Table 1 Coefficients of determination (R2).

From: Machine learning enables non-Gaussian investigation of changes to peripheral nerves related to electrical stimulation

Training vectors

Target values

Surgery

Shannon

Manual morphometrics

0.227

0.311

Manual fiber density (normal cells only)

0.150

0.392

Manual fiber density (normal, degenerated, & hypomyelinated cells)

0.593

0.536

Automated Morphometrics

0.121

0.149

Automated cell-wise windowed metrics

 Small window

0.504

0.431

 Large window

0.709

0.760

Automated morphometrics & cell-wise windowed metrics

 Small window

0.519

0.404

 Large window

0.677

0.605

Automated pixel-wise windowed metrics

 Small window

0.525

0.404

 Large window

0.724

0.730

  1. Support vector machine regressions were trained on subsets of measurements/features called training vectors. Performance was evaluated by measuring the coefficient of determination between model outputs and sample target values. For surgical regressions, the target value was whether samples had an electrode surgically implanted (1) or not (0). For stimulation regressions, model performance was evaluated by using each samples’ stimulation level (Shannon k) as the target value.
Back to article page