the behavior of the driver. A support vector machine and an artificial neural network were then trained and tested using several features calculated over more than 200 km of travel. The ground truth used to evaluate classification performances was obtained by means of an objective methodology based on the relationship between speed, and lateral and longitudinal acceleration of the vehicle. The classification results showed an average accuracy of about 88% using the SVM classifier and of about 90% using the neural network demonstrating the potential capability of the proposed methodology to identify unsafe driver behaviors.

Download Full-text

Post Cataract Surgery Refractive Error in Myopic Patients Using SRK/T Versus Holladay 1 Formula for IOL Power Calculation

Pakistan Journal of Ophthalmology ◽

10.36351/pjo.v34i2.947 ◽

2019 ◽

Vol 34 (2) ◽

Author(s):

Sidra Anwar, Atif Mansoor Ahmad, Irum Abbas, Zyeima Arif

Keyword(s):

Intraocular Lens ◽

Refractive Error ◽

Axial Length ◽

Power Calculation ◽

Mean Error ◽

Iol Power Calculation ◽

Group A ◽

Iol Power ◽

The Mean ◽

Group B

Purpose: To compare post-operative mean refractive error with SandersRetzlaff-Kraff/theoretical (SRK-T) and Holladay 1 formulae for intraocular lens (IOL) power calculation in cataract patients with longer axial lengths. Study Design: Randomized controlled trial. Place and Duration of Study: Department of Ophthalmology, Shaikh Zayed Hospital Lahore from 01 January 2017 01 January, 2018. Material and Methods: A total of 80 patients were selected from Ophthalmology Outdoor of Shaikh Zayed Hospital Lahore. The patients were randomly divided into two groups of 40 each by lottery method. IOL power calculation was done in group A using SRK-T formula and in group B using Holladay1 formula after keratomery and A-scan. All patients underwent phacoemulsification with foldable lens implantation. Post-operative refractive error was measured after one month and mean error was calculated and compared between the two groups. Results: Eighty cases were included in the study with a mean age of 55.8 ± 6.2 years. The mean axial length was 25.63 ± 0.78mm, and the mean keratometric power was 43.68 ± 1.1 D. The mean post-operative refractive error in group A (SRK/T) was +0.36D ± 0.33D and in group B (Holladay 1) it was +0.68 ± 0.43. The Mean Error in group A was +0.37D ± 0.31D as compared to +0.69D ± 0.44D in group B. Conclusion: SRK/T formula is superior to Holladay 1 formula for cases having longer axial lengths. Key words: Phacoemulsification, intraocular lens power, longer axial length, biometry.

Download Full-text