Mean Prediction Error Research Articles

Visual Outcomes After Cataract Surgery With the Light Adjustable Lens in Japanese Patients With and Without Prior Corneal Refractive Surgery.

To describe Light Adjustable Lens (RxSight, Inc) (LAL) visual and refractive outcomes in a Japanese patient population with and without prior refractive surgery. In this retrospective chart review of patients with cataract in a private eye clinic, 34 eyes from 21 patients underwent cataract surgery with LAL implantation. Eyes underwent light treatment to adjust the patients' refraction after implantation to meet individual visual goals. Standard visual and refractive outcomes were collected. Contrast sensitivity (to assess quality of vision), clinical defocus and vergence curves (to assess range of vision), and patient satisfaction were also collected. The mean prediction error of the light adjustment was -0.04 ± -0.48. Thirty-one eyes (91%) and 33 eyes (97%) were within ±0.25 and ±0.50 diopters (D) of manifest spherical equivalent target, respectively. Thirty-two eyes (94%) had residual refractive cylinder of 0.50 D or less. The full depth of focus was 3.68 D. Monocular contrast sensitivity shows frequencies equivalent to a normal population aged 50 to 60 years. The mean satisfaction score was 8.8 (scale: 1 to 10). There were no adverse events and no eyes lost corrected distance visual acuity from the light treatment. LALs in cataract refractive surgery hold great promise, including for the Japanese population, which have unique ocular biometrics. By providing a safe procedure and the ability for postoperative adjustments, the LAL provided a tailored approach to vision correction, ultimately improving patient outcomes and satisfaction. [J Refract Surg. 2024;40(11):e854-e862.].

Comparison of Precision, Agreement, and Accuracy of Two Swept-Source Optical Coherence Tomography Biometers.

Background/Objectives: This study's aim was to compare the precision, agreement, and accuracy in axial length (AL) measurements of Argos® (Alcon Healthcare, US) and Eyestar 900® (Haag-Streit, Switzerland) swept-source optical coherence tomography (SS-OCT) biometers. Methods: We performed a prospective evaluation of two diagnostic devices. Three consecutive measurements of AL with the Argos® and the Eyestar® 900 SS-OCT biometers were conducted in random order in eyes undergoing cataract surgery in Barcelona, Spain. The main endpoint was the median difference in AL between devices. Secondary endpoints included agreement on Bland-Altman plots and 95% limits of agreement (LoAs), repeatability as measured within-subject standard deviation (SW), percent of failed AL measurements, percent of eyes within ±0.50 D and ±1.00 D one month after surgery, and median and mean prediction error. Results: We included 107 eyes of 107 patients (60.8% females, mean age of 73.1 years). The median difference in AL (Argos®-Eyestar 900®) was -0.01 mm (interquartile range [IQR], 0.06), p = 0.01. The 95% LoAs were -0.11 to +0.08 mm, with a trend towards less extreme measurements with Argos® for very short and long eyes. The median (IQR) Sw was 0.0058 (0.0058) and 0.0000 (0.0058) for Argos® and Eyestar 900®, respectively. There were no failed AL measurements with either device (0%, 95% CI = 0% to 3.4%). Overall, 96.1% of eyes were within ±0.50 D and 100% were within ±1.00 D. Conclusions: Argos® and Eyestar 900® provided statistically different but clinically negligible differences in AL. However, they are not interchangeable in very long or short eyes, due to the different principles used to determine AL.

Comparison of intraocular lens power formulas for negative-diopter intraocular lens implantation for high myopia.

