Skip to main content

Table 4 Prediction performance in different levels of myopia in TICL cases (n = 205)

From: Artificial intelligence-based refractive error prediction and EVO-implantable collamer lens power calculation for myopia correction

Parameters 

Preop SE >  − 6.00 D

(n = 18)

Preop SE − 6.00 to − 10.00 D

(n = 93)

Preop SE <  − 10.00 D

(n = 94)

 

MPE (D)

SD (D)

MAE (D)

MedAE (D)

MPE (D)

SD (D)

MAE (D)

MedAE (D)

MPE (D)

SD (D)

MAE (D)

MedAE (D)

Postoperative SE prediction

 MVF

0.166

0.459

0.289

0.171

 − 0.039

0.365

0.284

0.251

0.007

0.563

0.408

0.293

 Random forest

 − 0.080

0.376

0.245

0.146

0.016

0.351

0.274

0.247

 − 0.001

0.565

0.407

0.322

 LASSO

 − 0.046

0.281

0.199

0.113

0.003

0.362

0.283

0.245

0.006

0.560

0.417

0.347

 SVR

 − 0.038

0.291

0.199

0.089

0.003

0.355

0.281

0.240

0.004

0.570

0.418

0.330

 XGBoost

 − 0.069

0.361

0.240

0.149

0.015

0.348

0.271

0.257

 − 0.002

0.552

0.395

0.290

Postoperative sphere prediction

 MVF

 − 0.178

0.254

0.236

0.183

0.021

0.367

0.283

0.247

0.013

0.559

0.379

0.250

 Random forest

 − 0.036

0.252

0.198

0.155

0.008

0.349

0.269

0.208

 − 0.001

0.553

0.378

0.254

 LASSO

 − 0.033

0.241

0.182

0.144

 − 0.007

0.378

0.289

0.235

0.014

0.561

0.389

0.275

 SVR

 − 0.008

0.256

0.194

0.162

 − 0.007

0.380

0.290

0.225

0.009

0.569

0.397

0.295

 XGBoost

 − 0.056

0.242

0.188

0.141

0.011

0.354

0.272

0.236

 − 0.001

0.543

0.366

0.264

  1. TICL = toric implantable collamer lens; MPE = mean prediction error; MAE = mean absolute error; MedAE = median absolute error; SD = standard deviation; D = diopters; MVF = modified vergence formula; SVR = support vector regression. The numerically smallest SD, MAE and MedAE are marked in bold for each subgroup