Multimodal Posterior Sampling-based Uncertainty in PD-L1 Segmentation from H&E Images
This work addresses the need for scalable and interpretable biomarker assessment in clinical workflows for immunotherapy guidance, though it is incremental as it builds upon existing nnUNet-v2 methods.
The paper tackles the problem of resource-intensive PD-L1 expression assessment by developing a Bayesian segmentation framework that infers PD-L1 expression from H&E images, achieving mean Dice Score of 0.805 and mean IoU of 0.709 on lung squamous cell carcinoma data.
Accurate assessment of PD-L1 expression is critical for guiding immunotherapy, yet current immunohistochemistry (IHC) based methods are resource-intensive. We present nnUNet-B: a Bayesian segmentation framework that infers PD-L1 expression directly from H&E-stained histology images using Multimodal Posterior Sampling (MPS). Built upon nnUNet-v2, our method samples diverse model checkpoints during cyclic training to approximate the posterior, enabling both accurate segmentation and epistemic uncertainty estimation via entropy and standard deviation. Evaluated on a dataset of lung squamous cell carcinoma, our approach achieves competitive performance against established baselines with mean Dice Score and mean IoU of 0.805 and 0.709, respectively, while providing pixel-wise uncertainty maps. Uncertainty estimates show strong correlation with segmentation error, though calibration remains imperfect. These results suggest that uncertainty-aware H&E-based PD-L1 prediction is a promising step toward scalable, interpretable biomarker assessment in clinical workflows.