What Makes High-Magnification Knowledge Transferable? A Study of Cross-Resolution Distillation in Whole-Slide Imaging

2026-09-28

AI Summary

This study addresses the unclear conditions under which high-magnification features can be effectively transferred to lower magnifications in cross-resolution knowledge distillation for whole slide images. We propose a decomposition-based cross-resolution distillation framework that decouples teacher access, representation loss, and model redundancy to localize transfer failure points. This approach reveals a disconnect between reconstruction error and downstream performance, challenging conventional evaluation paradigms that rely solely on reconstruction accuracy. Extensive experiments across ten pathological cohorts demonstrate that preserving native features while incorporating region-wise teacher means consistently improves performance in classification, grading, and survival prediction tasks. These findings provide a reliable diagnostic and optimization strategy for cross-resolution pathological analysis.

Abstract

Cross-resolution knowledge distillation aims to improve low-magnification whole- slide analysis by transferring high-magnification representations, yet the conditions for useful transfer remain unclear. We develop a decomposition-based analysis of teacher access, representation loss, and model excess, motivating three questions: whether (a) teacher targets help the task, (b) low-magnification students can predict them, and (c) slide models benefit from those predictions. We investigate them through controlled experiments across ten pathology cohorts spanning classifi- cation, grading, and survival prediction. In the main comparison, providing teacher regional means alongside native low-magnification features improves downstream performance in all ten cohorts. Direct prediction achieves lower reconstruction error than residual prediction, yet the predicted features underrepresent variation in the teacher targets. Moreover, better reconstruction does not consistently improve downstream scores, and retaining native features changes performance even when the predicted teacher features are held fixed. Together, these findings expose a gap between reconstructing teacher representations and realizing their downstream value. They challenge the sufficiency of reconstruction error as a measure of cross-resolution transfer and provide a diagnostic framework for examining where that transfer breaks down. Future distillation designs must account for both what students can predict and how slide models use those predictions.

Problem

• Cross-resolution knowledge distillation
• Whole-slide imaging
• Representation transferability
• Computational pathology

Innovation

• Cross-resolution knowledge distillation
• Whole-slide imaging
• Decomposition-based analysis framework
• Representation reconstruction
• Computational pathology