16 papers
Self-supervision drives representational convergence in medical foundation models more than clinical supervision
Soroosh Tayebi Arasteh, Sebastian Ziegelmayer, Mahshad Lotfinia +4
Medical image encoders from different groups are increasingly treated as interchangeable, on the assumption that scale and clinical supervision concentrate their representations on…
CANDOR: Chance-Calibrated Discordance in Frozen Foundation Encoders
Soroosh Tayebi Arasteh, Sven Nebelung, Daniel Truhn
Frozen encoders are chosen by how well a lightweight head reads a finding from their features, not whether the geometry separates it. Nearest-neighbor discordance does, but with un…
The strength of clinical evidence is recoverable from language model representations but not from their stated grades
Soroosh Tayebi Arasteh
Large language models (LLMs) increasingly summarize clinical evidence, where a claim's weight depends on how strongly it is supported. Yet these models convey confidence poorly, an…
Vision-language models for chest radiography do not always need the image
Mahshad Lotfinia, Sebastian Ziegelmayer, Lisa Adams +3
Medical vision-language models report strong chest radiograph accuracy, and this is increasingly read as evidence that they use the image. That inference is unsafe: a model exploit…
Cross-modal linkage risk in clinical vision-language models
Soroosh Tayebi Arasteh, Mahshad Lotfinia, Sven Nebelung +1
Vision-language models (VLMs) trained on paired chest radiographs and radiology reports learn a shared embedding space that can preserve instance-level image-report correspondence.…
Perceptual implications of automatic anonymization in pathological speech
Soroosh Tayebi Arasteh, Saba Afza, Tri-Thien Nguyen +11
Automatic anonymization is increasingly used to enable ethical sharing of clinical speech, yet its perceptual and clinical consequences remain undercharacterized. We present a huma…