Machine Learning Model Inputs
Variable Name	Definition/Calculation	Role in Model
Patient_id	Patient identifier	Not used in training (only for tracking)
S_auto	Sphere measured by autorefractor (D)	Raw input
C_auto	Cylinder measured by autorefractor (D)	Raw input
Axis_auto	Cylinder axis measured by autorefractor (°)	Raw input
M_auto	S_auto + C_auto/2 (spherical equivalent, SE)	Power vector input
J0_auto	−(C_auto/2) × cos (2 × Axis_auto)	Power vector input (horizontal/vertical astigmatism)
J45_auto	−(C_auto/2) × sin (2 × Axis_auto)	Power vector input (oblique astigmatism)
M_a2, J0_a2, J45_a2	Squared terms of power vector components	Non-linear feature expansion
M_aJ0_a, M_aJ45_a, J0_a*J45_a	Interaction terms among power vectors	Feature interactions
Cabs_a	Absolute value of C_auto	C
age	Patient age (years)	Demographic input
age2	Squared age	Non-linear age effect
sex_Female/Male/Other	Dummy variables derived from sex	Demographic input
K1, K2	Keratometry values (D)	Corneal shape input
Kmean	(K1 + K2)/2	Mean corneal curvature
Kdelta	K1 − K2	K1 − K2
Kratio	K1/K2	Corneal asymmetry
age Kmean, age Kdelta	Interaction of age with corneal shape	Interaction features
S_man	Sphere measured manually by clinician (D)	Target (training only)
C_man	Cylinder measured manually by clinician (D)	Target (training only)
Axis_man	Cylinder axis measured manually by clinician (°)	Target (training only)
M_true	Power vector M from manual refraction	Target (training only)
J0_true	Power vector J0 from manual refraction	Target (training only)
J45_true	Power vector J45 from manual refraction	Target (training only)
SE_true	Manual spherical equivalent (S_man + C_man/2)	Target (training only)
Cabs_true	Absolute cylinder from manual refraction	Target (training only)