Table 6 Token-level similarity metrics between generated and expert reports

From: PlaqueCap: lesion-centered captioning of atherosclerotic plaques in intravascular ultrasound using vision-language models and prompt injection

Model

BLEU-4 ↑

ROUGE-L ↑

METEOR ↑

CIDEr ↑

BERTScore ↑

BLIP-2

0.342

0.491

0.267

1.17

0.841

GIT

0.319

0.472

0.243

1.08

0.832

PathoCap

0.289

0.453

0.225

0.96

0.825

PMC-VQA

0.276

0.432

0.221

0.91

0.819

PlaqueCap (Ours)

0.396

0.522

0.301

1.33

0.854

  1. Our method achieves the highest scores across all standard metrics, indicating superior alignment with expert-written language.
Back to article page