Spectral Prognostic Score

The score is Neyman-Pearson optimal only under Gaussian, zero-mean, isotropic-healthy assumptions. Modes with disease variance above healthy isotropic variance increase the score, while modes below it decrease the score. The prognostic score is computed from squared patient projections onto disease eigenmodes. For non-isotropic healthy covariance, the optimal statistic requires a generalized eigenvalue problem involving both healthy and disease Hamiltonians.