PURPOSE:To develop a decision forest for prediction of laser refractive surgery outcome.
METHODS:Data from consecutive cases of patients who underwent LASIK or photorefractive surgeries during a 12-year period in a single center were assembled into a single dataset. Training of machine-learning classifiers and testing were performed with a statistical classifier algorithm. The decision forest was created by feature vectors extracted from 17,592 cases and 38 clinical parameters for each patient. A 10-fold cross-validation procedure was applied to estimate the predictive value of the decision forest when applied to new patients.
RESULTS:Analysis included patients younger than 40 years who were not treated for monovision. Efficacy of 0.7 or greater and 0.8 or greater was achieved in 16,198 (92.0%) and 14,945 (84.9%) eyes, respectively. Efficacy of less than 0.4 and less than 0.5 was achieved in 322 (1.8%) and 506 (2.9%) eyes, respectively. Patients in the low efficacy group (< 0.4) had statistically significant differences compared with the high efficacy group (≥ 0.8), yet were clinically similar (mean differences between groups of 0.7 years, of 0.43 mm in pupil size, of 0.11 D in cylinder, of 0.22 logMAR in preoperative CDVA, of 0.11 mm in optical zone size, of 1.03 D in actual sphere treatment, and of 0.64 D in actual cylinder treatment). The preoperative subjective CDVA had the highest gain (most important to the model). Correlations analysis revealed significantly decreased efficacy with increased age (r = −0.67, P < .001), central corneal thickness (r = −0.40, P < .001), mean keratometry (r = −0.33, P < .001), and preoperative CDVA (r = −0.47, P < .001). Efficacy increased with pupil size (r = 0.20, P < .001).
CONCLUSIONS:This model could support clinical decision making and may lead to better individual risk assessment. Expanding the role of machine learning in analyzing big data from refractive surgeries may be of interest.