Test-Time Instance Selection for Improved Whole Slide Image Analysis

2026-08-14

AI Summary

This study addresses patch redundancy and the reliance on additional training for instance selection in multiple instance learning models for whole slide image analysis. To overcome these limitations, we propose TTIS, a training-free, plug-and-play framework that dynamically selects representative patches and fuses multi-view features during inference to optimize instance selection without retraining. Experiments demonstrate that TTIS significantly improves or matches existing baselines in classification and subtyping performance across multiple benchmark datasets. By enhancing WSI analysis efficacy at zero training cost, this work establishes a novel test-time adaptation paradigm for efficient pathological diagnosis.

Abstract

Whole Slide Image (WSI) analysis has been widely studied for cancer diagnosis. Conventionally, a gigapixel WSI is divided into small patches and processed by Multiple Instance Learning (MIL) models. However, existing MIL models typically process all patches, many of which contain redundant or non-informative tissue patterns. Although recent approaches have focused on instance selection to identify discriminative patches and reduce redundancy, these selection modules still require additional training. In this work, we propose Test-Time Instance Selection (TTIS), a training-free, plug-and-play framework that selects compact yet representative patches during inference. TTIS further incorporates a multi-view ensemble strategy to integrate distinct facets of tissue morphology, enhancing robustness. Importantly, TTIS can be seamlessly integrated into existing MIL models without retraining or architectural changes, enabling flexible deployment. Extensive evaluations across multiple benchmarks demonstrate that our approach improves or matches baseline MIL performance across a range of classification and subtyping tasks. Our implementation code is available at https://github.com/QuIIL/TTIS

Problem

• Whole Slide Image Analysis
• Multiple Instance Learning
• Instance Selection
• Test-Time Adaptation
• Redundancy Reduction

Innovation

• Test-Time Instance Selection
• Training-free
• Plug-and-play
• Multi-view Ensemble
• Whole Slide Image Analysis