evaluation/baseline_vs_se_comparison.csv

Metric,Baseline 3D U-Net,SE Attention 3D U-Net,Difference (SE - Baseline),Better Model,Unit
Best Model Epoch,45.0,48.0,3.0,—,Epoch
Training-Reported Best Validation Dice,0.772125,0.773046,0.0009210000000000607,SE Attention 3D U-Net,Dice
Foreground Macro Dice,0.7752109928303947,0.7610692589733999,-0.014141733856994798,Baseline 3D U-Net,Dice
Foreground Macro IoU,0.6382248216246319,0.6221029962007719,-0.01612182542385998,Baseline 3D U-Net,IoU
Foreground Macro Precision,0.8174835330605084,0.8114060556626961,-0.006077477397812281,Baseline 3D U-Net,Precision
Foreground Macro Recall,0.7473307122885059,0.7258300177342824,-0.021500694554223543,Baseline 3D U-Net,Recall
Tumor Dice,0.9170132244124318,0.9216496372309628,0.004636412818530999,SE Attention 3D U-Net,Dice
Tumor IoU,0.846744618755766,0.8546847750524318,0.007940156296665779,SE Attention 3D U-Net,IoU
Tumor Precision,0.935361335324936,0.9470200761359134,0.011658740810977397,SE Attention 3D U-Net,Precision
Tumor Recall / Sensitivity,0.899371100371264,0.897603068027069,-0.0017680323441949808,Baseline 3D U-Net,Recall
Tumor FNR,0.10062889962873361,0.1023969319729286,0.0017680323441949808,Baseline 3D U-Net,FNR
Tumor FPR,0.01735648436905456,0.013822529972448414,-0.003533954396606146,SE Attention 3D U-Net,FPR
Tumor TP,3808565.0,3758431.0,-50134.0,Baseline 3D U-Net,Voxels
Tumor TN,14900765.0,15001209.0,100444.0,SE Attention 3D U-Net,Voxels
Tumor FP,263193.0,210261.0,-52932.0,SE Attention 3D U-Net,Voxels
Tumor FN,426133.0,428755.0,2622.0,Baseline 3D U-Net,Voxels