SeqSAM: Autoregressive Multiple Hypothesis Prediction for Medical Image Segmentation using SAM
Published in IEEE ISBI, 2025
Recommended citation: SeqSAM: Autoregressive Multiple Hypothesis Prediction for Medical Image Segmentation using SAM (Towle, Chen & Zhou, ISBI 2025). https://arxiv.org/abs/2503.09797
We introduce SeqSAM, an autoregressive approach to medical image segmentation that predicts multiple hypotheses sequentially using the Segment Anything Model. By generating diverse segmentation candidates autoregressively, SeqSAM better captures the inherent ambiguity in medical image segmentation tasks and produces more robust predictions.
Recommended citation: SeqSAM: Autoregressive Multiple Hypothesis Prediction for Medical Image Segmentation using SAM (Towle, Chen & Zhou, ISBI 2025).