PiPViT: Patch-based Visual Interpretable Prototypes for Retinal Image Analysis

Background and Objective: Prototype-based methods improve interpretability by learning fine-grained part-prototypes; however, their visualization in the input pixel space is not always consistent with human-understandable biomarkers. In addition, well-known prototype-based approaches typically learn extremely granular prototypes that are less interpretable in medical imaging, where both the presence and extent of biomarkers and lesions are critical.

@article{oghbaie2025_2506.10669,
  title={ PiPViT: Patch-based Visual Interpretable Prototypes for Retinal Image Analysis },
  author={ Marzieh Oghbaie and Teresa Araújo and Hrvoje Bogunović },
  journal={arXiv preprint arXiv:2506.10669},
  year={ 2025 }
}