alignment.alignment_metrics

alignment.alignment_metrics(
    true_xyz,
    est_xyz,
    brain_atlas,
    *,
    p_good=None,
    valid=None,
    threshold=CONFIDENCE_THRESHOLD,
)

Per-probe scores of estimated channel positions against the reference positions.

Channels count when both positions are finite and non-zero (and valid when given). The region accuracies are over channels whose reference position is inside the brain; an estimate outside the brain counts as wrong. Distances are Euclidean, per channel. With p_good the same scores are also given for the high-confidence (p_good >= threshold) and low-confidence channels, with the high-confidence fraction.

Returns: dict: cosmos_acc, beryl_acc, mean_distance_um, median_distance_um, n_channels, and the *_high_conf / *_low_conf variants and high_conf_fraction when p_good is given.