Skip to main content

Table 1 Performance of the different methods for classifying fixation as central vs para-central

From: New pediatric vision screener, part II: electronics, software, signal processing and validation

 

Simple threshold (method 1)

Discriminant analysis (method 2)

 

2D

3D

4D

\(\frac{{\left( {P_{2.5} + P_{6.5} } \right)}}{{P_{4.5} }}\)

\(\frac{{\left( {P_{2.5} + P_{6.5} } \right)}}{{P_{4.5} }} vs \frac{{\left( {P_{3.5} + P_{5.5} } \right)}}{{P_{4.5} }}\)

\(\frac{{P_{6.5} }}{{P_{4.5} }} vs \frac{{P_{2.5} }}{{P_{4.5} }} vs \frac{{\left( {P_{3.5} + P_{5.5} } \right)}}{{P_{4.5} }}\)

\(\frac{{P_{6.5} }}{{P_{4.5} }}\, vs \frac{{P_{2.5} }}{{P_{4.5} }}\, vs \frac{{P_{3.5} }}{{P_{4.5} }}\, vs\frac{{P_{5.5} }}{{P_{4.5} }}\)

θ = 0.8750

a 0  = 0.743, a 1  = 0.318

a 0  = 0.768, a 1  = 0.056, a 2  = −0.289

a 0  = 0.779, a 1  = 0.016, a 2  = −0.283, a 3  = 0.083

Sensitivity

0.9917

0.9083

0.9417

0.9417

Specificity

0.9625

0.9771

1.0000

1.0000

  1. All methods use spectral power at odd multiples of half the scanning frequency f s . Method 1 uses optimized simple threshold, whereas Method 2 utilizes 2-, 3-, and 4-way linear discriminant functions, whose coefficients are shown in the table