基于轻量高频Transformer与特征互补融合的视网膜血管分割

梁礼明,王成斌,钟奕,陈林俊,吴健

Retinal vessel segmentation based on lightweight high-frequency Transformer and feature complementary fusion

Liming LIANG,Chengbin WANG,Yi ZHONG,Linjun CHEN,Jian WU

表 3 采取不同模块时模型在不同数据集上的血管分割性能

Tab.3 Vessel segmentation performance of model with different modules on different datasets

数据集 FCFM LHFT DEAM ACC/% SE/% SP/% AUC/% F1/% DRIVE — — — 97.08 78.57 98.86 98.80 82.53 √ — — 97.10 79.61 98.77 98.81 82.78 — √ — 97.09 79.92 98.75 98.82 82.79 — — √ 97.10 79.75 98.77 98.81 82.76 √ √ — 97.10 79.79 98.77 98.84 82.86 √ — √ 97.10 79.71 98.78 98.81 82.82 — √ √ 97.11 79.78 98.80 98.82 82.87 √ √ √ 97.12 80.23 98.74 98.83 82.99 STARE — — — 97.52 79.26 99.02 99.01 82.98 √ — — 97.57 79.62 99.04 99.08 83.30 — √ — 97.53 79.88 99.04 99.05 83.41 — — √ 97.54 79.18 99.07 99.01 83.20 √ √ — 97.60 80.35 99.02 99.10 83.62 √ — √ 97.59 80.33 99.01 99.10 83.57 — √ √ 97.60 80.22 99.02 99.08 83.58 √ √ √ 97.62 80.48 99.02 99.12 83.73 CHASE_DB1 — — — 97.46 80.90 98.57 98.85 80.09 √ — — 97.59 81.05 98.71 98.99 80.97 — √ — 97.54 81.14 98.77 98.99 81.03 — — √ 97.57 80.35 98.78 98.95 80.86 √ √ — 97.64 81.17 98.75 99.03 81.30 √ — √ 97.61 80.52 98.76 98.98 80.97 — √ √ 97.62 81.19 98.72 98.97 81.16 √ √ √ 97.65 81.30 98.75 98.99 81.37