outputs/benchmark_convnext_tiny_eval/metrics.json

{
  "config": "benchmark_convnext_tiny.yaml",
  "backbone": "convnext_tiny",
  "checkpoint": "/mnt/outputs/benchmark_convnext_tiny/checkpoints/best_recall.pt",
  "epoch": 25,
  "val_auroc_slice": 0.5558158159255981,
  "n_slices": 37119,
  "strategies": [
    {
      "strategy": "max",
      "patient_auc": 0.5609,
      "opt_threshold": 0.9769,
      "n_patients": 152,
      "opt_tp": 115,
      "opt_fp": 17,
      "opt_tn": 7,
      "opt_fn": 13,
      "opt_sensitivity": 0.8984,
      "opt_specificity": 0.2917,
      "opt_ppv": 0.8712,
      "opt_npv": 0.35,
      "opt_f1": 0.8846
    },
    {
      "strategy": "topk_mean",
      "patient_auc": 0.5938,
      "opt_threshold": 0.9753,
      "n_patients": 152,
      "opt_tp": 123,
      "opt_fp": 18,
      "opt_tn": 6,
      "opt_fn": 5,
      "opt_sensitivity": 0.9609,
      "opt_specificity": 0.25,
      "opt_ppv": 0.8723,
      "opt_npv": 0.5455,
      "opt_f1": 0.9145
    },
    {
      "strategy": "topk_percent",
      "patient_auc": 0.5929,
      "opt_threshold": 0.9785,
      "n_patients": 152,
      "opt_tp": 58,
      "opt_fp": 6,
      "opt_tn": 18,
      "opt_fn": 70,
      "opt_sensitivity": 0.4531,
      "opt_specificity": 0.75,
      "opt_ppv": 0.9062,
      "opt_npv": 0.2045,
      "opt_f1": 0.6042
    },
    {
      "strategy": "percentile",
      "patient_auc": 0.617,
      "opt_threshold": 0.9778,
      "n_patients": 152,
      "opt_tp": 71,
      "opt_fp": 8,
      "opt_tn": 16,
      "opt_fn": 57,
      "opt_sensitivity": 0.5547,
      "opt_specificity": 0.6667,
      "opt_ppv": 0.8987,
      "opt_npv": 0.2192,
      "opt_f1": 0.686
    },
    {
      "strategy": "lse",
      "patient_auc": 0.6432,
      "opt_threshold": 0.9741,
      "n_patients": 152,
      "opt_tp": 88,
      "opt_fp": 10,
      "opt_tn": 14,
      "opt_fn": 40,
      "opt_sensitivity": 0.6875,
      "opt_specificity": 0.5833,
      "opt_ppv": 0.898,
      "opt_npv": 0.2593,
      "opt_f1": 0.7788
    },
    {
      "strategy": "noisy_or",
      "patient_auc": 0.5208,
      "opt_threshold": 1.0,
      "n_patients": 152,
      "opt_tp": 128,
      "opt_fp": 23,
      "opt_tn": 1,
      "opt_fn": 0,
      "opt_sensitivity": 1.0,
      "opt_specificity": 0.0417,
      "opt_ppv": 0.8477,
      "opt_npv": 1.0,
      "opt_f1": 0.9176
    },
    {
      "strategy": "mean",
      "patient_auc": 0.637,
      "opt_threshold": 0.9688,
      "n_patients": 152,
      "opt_tp": 94,
      "opt_fp": 11,
      "opt_tn": 13,
      "opt_fn": 34,
      "opt_sensitivity": 0.7344,
      "opt_specificity": 0.5417,
      "opt_ppv": 0.8952,
      "opt_npv": 0.2766,
      "opt_f1": 0.8069
    },
    {
      "strategy": "variance_weighted",
      "patient_auc": 0.6115,
      "opt_threshold": 0.9756,
      "n_patients": 152,
      "opt_tp": 118,
      "opt_fp": 17,
      "opt_tn": 7,
      "opt_fn": 10,
      "opt_sensitivity": 0.9219,
      "opt_specificity": 0.2917,
      "opt_ppv": 0.8741,
      "opt_npv": 0.4118,
      "opt_f1": 0.8973
    }
  ],
  "best_strategy": "lse",
  "best_patient_auc": 0.6432,
  "best_threshold": 0.9741,
  "metrics_optimal": {
    "threshold": 0.9741,
    "tp": 88,
    "fp": 10,
    "tn": 14,
    "fn": 40,
    "sensitivity": 0.6875,
    "specificity": 0.5833,
    "ppv": 0.898,
    "npv": 0.2593,
    "f1": 0.7788
  },
  "metrics_0.50": {
    "threshold": 0.5,
    "tp": 128,
    "fp": 23,
    "tn": 1,
    "fn": 0,
    "sensitivity": 1.0,
    "specificity": 0.0417,
    "ppv": 0.8477,
    "npv": 1.0,
    "f1": 0.9176
  },
  "metrics_0.30": {
    "threshold": 0.3,
    "tp": 128,
    "fp": 23,
    "tn": 1,
    "fn": 0,
    "sensitivity": 1.0,
    "specificity": 0.0417,
    "ppv": 0.8477,
    "npv": 1.0,
    "f1": 0.9176
  }
}