To compare the accuracy of intraocular lens (IOL) power calculation formulas for myopic eyes requiring negative diopter powered IOLs. Retrospective case series. K… hospital and Y… Hospital, …, …. Sixty-one eyes that underwent phacoemulsification with implantation of a negative power IOL were investigated. The trueness, precision and accuracy of IOL power calculation were assessed for the Barrett Universal II (BUII), EVO 2.0, Haigis, Hoffer QST, Holladay 1 and SRK/T formulas using the Eyetemis online tool. The analysis was performed using 1) the ULIB IOL constants and 2) after constant optimization. With ULIB constants, the Haigis, Holladay 1 and SRK/T resulted in a hyperopic mean prediction error (PE) >1.00 diopter (D), which was significantly different from zero (adjusted p <0.05). The mean PE of the remaining formulas was closer to zero. The absolute PE was significantly higher with the Holladay 1 and SRK/T (adjusted p <0.05) with respect to the remaining formulas. After constant optimization, the outcomes of traditional formulas improved and no statistically significant differences were found among any of the formulas in terms of trueness, precision and accuracy. The percentage of eyes with an absolute PE within 0.50 D was low (<50%) even after constant optimization. With ULIB constants, the BUII, EVO 2.0 and Hoffer QST were more accurate than traditional formulas in eyes with negative-diopter IOLs. The results of IOL power calculation in these eyes remain poor even after constant optimization.

Detection of piperine content in black pepper using a molecular imprinted poly(N,N-dimethylacrylamide) embedded graphite electrode: A machine learning based prediction approach

The paper presents a molecular imprinted graphite electrode for selective and sensitive detection of piperine (1-Propylpiperidine) in Piper nigrum (black pepper). Poly(N,N-dimethylacrylamide) (PDMAM) was utilized as a functional monomer directing polymerization of uniform molecularly imprinted polymer (MIP) layers on the surface of the graphite layer using ethylene glycol dimethacrylate (EGDMA) as cross-linker with piperine (PIP) as template molecules. The formation of imprinting, non-imprinting, and elution of analytes within the polymer matrix were studied by field emission scanning electron microscope (FESEM), Fourier transform infrared (FTIR), UV–visible (UV–Vis) and X-ray photoelectron spectroscopy (XPS). The electrochemical response of piperine was evaluated through differential pulse voltammetry (DPV) using a PDMAM-MIP@G electrode that showed an oxidation peak at 1.25 ± 0.15 V. Under optimized experimental conditions, DPV peak currents were linear over two concentration ranges (0.001–1 µM and 1–200 µM). A limit of detection (LOD) was found to be 0.0006 µM. Developed electrode yielded high selectivity, sensitivity (12.92 µA/µM (lower linear range) and 0.33 µA/µM (upper linear range)), promising reproducibility, long-term stability (3.85 % decrease in response over 28 days), and interference immunity. The developed PDMAM-MIP@G electrode was applied to real samples using four branded pepper powder procured from local market and the sensor data was recorded. A convolutional neural network (CNN) driven prediction model has been proposed to predict the piperine concentration. Among the four CNN models tested, the best model, averaged over 50 runs, achieved a mean absolute error (MAE) of 0.268, a mean prediction error of 2.35 %, and a R-squared value of 0.93 on the testing set when compared with reference values obtained from reverse-phase high-performance liquid chromatography (RP-HPLC).

Impact of the Minimization of Standard Deviation Before Zeroization of the Mean Bias on the Performance of IOL Power Formulas.

In cataract surgery, accurate intraocular lens (IOL) power calculations are crucial for optimal postoperative refractive outcomes. This study explores the impact of prioritizing the reduction of the standard deviation (SD) of prediction errors before mean prediction error (PE) adjustment on IOL calculation formula precision and accuracy. We conducted a retrospective analysis of 4885 eyes from 2611 patients, all implanted with the same IOL model, comparing four traditional IOL power calculation formulas: SRK/T, Holladay 1, Haigis, and Hoffer Q. We introduced new constants aiming to minimize the SD of PE (new_const) against traditionally optimized constants (classic_const), using a heteroscedastic statistical method for comparison. Validation of precision improvements used a secondary dataset of 262 eyes from 132 patients. We observed significant reductions in mean absolute error (MAE) across training and test sets for Hoffer Q, Holladay, and Haigis formulas, indicating accuracy enhancements. Optimized constants significantly reduced SDs for Haigis from 0.3255 to 0.3153 and for Hoffer Q from 0.3521 to 0.3387. These optimizations also increased the proportion of eyes achieving PE within ±0.25D. SRK/T showed improved SD from 0.3596 to 0.3585. However, Holladay 1 showed minimal change with no significant improvement. In the test dataset, significant reductions in SD were observed for Haigis and Hoffer Q. Prioritizing SD minimization before adjusting mean PE significantly improves the precision of selected IOL power formulas, enhancing postoperative refractive outcomes. The effectiveness varies among formulas, underscoring the need for formula-specific adjustments. The study presents a novel two-step approach for optimizing IOL power calculations.

