PIKACHU: Prototypical In-context Knowledge Adaptation for Clinical Heterogeneous Usage

Amar Kumar, Zahra TehraniNasab, Emily Kaczmarek, Tal Arbel
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:2922-2940, 2026.

Abstract

Medical imaging systems increasingly rely on large vision language foundation models (VLFMs) trained on diverse biomedical corpora, yet these models remain difficult to adapt to new clinical tasks without costly fine-tuning and large annotated datasets. We present (Prototypical In-Context Knowledge Adaptation for Clinical Heterogeneous Usage), a lightweight and generalizable framework that enables rapid few-shot adaptation of frozen medical FMs using only a handful of labelled examples. Unlike prior approaches that modify backbone weights or introduce heavy attention-based adapters, PIKACHU performs all task adaptation directly in the FM feature space through in-context prototypical reasoning. Given a small support set, the framework constructs class prototypes by averaging normalized embeddings from a frozen VLFM image encoder and performs prediction on query images using temperature-scaled cosine similarity. Only a single temperature parameter is learned. We evaluate PIKACHU across three heterogeneous medical imaging datasets - dermatological images (ISIC), Optical Coherence Tomography (OCT), and Diabetic Retinopathy (DR), using established vision models (SigLIP, PubMedCLIP, DinoV2, and ViT) as backbones. The proposed in-context learning (ICL) strategy consistently outperforms the baseline (zero-shot) approaches across all datasets and architectures, achieving substantial improvements in both accuracy and AUC. Notably, with PubMedCLIP as the backbone, PIKACHU achieves 0.69 accuracy on the ISIC dataset, 0.72 on OCT, and 0.79 on DR, demonstrating robust generalization across diverse clinical imaging modalities. These results highlight the promise of feature-space in-context learning as an efficient and deployable paradigm for test-time adaptation of foundation models, without the need for extensive retraining.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-kumar26b, title = {PIKACHU: Prototypical In-context Knowledge Adaptation for Clinical Heterogeneous Usage}, author = {Kumar, Amar and TehraniNasab, Zahra and Kaczmarek, Emily and Arbel, Tal}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {2922--2940}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/kumar26b/kumar26b.pdf}, url = {https://proceedings.mlr.press/v315/kumar26b.html}, abstract = {Medical imaging systems increasingly rely on large vision language foundation models (VLFMs) trained on diverse biomedical corpora, yet these models remain difficult to adapt to new clinical tasks without costly fine-tuning and large annotated datasets. We present (Prototypical In-Context Knowledge Adaptation for Clinical Heterogeneous Usage), a lightweight and generalizable framework that enables rapid few-shot adaptation of frozen medical FMs using only a handful of labelled examples. Unlike prior approaches that modify backbone weights or introduce heavy attention-based adapters, PIKACHU performs all task adaptation directly in the FM feature space through in-context prototypical reasoning. Given a small support set, the framework constructs class prototypes by averaging normalized embeddings from a frozen VLFM image encoder and performs prediction on query images using temperature-scaled cosine similarity. Only a single temperature parameter is learned. We evaluate PIKACHU across three heterogeneous medical imaging datasets - dermatological images (ISIC), Optical Coherence Tomography (OCT), and Diabetic Retinopathy (DR), using established vision models (SigLIP, PubMedCLIP, DinoV2, and ViT) as backbones. The proposed in-context learning (ICL) strategy consistently outperforms the baseline (zero-shot) approaches across all datasets and architectures, achieving substantial improvements in both accuracy and AUC. Notably, with PubMedCLIP as the backbone, PIKACHU achieves 0.69 accuracy on the ISIC dataset, 0.72 on OCT, and 0.79 on DR, demonstrating robust generalization across diverse clinical imaging modalities. These results highlight the promise of feature-space in-context learning as an efficient and deployable paradigm for test-time adaptation of foundation models, without the need for extensive retraining.} }
Endnote
%0 Conference Paper %T PIKACHU: Prototypical In-context Knowledge Adaptation for Clinical Heterogeneous Usage %A Amar Kumar %A Zahra TehraniNasab %A Emily Kaczmarek %A Tal Arbel %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-kumar26b %I PMLR %P 2922--2940 %U https://proceedings.mlr.press/v315/kumar26b.html %V 315 %X Medical imaging systems increasingly rely on large vision language foundation models (VLFMs) trained on diverse biomedical corpora, yet these models remain difficult to adapt to new clinical tasks without costly fine-tuning and large annotated datasets. We present (Prototypical In-Context Knowledge Adaptation for Clinical Heterogeneous Usage), a lightweight and generalizable framework that enables rapid few-shot adaptation of frozen medical FMs using only a handful of labelled examples. Unlike prior approaches that modify backbone weights or introduce heavy attention-based adapters, PIKACHU performs all task adaptation directly in the FM feature space through in-context prototypical reasoning. Given a small support set, the framework constructs class prototypes by averaging normalized embeddings from a frozen VLFM image encoder and performs prediction on query images using temperature-scaled cosine similarity. Only a single temperature parameter is learned. We evaluate PIKACHU across three heterogeneous medical imaging datasets - dermatological images (ISIC), Optical Coherence Tomography (OCT), and Diabetic Retinopathy (DR), using established vision models (SigLIP, PubMedCLIP, DinoV2, and ViT) as backbones. The proposed in-context learning (ICL) strategy consistently outperforms the baseline (zero-shot) approaches across all datasets and architectures, achieving substantial improvements in both accuracy and AUC. Notably, with PubMedCLIP as the backbone, PIKACHU achieves 0.69 accuracy on the ISIC dataset, 0.72 on OCT, and 0.79 on DR, demonstrating robust generalization across diverse clinical imaging modalities. These results highlight the promise of feature-space in-context learning as an efficient and deployable paradigm for test-time adaptation of foundation models, without the need for extensive retraining.
APA
Kumar, A., TehraniNasab, Z., Kaczmarek, E. & Arbel, T.. (2026). PIKACHU: Prototypical In-context Knowledge Adaptation for Clinical Heterogeneous Usage. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:2922-2940 Available from https://proceedings.mlr.press/v315/kumar26b.html.

Related Material