TriDeNT: Triple Deep Network Training for Privileged Knowledge Distillation in Histopathology
This addresses the challenge of leveraging scarce or costly data in histopathology to improve model performance for routine inputs, representing a domain-specific advancement.
The paper tackles the problem of computational pathology models not using data unavailable during inference, such as additional stains and spatial transcriptomics, by introducing TriDeNT, a self-supervised method for privileged knowledge distillation, which outperforms state-of-the-art methods with improvements of up to 101% in downstream tasks.
Computational pathology models rarely utilise data that will not be available for inference. This means most models cannot learn from highly informative data such as additional immunohistochemical (IHC) stains and spatial transcriptomics. We present TriDeNT, a novel self-supervised method for utilising privileged data that is not available during inference to improve performance. We demonstrate the efficacy of this method for a range of different paired data including immunohistochemistry, spatial transcriptomics and expert nuclei annotations. In all settings, TriDeNT outperforms other state-of-the-art methods in downstream tasks, with observed improvements of up to 101%. Furthermore, we provide qualitative and quantitative measurements of the features learned by these models and how they differ from baselines. TriDeNT offers a novel method to distil knowledge from scarce or costly data during training, to create significantly better models for routine inputs.