Application of machine-learning models to predict the ganciclovir and valganciclovir exposure in children using a limited sampling strategy.

Intravenous ganciclovir and oral valganciclovir display significant variability in ganciclovir pharmacokinetics, particularly in children. Therapeutic drug monitoring currently relies on the area under the concentration-time (AUC). Machine-learning (ML) algorithms represent an interesting alternative to Maximum-a-Posteriori Bayesian-estimators for AUC estimation. The goal of our study was to develop and validate an ML-based limited sampling strategy (LSS) approach to determine ganciclovir AUC0-24 after administration of either intravenous ganciclovir or oral valganciclovir in children. Pharmacokinetic parameters from four published population pharmacokinetic models, in addition to the World Health Organization growth curve for children, were used in the mrgsolve R package to simulate 10,800 pharmacokinetic profiles of children. Different ML algorithms were trained to predict AUC0-24 based on different combinations of two or three samples. Performances were evaluated in a simulated test set and in an external data set of real patients. The best estimation performances in the test set were obtained with the Xgboost algorithm using a 2 and 6 hours post dose LSS for oral valganciclovir (relative mean prediction error [rMPE] = 0.4% and relative root mean square error [rRMSE] = 5.7%) and 0 and 2 hours post dose LSS for intravenous ganciclovir (rMPE = 0.9% and rRMSE = 12.4%). In the external data set, the performance based on these two sample LSS was acceptable: rMPE = 0.2% and rRMSE = 16.5% for valganciclovir and rMPE = -9.7% and rRMSE = 17.2% for intravenous ganciclovir. The Xgboost algorithm developed resulted in a clinically relevant individual estimation using only two blood samples. This will improve the implementation of AUC-targeted ganciclovir therapeutic drug monitoring in children.

A comparison of the renal function biomarkers serum creatinine, pro-enkephalin and cystatin C to predict clearance of pemetrexed.

For drugs with a narrow therapeutic window, there is a delicate balance between efficacy and toxicity, thus it is pivotal to administer the right dose from the first administration onwards. Exposure of pemetrexed, a cytotoxic drug used in lung cancer treatment, is dictated by kidney function. To facilitate optimized dosing of pemetrexed, accurate prediction of drug clearance is pivotal. Therefore, the aim of this study was to investigate the performance of the kidney function biomarkers serum creatinine, cystatin C and pro-enkephalin in terms of predicting the elimination of pemetrexed. We performed a population pharmacokinetic analysis using a dataset from two clinical trials containing pharmacokinetic data of pemetrexed and measurements of all three biomarkers. A three-compartment model without covariates was fitted to the data and the obtained individual empirical Bayes estimates for pemetrexed clearance were considered the "true" values (Cltrue). Subsequently, the following algorithms were tested as covariates for pemetrexed clearance: the Chronic Kidney Disease Epidemiology Collaboration equation using creatinine (CKD-EPICR), cystatin C (CKD-EPICYS), a combination of both (CKD-EPICR-CYS), pro-enkephalin as an absolute valueor in a combined algorithm with age and serum creatinine, and lastly, a combination of pro-enkephalin with cystatin C. The dataset consisted of 66 subjects with paired observations for all three kidney function biomarkers. Inclusion of CKD-EPICR-CYS as a covariate on pemetrexed clearance resulted in the best model fit, with the largest decrease in objective function (p < 0.00001) and explaining 35% of the total inter-individual variability in clearance. The predictive performance of the model to containing CKD-EPICR-CYS to predict pemetrexed clearance was good with a normalized root mean squared error and mean prediction error of 19.9% and 1.2%, respectively. In conclusion, this study showed that the combined CKD-EPICR-CYS performs best in terms predicting pharmacokinetics of pemetrexed. Despite the hypothesized disadvantages, creatinine remains to be a suitable and readily available marker to predict pemetrexed clearance in clinical practice.

