Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces Temporal Inversion-aware Learning and Alignment (TILA), a framework that leverages temporal inversion (reversing image pairs) as a supervisory signal to improve vision-language models' sensitivity to directional changes in chest X-rays. TILA incorporates inversion-aware objectives during pretraining, fine-tuning, and inference to explicitly learn temporal order alongside appearance modeling. Experiments on multiple datasets demonstrate that TILA enhances progression classification and temporal embedding alignment across various architectures.
Flipping the order of before-and-after chest X-rays during training dramatically improves a model's ability to detect disease progression.
Recent advances in vision--language pretraining have enabled strong medical foundation models, yet most analyze radiographs in isolation, overlooking the key clinical task of comparing prior and current images to assess interval change. For chest radiographs (CXRs), capturing interval change is essential, as radiologists must evaluate not only the static appearance of findings but also how they evolve over time. We introduce TILA (Temporal Inversion-aware Learning and Alignment), a simple yet effective framework that uses temporal inversion, reversing image pairs, as a supervisory signal to enhance the sensitivity of existing temporal vision-language models to directional change. TILA integrates inversion-aware objectives across pretraining, fine-tuning, and inference, complementing conventional appearance modeling with explicit learning of temporal order. We also propose a unified evaluation protocol to assess order sensitivity and consistency under temporal inversion, and introduce MS-CXR-Tretrieval, a retrieval evaluation set constructed through a general protocol that can be applied to any temporal CXR dataset. Experiments on public datasets and real-world hospital cohorts demonstrate that TILA consistently improves progression classification and temporal embedding alignment when applied to multiple existing architectures.