# Results

## Stage 1 - Clinical

### Late AMD

```text
Endpoint: late_amd (Status_late_amd)
Feature set: grading
Loaded rows: 2246
Initial number of features: 4
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=303, val=40, test=90
Features after variance filter: 4
Features after top-k selection: 4

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 303,
    "c_index": 0.8855778758440338
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 40,
    "c_index": 0.8816348923330133
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 90,
    "c_index": 0.8732695005183875
  }
}

Saved outputs to: runs/stage1_late_amd_grading
```

### Any GA

```text
Endpoint: anyga (Status_anyga)
Feature set: grading
Loaded rows: 2246
Initial number of features: 4
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=188, val=27, test=58
Features after variance filter: 4
Features after top-k selection: 4

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 188,
    "c_index": 0.8701093069888003
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 27,
    "c_index": 0.8747786740114106
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 58,
    "c_index": 0.8789303153379272
  }
}

Saved outputs to: runs/stage1_anyga_grading
```

### NV

```text
Endpoint: nv (Status_nv)
Feature set: grading
Loaded rows: 2246
Initial number of features: 4
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=181, val=23, test=48
Features after variance filter: 4
Features after top-k selection: 4

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 181,
    "c_index": 0.8859630279871941
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 23,
    "c_index": 0.8588550151268327
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 48,
    "c_index": 0.8415194442925121
  }
}

Saved outputs to: runs/stage1_nv_grading
```

## Stage 1 - Grading Clinical

### Late AMD

```text
Endpoint: late_amd (Status_late_amd)
Feature set: grading_clinical
Loaded rows: 2246
Initial number of features: 6
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=303, val=40, test=90
Features after variance filter: 6
Features after top-k selection: 6

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 303,
    "c_index": 0.889178746182056
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 40,
    "c_index": 0.881909203127143
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 90,
    "c_index": 0.8796731109349272
  }
}

Saved outputs to: runs/stage1_late_amd_grading_clinical
```

### Any GA

```text
Endpoint: anyga (Status_anyga)
Feature set: grading_clinical
Loaded rows: 2246
Initial number of features: 6
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=188, val=27, test=58
Features after variance filter: 6
Features after top-k selection: 6

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 188,
    "c_index": 0.8721511713209762
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 27,
    "c_index": 0.8756639779657682
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 58,
    "c_index": 0.8843152466572354
  }
}

Saved outputs to: runs/stage1_anyga_grading_clinical
```

### NV

```text
Endpoint: nv (Status_nv)
Feature set: grading_clinical
Loaded rows: 2246
Initial number of features: 6
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=181, val=23, test=48
Features after variance filter: 6
Features after top-k selection: 6

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 181,
    "c_index": 0.8898688423009398
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 23,
    "c_index": 0.8587386548754945
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 48,
    "c_index": 0.8480008751299021
  }
}

Saved outputs to: runs/stage1_nv_grading_clinical
```

## Stage 2 - Deep Clinical

### Late AMD

```text
>>> Late AMD: block4
Endpoint: late_amd (Status_late_amd)
Feature set: deep_clinical
Loaded rows: 2246
Deep feature sanity check:
  shape: (2246, 512)
  NaN:   0
  Inf:   0
  min:   0
  max:   23.0515
  mean:  0.513523
  std:   0.945009
Rows after feature merge: 2246
Initial number of features: 514
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=303, val=40, test=90
[info] Dropping low-variance features: 14 / 514
Features after variance filter: 500
[info] Selected 4 features from LE_DRUS
  LE_DRUS_106: score=0.8624, raw_c=0.1376
  LE_DRUS_084: score=0.8620, raw_c=0.1380
  LE_DRUS_125: score=0.8608, raw_c=0.1392
  LE_DRUS_111: score=0.8606, raw_c=0.1394
[info] Selected 4 features from RE_DRUS
  RE_DRUS_084: score=0.8580, raw_c=0.1420
  RE_DRUS_111: score=0.8576, raw_c=0.1424
  RE_DRUS_125: score=0.8573, raw_c=0.1427
  RE_DRUS_106: score=0.8568, raw_c=0.1432
[info] Selected 4 features from LE_PIG
  LE_PIG_013: score=0.8566, raw_c=0.8566
  LE_PIG_066: score=0.8565, raw_c=0.8565
  LE_PIG_042: score=0.8563, raw_c=0.8563
  LE_PIG_125: score=0.8562, raw_c=0.8562
[info] Selected 4 features from RE_PIG
  RE_PIG_013: score=0.8410, raw_c=0.8410
  RE_PIG_042: score=0.8409, raw_c=0.8409
  RE_PIG_006: score=0.8409, raw_c=0.8409
  RE_PIG_125: score=0.8406, raw_c=0.8406
[info] Block-balanced selection kept 18 / 500 features including 2 non-deep features
Features after block-balanced top-k selection: 18

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 303,
    "c_index": 0.8985550550090405
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 40,
    "c_index": 0.8714853929502125
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 90,
    "c_index": 0.887235469903031
  }
}

Saved outputs to: runs_2/stage2_late_amd_deep_clinical_block4_p001
```