Non-invasive regional parameter identification of degenerated human meniscus

Accurate identification of local changes in the biomechanical properties of the normal and degenerative meniscus is critical to better understand knee joint osteoarthritis onset and progression. Ex-vivo material characterization is typically performed on specimens obtained from different locations, compromising the tissue's structural integrity and thus altering its mechanical behavior. Therefore, the aim of this in-silico study was to establish a non-invasive method to determine the region-specific material properties of the degenerated human meniscus. In a previous experimental magnetic resonance imaging (MRI) study, the spatial displacement of the meniscus and its root attachments in mildly degenerated (n = 12) and severely degenerated (n = 12) cadaveric knee joints was determined under controlled subject-specific axial joint loading. To simulate the experimental response of the lateral and medial menisci, individual finite element models were created utilizing a transverse isotropic hyper-poroelastic constitutive material formulation. The superficial displacements were applied to the individual models to calculate the femoral reaction force in an inverse finite element analysis. During particle swarm optimization, the four most sensitive material parameters were varied to minimize the error between the femoral reaction force and the force applied in the MRI loading experiment. Individual global and regional parameter sets were identified. In addition to in-depth model verification, prediction errors were determined to quantify the reliability of the identified parameter sets. Both compressibility of the solid meniscus matrix (+141 %, p ≤ 0.04) and hydraulic permeability (+53 %, p ≤ 0.04) were significantly increased in the menisci of severely degenerated knees compared to mildly degenerated knees, irrespective of the meniscus region. By contrast, tensile and shear properties were unaffected by progressive knee joint degeneration. Overall, the optimization procedure resulted in reliable and robust parameter sets, as evidenced by mean prediction errors of <1 %. In conclusion, the proposed approach demonstrated high potential for application in clinical practice, where it might provide a non-invasive diagnostic tool for the early detection of osteoarthritic changes within the knee joint.

Updating the no-history method in intraocular lens power calculation after myopic laser vision correction.

To describe the Shammas-Cooke formula, an updated no-history (NH) formula for IOL calculation in eyes with prior myopic laser vision correction (M-LVC), and to compare the results with the Shammas PL, Haigis-L, and Barrett True-K NH formulas. Bascom Palmer Eye Institute (BPEI), The Lennar Foundation Medical Center, University of Miami, Miami, Florida; Dean A. McGee Eye Institute (DMEI), University of Oklahoma, Oklahoma City, Oklahoma; and private practice, Lynwood, California, and St Joseph, Michigan. Retrospective observational study. We analyzed 2 large series of cataractous eyes with prior M-LVC. The training set (BPEI series of 330 eyes) was used to derive the new corneal power conversion equation to be used in the new Shammas-Cooke formula and the testing set (165 eyes of 165 patients in the DMEI series) to compare the updated formula with 3 other M-LVC NH formulas on the ASCRS calculator: Shammas PL, Haigis-L, and Barrett True-K NH. Mean prediction error was 0.09 ± 0.56 diopters (D), -0.44 ± 0.61 D, -0.47 ± 0.59 D, and -0.18 ± 0.56 D and the mean absolute error was 0.43 D, 0.60 D, 0.61 D, and 0.45 D for the Shammas-Cooke, Shammas PL, Haigis-L, and Barrett True-K NH, respectively. The percentage of eyes within ±0.50 D was 66.7% vs 47.9%, 48.5%, and 65.5%, respectively. The Shammas-Cooke formula performed better than the Shammas PL and Haigis-L (P < .001 for both) and as well as the Barrett True-K NH formula (P = .923).

