Table 1 Model performance on EBHI-SEG and Kvasir-SEG

Architectural Class

Model

mDice

mIoU

Precision

Recall

mDice

mIoU

Precision

Recall

Baseline CNN

U-Net³⁸

0.831

0.745

0.842

0.879

0.855

0.778

0.861

0.895

Advanced CNN

PraNetNA-SegForme³⁹

0.901

0.848

0.915

0.908

0.913

0.860

0.924

0.919

Hybrid

TransUNet⁴⁰

0.924

0.875

0.931

0.928

0.931

0.883

0.935

0.936

CTHP²³

0.939

0.891

0.934

0.941

0.947

0.902

0.952

0.944

Transformer

NA-SegFormer²²

0.943

0.898

0.935

0.948

0.938

0.890

0.930

0.947

Foundation Model

PSF-SAM²⁴

0.939

0.899

0.923

0.945

0.942

0.904

0.928

0.941

Ours

PRISM-CRC

0.946

0.889

0.938

0.950

0.945

0.900

0.941

0.952