### Any GA

```text
>>> Any GA: block2
Endpoint: anyga (Status_anyga)
Feature set: deep_clinical
Loaded rows: 2246
Deep feature sanity check:
  shape: (2246, 512)
  NaN:   0
  Inf:   0
  min:   0
  max:   23.0515
  mean:  0.513523
  std:   0.945009
Rows after feature merge: 2246
Initial number of features: 514
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=188, val=27, test=58
[info] Dropping low-variance features: 14 / 514
Features after variance filter: 500
[info] Selected 2 features from LE_DRUS
  LE_DRUS_106: score=0.8300, raw_c=0.1700
  LE_DRUS_125: score=0.8299, raw_c=0.1701
[info] Selected 2 features from RE_DRUS
  RE_DRUS_084: score=0.8279, raw_c=0.1721
  RE_DRUS_111: score=0.8275, raw_c=0.1725
[info] Selected 2 features from LE_PIG
  LE_PIG_066: score=0.8525, raw_c=0.8525
  LE_PIG_013: score=0.8524, raw_c=0.8524
[info] Selected 2 features from RE_PIG
  RE_PIG_006: score=0.8268, raw_c=0.8268
  RE_PIG_042: score=0.8266, raw_c=0.8266
[info] Block-balanced selection kept 10 / 500 features including 2 non-deep features
Features after block-balanced top-k selection: 10

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 188,
    "c_index": 0.8855251789178527
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 27,
    "c_index": 0.8853039543576628
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 58,
    "c_index": 0.8996029754027289
  }
}

Saved outputs to: runs_2/stage2_anyga_deep_clinical_block2_p001
```

### NV

```text
>>> NV: block2
Endpoint: nv (Status_nv)
Feature set: deep_clinical
Loaded rows: 2246
Deep feature sanity check:
  shape: (2246, 512)
  NaN:   0
  Inf:   0
  min:   0
  max:   23.0515
  mean:  0.513523
  std:   0.945009
Rows after feature merge: 2246
Initial number of features: 514
Rows after cleaning: 2246
Split sizes: train=1580, val=221, test=445
Events: train=181, val=23, test=48
[info] Dropping low-variance features: 14 / 514
Features after variance filter: 500
[info] Selected 2 features from LE_DRUS
  LE_DRUS_084: score=0.8713, raw_c=0.1287
  LE_DRUS_106: score=0.8701, raw_c=0.1299
[info] Selected 2 features from RE_DRUS
  RE_DRUS_026: score=0.8624, raw_c=0.1376
  RE_DRUS_041: score=0.8619, raw_c=0.1381
[info] Selected 2 features from LE_PIG
  LE_PIG_114: score=0.8483, raw_c=0.8483
  LE_PIG_082: score=0.8482, raw_c=0.8482
[info] Selected 2 features from RE_PIG
  RE_PIG_066: score=0.8421, raw_c=0.8421
  RE_PIG_013: score=0.8416, raw_c=0.8416
[info] Block-balanced selection kept 10 / 500 features including 2 non-deep features
Features after block-balanced top-k selection: 10

Metrics
{
  "train": {
    "split": "train",
    "n": 1580,
    "events": 181,
    "c_index": 0.8887245688319736
  },
  "val": {
    "split": "val",
    "n": 221,
    "events": 23,
    "c_index": 0.8161508028857343
  },
  "test": {
    "split": "test",
    "n": 445,
    "events": 48,
    "c_index": 0.8400700103921676
  }
}

Saved outputs to: runs_2/stage2_nv_deep_clinical_block2_p001
```