Comparison of the Accuracy Between the Z CALC2 Calculator and Barrett Toric Calculator in Toric IOL Calculation.

To compare the accuracy of the Z CALC2 calculator and Barrett toric calculator in toric intraocular lens (IOL) calculation. Eighty-five eyes of 85 patients who underwent uneventful cataract surgery with toric IOL implantation were included. The accuracy was compared between the Z CALC2 calculator and Barrett toric calculator under two calculation modes: using simulated keratometry (SimK) from the IOLMaster 700 (Carl Zeiss Meditec AG) for posterior corneal astigmatism (PCA) prediction and employing total corneal astigmatism (total corneal refractive power [TCRP] or measured PCA) obtained from the Pentacam (Oculus Optikgeräte GmbH). The centroid of prediction errors, percentage of eyes with prediction errors within ±0.50 diopter (D), mean prediction error, and mean absolute prediction error were calculated. Subgroup analysis was conducted based on the orientation and magnitude of anterior corneal astigmatism and axial length. When using SimK, the two calculators with predicted PCA showed comparable accuracy. When employing total corneal astigmatism, the Barrett toric calculator with measured PCA showed a lower centroid error (0.15 vs 0.38 D), a higher percentage of eyes with prediction errors within ±0.50 D (47.1% vs 32.9%, P = .018), and a lower mean prediction error (0.57 vs 0.71 D, P = .033) compared to the Z CALC2 calculator with TCRP in the 4-mm zone. In subgroup analysis, when employing total corneal astigmatism, the Barrett toric calculator with measured PCA exhibited superior accuracy, especially in the with-the-rule and anterior corneal astigmatism of 2.00 D or less subgroups. When using SimK, the Z CALC2 calculator and Barrett toric calculator yield comparable accuracy. The Barrett toric calculator with measured PCA may be more recommended when employing total corneal astigmatism. [J Refract Surg. 2024;40(10):e681-e691.].

Bayesian impedance deconvolution using timescale distribution for lithium-ion battery state estimation

The distribution of relaxation times (DRT) and distribution of diffusion times (DDT) are widely recognized as effective model-free methods for deconvolving the internal properties of complex electrochemical systems using electrochemical impedance spectroscopy (EIS) data. This study proposes an integrated framework that employs a Bayesian approach to accurately estimate both DRT and DDT and machine learning techniques to enhance capacity estimation, incorporating Gaussian process regression and transformer networks. These methods utilize the inferred DRT and DDT as inputs to predict the discharge capacity. In addition, we perform a peak analysis on the estimated DRT and DDT to extract additional physically meaningful features, which have also proven to be effective inputs for capacity prediction. The applicability of the proposed framework to the EIS experimental data of lithium-ion cells is demonstrated and compared with existing capacity estimation methods. The proposed framework demonstrates a mean absolute error in capacity prediction below 3.03% when using DRT and DDT directly and 2.71% when using extracted features, outperforming existing methods by approximately one to three percentage points. Our Bayesian approach, combined with peak analysis and the integration of machine learning techniques, enables a more robust diagnosis of the internal state of lithium-ion batteries through EIS measurements.

