Spaces:
Sleeping
Sleeping
| """ | |
| Script includes functions to compute evaluation metrics | |
| """ | |
| import numpy as np | |
| from sklearn.metrics import roc_auc_score | |
| # Function to compute Dice index score for a tissue region indicated by label | |
| def compute_dice(pred, gt, label): | |
| pred = pred.flatten() | |
| gt = gt.flatten() | |
| pred[pred!=label] = 0 | |
| pred[pred==label] = 1 | |
| gt[gt != label] = 0 | |
| gt[gt == label] = 1 | |
| pred_pixels = np.sum(pred) | |
| gt_pixels = np.sum(gt) | |
| denom = (pred_pixels + gt_pixels) | |
| if (gt_pixels == 0): | |
| return -1 | |
| return np.sum(pred[gt == 1]) * 2.0 / denom | |
| # Function to compute AUC-ROC for a tissue region indicated by label | |
| def compute_auc_roc(pred, gt, label): | |
| pred_binary = np.where(pred == label, 1, 0) | |
| gt_binary = np.where(gt == label, 1, 0) | |
| if (np.sum(gt_binary) == 0 or np.all(gt_binary == 1) or np.all(pred_binary == 1)): | |
| return -1 | |
| if (np.sum(pred_binary) == 0): | |
| return 0 | |
| pred_flat = pred_binary.flatten() | |
| gt_flat = gt_binary.flatten() | |
| auc_roc = roc_auc_score(gt_flat, pred_flat) | |
| return auc_roc | |
| # Function to compute Accuracy for a tissue region indicated by label | |
| def compute_accuracy_metrics(pred, gt): | |
| correct_predictions = np.sum(pred == gt) | |
| overall_count = pred.size | |
| return correct_predictions, overall_count |