Table 2 Ablation study results of the proposed model.

From: Deep intelligence: a four-stage deep network for accurate brain tumor segmentation

Techniques

Loss

Dice coefficient

Jaccard similarity

Tversky index

Accuracy

2D VNET

0.00694

0.91926

0.95793

0.99305

0.99375

2D VNET with custom LCFT loss function

0.00655

0.98976

0.99485

0.99678

0.99646

2D VNET with CBF Module

0.00633

0.93393

0.96583

0.99366

0.99381

2D VNET—custom LCFT loss function—4 stages

0.00616

0.99051

0.99522

0.99693

0.99657

2D VNET—custom LCFT loss function—5 stages

0.4161

0.6386

0.4701

0.6386

0.913448

Proposed model (2D VNET—custom LCFT loss function—CBF module—4 stages)

0.00614

0.99287

0.99642

0.99743

0.99717

Back to article page