Parallel and scalable AI in HPC systems for CFD applications and beyond

This manuscript presents the library AI4HPC with its architecture and components. The library enables large-scale trainings of AI models on High-Performance Computing systems. It addresses challenges in handling non-uniform datasets through data manipulation routines, model complexity through specialized ML architectures, scalability through extensive code optimizations that augment performance, HyperParameter Optimization (HPO), and performance monitoring. The scalability of the library is demonstrated by strong scaling experiments on up to 3,664 Graphical Processing Units (GPUs) resulting in a scaling efficiency of 96%, using the performance on 1 node as baseline. Furthermore, code optimizations and communication/computation bottlenecks are discussed for training a neural network on an actuated Turbulent Boundary Layer (TBL) simulation dataset (8.3 TB) on the HPC system JURECA at the Jülich Supercomputing Centre. The distributed training approach significantly influences the accuracy, which can be drastically compromised by varying mini-batch sizes. Therefore, AI4HPC implements learning rate scaling and adaptive summation algorithms, which are tested and evaluated in this work. For the TBL use case, results scaled up to 64 workers are shown. A further increase in the number of workers causes an additional overhead due to too small dataset samples per worker. Finally, the library is applied for the reconstruction of TBL flows with a convolutional autoencoder-based architecture and a diffusion model. In case of the autoencoder, a modal decomposition shows that the network provides accurate reconstructions of the underlying field and achieves a mean drag prediction error of ≈5%. With the diffusion model, a reconstruction error of ≈4% is achieved when super-resolution is applied to 5-fold coarsened velocity fields. The AI4HPC library is agnostic to the underlying network and can be adapted across various scientific and technical disciplines.

Quantification Approaches in Non-Target LC/ESI/HRMS Analysis: An Interlaboratory Comparison.

Nontargeted screening (NTS) utilizing liquid chromatography electrospray ionization high-resolution mass spectrometry (LC/ESI/HRMS) is increasingly used to identify environmental contaminants. Major differences in the ionization efficiency of compounds in ESI/HRMS result in widely varying responses and complicate quantitative analysis. Despite an increasing number of methods for quantification without authentic standards in NTS, the approaches are evaluated on limited and diverse data sets with varying chemical coverage collected on different instruments, complicating an unbiased comparison. In this interlaboratory comparison, organized by the NORMAN Network, we evaluated the accuracy and performance variability of five quantification approaches across 41 NTS methods from 37 laboratories. Three approaches are based on surrogate standard quantification (parent-transformation product, structurally similar or close eluting) and two on predicted ionization efficiencies (RandFor-IE and MLR-IE). Shortly, HPLC grade water, tap water, and surface water spiked with 45 compounds at 2 concentration levels were analyzed together with 41 calibrants at 6 known concentrations by the laboratories using in-house NTS workflows. The accuracy of the approaches was evaluated by comparing the estimated and spiked concentrations across quantification approaches, instrumentation, and laboratories. The RandFor-IE approach performed best with a reported mean prediction error of 15× and over 83% of compounds quantified within 10× error. Despite different instrumentation and workflows, the performance was stable across laboratories and did not depend on the complexity of water matrices.

Machine learning approach for predicting the machining characteristics of additive manufactured ABS material

An Additive manufacturing process offers advantage for the production of intricate designs with increased efficiency and reduced material waste. In this research, an experimental and analysis work has been conducted to identify the significant 3-D printed input parameters effecting the output response characteristics as kerf width. The correlation between experimental and predicted values has been established through the application of Artificial Neural Network (ANN), Support Vector Machine Regression (SVMR), and Response Surface Methodology (RSM) techniques. The validation of predicted models conducted with the help of mean prediction error, maximum error and minimum error. The performance of ANN model was analyzed by feedforward-propagation method using Levenberg- Marquardt algorithm. The maximum prediction error investigated by ANN, SVM, and RSM were noted as 4.28, 5.18, and 5.84 respectively. The optical micrographs have been investigated to understand the hole and crater formation phenomenon.

Optimizing High-Performance Predictive Modeling of the Medium-Speed WEDM Processing of Inconel 718

The purpose of this research was to create a predictive model for a medium-speed wire electrical discharge machine (WEDM) utilizing an artificial neural network (ANN). Medium-speed WEDM experiments were developed based on the I-optimal mixture design for machining, the Inconel 718 superalloy. During the experiment, the input parameters were the spark ontime, spark offtime, wire feed, and current, with the material removal rate (MRR) and surface roughness (Ra) selected as performance indicators. The ANN model was trained on experimental data and built using a feed-forward backpropagation neural network with a (4-8-2) structure and the Bayesian regularization (BR) learning approach. The model correctly predicted the relationship between the medium-speed WEDM’s primary process parameters and machining performance. An integrated ANN model and the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) were used to determine the ideal parameters for the MRR and Ra, resulting in a set of Pareto-optimal solutions. The confirmation experiment revealed that the mean prediction error between the experimental and ideal solutions had a maximum error percentage of 1% for the MRR and 2% for the Ra, which are within acceptable ranges. This showed that the best process–parameter combinations were better for the MRR and Ra.

