Fig. 4 | Scientific Reports

Fig. 4

From: Vision language models versus machine learning models performance on polyp detection and classification in colonoscopy images

Fig. 4

Polyp classification performance of top-performing models. Confusion matrices of polyp classification are provided for the top-performing classical machine learning model (a: Random Forest), convolutional neural network (b: ResNet-50), highest-performing vision-language model (c: GPT-4.1), and the contrastive vision-language encoder fine-tuned on external general medical imaging data (d: BiomedCLIP). Abbreviations: AC, Adenocarcinoma; TA, Tubular Adenoma; TVA, Tubulovillous Adenoma; VA, Villous Adenoma; HP, Hyperplastic Polyp; IP, Inflammatory Polyp; No-A: No answer provided; 2OP: two options (polyp type) were selected.

Back to article page