MUL-T is a lightweight Transformer framework for modelling spatial cellular architecture in multiplexed tissue images. It represents cells as discrete tokens and learns contextualized [CLS] embeddings through masked contextual prediction without task-specific supervision. The abstract reports evaluations on tumour pattern classification, patient-level grading, PD-L1 positivity prediction, and cross-dataset treatment-response prediction. Across these tasks, MUL-T reportedly outperforms classical feature-based baselines and approaches the performance of a foundation ViT model, while using substantially fewer parameters and lower training cost. The supplied record does not include detailed metrics, cohort sizes, or implementation specifics.
No heat snapshots are available in the last 24 hours.