Farndale, Lucas and Insall, Robert and Yuan, Ke (2025) TriDeNT: Triple deep network training for privileged knowledge distillation in histopathology. Medical Image Analysis, 102: 103479. ISSN 1361-8415
AI Summary:
TriDeNT is a novel self-supervised method that utilises privileged data not available during inference to improve performance. The method outperforms other state-of-the-art methods in downstream tasks, with observed improvements of up to 101.AI Topics:
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.
Farndale, Lucas
Author
Farndale, Lucas and Insall, Robert and Yuan, Ke (2025) TriDeNT: Triple deep network training for privileged knowledge distillation in histopathology. Medical Image Analysis, 102: 103479. ISSN 1361-8415
See full publications listInsall, Robert
Author
Farndale, Lucas and Insall, Robert and Yuan, Ke (2025) TriDeNT: Triple deep network training for privileged knowledge distillation in histopathology. Medical Image Analysis, 102: 103479. ISSN 1361-8415
Mandrou, Elena and Thomason, Peter A. and Paschke, Peggy I. and Paul, Nikki R. and Tweedy, Luke and Insall, Robert (2024) A reliable system for quantitative g-protein activation imaging in cancer cells. Cells, 13 (13): 1114. ISSN 2073-4409
Kapitany, Valentin and Fatima, Areeba and Zickus, Vytautas and Whitelaw, Jamie and McGhee, Ewan and Insall, Robert and Machesky, Laura and Faccio, Daniele (2024) Single-sample image-fusion upsampling of fluorescence lifetime images. Science Advances, 10 (21): eadn0139. ISSN 2375-2548
See full publications listYuan, Ke
Author
Farndale, Lucas and Insall, Robert and Yuan, Ke (2025) TriDeNT: Triple deep network training for privileged knowledge distillation in histopathology. Medical Image Analysis, 102: 103479. ISSN 1361-8415
Ji, Yanni and Cutiongco, Marie F.A. and Jensen, Bjørn Sand and Yuan, Ke (2025) Generating realistic single-cell images from CellProfiler representations. Medical Image Analysis. ISSN 1361-8415 (In Press)
Coudray, Nicolas and Juarez, Michelle C. and Criscito, Maressa C. and Claudio Quiros, Adalberto and Wilken, Reason and Jackson Cullison, Stephanie R. and Stevenson, Mary L. and Doudican, Nicole A. and Yuan, Ke and Aquino, Jamie D. and Klufas, Daniel M. and North, Jeffrey P. and Yu, Siegrid S. and Murad, Fadi and Ruiz, Emily and Schmults, Chrysalyne D. and Cardona Machado, Cristian D. and Cañueto, Javier and Choudhary, Anirudh and Hughes, Alysia N. and Stockard, Alyssa and Leibovit-Reiben, Zachary and Mangold, Aaron R. and Tsirigos, Aristotelis and Carucci, John A. (2025) Self supervised artificial intelligence predicts poor outcome from primary cutaneous squamous cell carcinoma at diagnosis. npj Digital Medicine, 8 (1): 105. ISSN 2398-6352
See full publications listAvailable under License Creative Commons Attribution.
Download (3MB) | Preview
 |