Effect of Posterior Keratometry and Corneal Radius Ratio on the Accuracy of Intraocular Lens Formulas After Myopic LASIK/PRK.

To investigate the impact of back-to-front corneal radius ratio (B/F ratio) and posterior keratometry (PK) on the accuracy of intraocular lens power calculation formulas in eyes after myopic laser in situ keratomileusis (LASIK)/photorefractive keratectomy (PRK) surgery. A retrospective, consecutive case series study included 101 patients (132 eyes) with cataract after myopic LASIK/PRK. Mean prediction error (PE), mean absolute PE (MAE), median absolute error (MedAE), and the percentage of eyes within ±0.25, ±0.50, and ±1.00 diopters (D) of PE were determined. The Barrett True K-TK formula exhibited the lowest MAE (0.59 D) and MedAE (0.48 D) and the highest percentage of eyes within ±0.50 D of PE (54.55%) in total. In eyes with a B/F ratio of 0.70 or less and PK of -5.70 D or greater, the Potvin-Hill formula displayed the lowest MAE (0.46 to 0.67 D). The Barrett True-TK exhibited the highest prediction accuracy in eyes after myopic LASIK/PRK overall. However, for eyes with a low B/F ratio and flat PK, the Potvin-Hill performed best. [J Refract Surg. 2024;40(9):e635-e644.].

A Novel Method to Optimize Personal IOL Constants

ObjectiveTo describe a novel method called ‘three variable optimization’ that entails a process of doing just one calculation to zero out the mean prediction error of an entire dataset (regardless of size), using only 3 variables: 1) the constant used, 2) the average intraocular lens (IOL) power and 3) the average PE. DesignDevelopment, evaluation, and testing of a method to optimize personal IOL constants. MethodsA dataset of 876 eyes was used as a training set, and another dataset of 1,079 eyes was used to test the method. The Barrett Universal II, Cooke K6, Haigis, RBF 3.0, Hoffer Q, Holladay 1, Holladay 2, SRK/T and T2 were analyzed. The same dataset was also divided into 3 subgroups (short, medium and long eyes). The three variable optimization process was applied to each dataset and subset, and the obtained optimized constants were then used to obtain the mean PE of each dataset. We then compared those results with those obtained by zeroing out the mean PE in the classical method. ResultsThe three variable optimization showed similar results to classical optimization with less data needed to optimize and no clinically significant difference. Dividing the dataset into subsets of short, medium and long eyes, also shows that the method is useful even in those situations. Finally, the method was tested in multiple formulas and it was able to reduce the PE with no clinical significant difference from classical optimization. ConclusionThis method could then be applied by surgeons to optimize their constants by reducing the mean prediction error to zero without prior technical knowledge and it is available online for free at http://wwww.ioloptimization.com.

Prediction of body condition score throughout lactation by random regression test-day models.

