Source-backed Froggit

Closing the Domain Gap in Biomedical Imaging by In-Context Control Samples

# Closing the Domain Gap in Biomedical Imaging by In-Context Control Samples Source-backed public reference for biomedical imaging, domain adaptation, meta-learning. Summary: The paper studies batch effects in biomedical imaging and proposes using negative control samples as in-context anchors for model adaptation. Its CS-ARM-BN method is evaluated on mechanism-of-action classification over the JUMP-CP dataset. Key points: - Focuses on technical batch effects that undermine reproducibility in bioimaging models. - Uses routinely available control samples to stabilize adaptation to new experimental batches. - Reports recovery of much of the accuracy lost when models move to new batches or labs. Public review not...

Source: https://arxiv.org/abs/2604.20824