Human visual attention-inspired knowledge distillation underlying interpretable computational pathology
Open Access
- 26 November 2025
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 42124
Abstract
Computational pathology leverages advanced deep-learning techniques to analyze medical images with high resolution. However, a trade-off exists between model lightweight, interpretability, and task performance in such real-world scenarios. Knowledge distillation (KD) is widely applied to compress deep learning models while preserving high performance. Despite this, deep learning-based KD often lacks interpretable design, leading to inaccurate attention to images. Inspired by human vision processing, we developed a human vision attention-inspired knowledge distillation (HVisKD) strategy that captures local and global patch relations to construct differentiated features. We employed it in pathological analysis to balance the tradeoff. HVisKD improves performance across various lightweight models in segmentation tasks. More importantly, the attention map of HVisKD showed promoted consistency with human expert-labeled segmentation. Furthermore, we examined HVisKD in a real-world intraoperative pathological diagnosis scenario and achieved an interpretable and fast analysis. Together, HVisKD offers a lightweight and interpretable strategy for computational pathology, aligning deep learning with brain-like information processing for more dependable output.Keywords
Funding Information
- the Institute for Interdisciplinary Information Core Technology (IIISCT), Xi’an, China
- Yu Wang (2022-PUMCH-B-113)
This publication has 67 references indexed in Scilit:
- An anatomic transcriptional atlas of human glioblastomaScience, 2018
- Association of the Extent of Resection With Survival in GlioblastomaJAMA Oncology, 2016
- Multi-class texture analysis in colorectal cancer histologyScientific Reports, 2016
- Attention to local and global levels of hierarchical Navon figures affects rapid scene categorizationFrontiers in Psychology, 2014
- Establishing percent resection and residual volume thresholds affecting survival and recurrence for patients with newly diagnosed intracranial glioblastomaNeuro-Oncology, 2013
- Angiogenesis in GlioblastomaNew England Journal of Medicine, 2013
- Approach to the Intraoperative Consultation for Neurosurgical SpecimensAdvances in Anatomic Pathology, 2011
- Statistical validation of image segmentation quality based on a spatial overlap index1Academic Radiology, 2004
- Spatial Frequency and Selective Attention to Local and Global InformationPerception, 1987
- Perceiving Real-World ScenesScience, 1972