Regular monitoring of body condition score (BCS) changes during lactation is a crucial management tool in dairy cattle; however, the current BCS measurements are often discontinuous and unevenly spaced in time. The aim of this study was to investigate the ability of random regression test-day models (RR-TDM) to predict BCS for the entire lactation in dairy cows even if the actual scoring is limited to one BCS record. The data consisted of test-day records of milk yield (MY), fat percentage (FP), protein percentage (PP) and BCS (based on a 9-point scale with unit increments; 1-9) collected from 2014 to 2022 in 128 herds in the Walloon Region of Belgium. In total, 20,698 test-day records on 2166 first-parity Holstein cows (2-12 with an average of 9.42 test-day records per cow) were available for MY, FP and PP; and 7985 records on the same animals (2-12 with an average of 3.68 records per cow) were available for BCS. To estimate the solutions, only one randomly selected BCS record per animal along with all her MY, FP and PP records were used, which were then used to predict BCS data (calibration set). The remaining BCS (1-11 with an average 2.86 BCS records per animal) were used to evaluate the goodness of the predictions (validation set). Multiple-trait RR-TDM was used to estimate (co)variance components through the average information restricted maximum likelihood (AI-REML) algorithm. Predicted BCS were grouped into nine classes as the original observed BCS used for comparison. Pearson correlation between the predicted and observed BCS, prediction error (PE), absolute prediction error (APE) and root mean squared prediction error (RMSE) were calculated. Mean (standard deviation; SD) BCS was 4.97 (1.01), 4.95 (1.07) and 4.98 (1.00) BCS units in the full, calibration and validation datasets, respectively. Pearson correlation between the observed and predicted BCS was 0.71, mean (SD) PE was 0.04 (0.52) BCS units, mean (SD) APE was 0.48 (0.53) BCS units and RMSE was 0.72 BCS units. These findings demonstrate the ability of RR-TDM to predict BCS for the entire lactation using a single BCS record along with available test-day records of milk yield and composition in Holstein dairy cows.

Integrating machine learning for enhanced wildfire severity prediction: A study in the Upper Colorado River basin

Wildfires pose significant threats worldwide, requiring accurate prediction for mitigation. This study uses machine learning techniques to forecast wildfire severity in the Upper Colorado River basin. Datasets from 1984 to 2019 and key indicators like weather conditions and land use were employed. Random Forest outperformed Artificial Neural Network, achieving 72 % accuracy. Influential predictors include air temperature, vapor pressure deficit, NDVI, and fuel moisture. Solar radiation, SPEI, precipitation, and evapotranspiration also contribute significantly. Validation against actual severities from 2016 to 2019 showed mean prediction errors of 11.2 %, affirming the model's reliability. These results highlight the efficacy of machine learning in understanding wildfire severity, especially in vulnerable regions.

Accuracy Validation of the New Barrett True Axial Length Formula and the Optimized Lens Factor Using Sum-of-Segment Biometry.

Objectives: This study aims to verify the accuracy of a new calculation formula, Barrett true axial length formula (T-AL), and the optimized lens factor (LF) for predicting postoperative refraction after cataract surgery. Methods: We included 156 Japanese patients who underwent cataract surgery using Clareon monofocal intraocular lenses at our clinic between January 2022 and June 2023. Postoperative spherical equivalent was calculated using subjective refraction values obtained 1 month post-surgery. The LFs were optimized so that the mean prediction error (PE) of each calculation formula was zero (zero optimization). We calculated the mean absolute PE (MAE) to assess accuracy and used a Friedman test for statistical comparisons. The accuracy of T-AL and the optimized LFs was compared with that of the conventional Barrett Universal II formula for ARGOS (AR-B) and OA-2000 (OA-B) with equivalent refractive index. Results: For T-AL, AR-B, and OA-B, the MAEs ± standard deviations were 0.225 ± 0.179, 0.219 ± 0.168, and 0.242 ± 0.206 D, respectively. The Friedman test showed no statistically significant differences among the three groups. The device-optimized LFs were 2.248-2.289 (T-AL), 2.236-2.246 (AR-B), and 2.07-2.08 (OA-B); the corresponding zero-optimized LFs were 2.262-2.287 (T-AL), 2.287-2.303 (AR-B), and 2.160-2.170 (OA-B). Conclusion: There were no significant differences in prediction accuracy among the formulas. However, the accuracy of LF optimization varied by device, with T-AL being closest to the value under zero optimization. This suggests that T-AL is clinically useful for predicting an accurate postoperative refraction without zero optimization.