When normalization hallucinates: unseen risks in AI-powered whole slide image processing

2025-12-08

AI Summary

This work identifies a long-overlooked “hallucination” risk in AI-driven whole-slide image (WSI) normalization: models generate visually imperceptible yet structurally spurious artifacts on clinical histopathological images, severely compromising diagnostically critical features. To address the inability of conventional evaluation metrics to detect such latent hallucinations, we propose a novel hallucination quantification metric based on deep feature contrast. We further conduct the first systematic retraining and evaluation of mainstream normalization models on real-world clinical WSI datasets. Results reveal pervasive hallucination across multiple established methods, while traditional metrics—including MSE and SSIM—fail entirely to capture these distortions. Our study establishes a new paradigm for hallucination assessment in WSI normalization and provides both a critical reliability warning and technical foundation for preprocessing pipelines in computational pathology.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Whole slide image (WSI) normalization remains a vital preprocessing step in computational pathology. Increasingly driven by deep learning, these models learn to approximate data distributions from training examples. This often results in outputs that gravitate toward the average, potentially masking diagnostically important features. More critically, they can introduce hallucinated content, artifacts that appear realistic but are not present in the original tissue, posing a serious threat to downstream analysis. These hallucinations are nearly impossible to detect visually, and current evaluation practices often overlook them. In this work, we demonstrate that the risk of hallucinations is real and underappreciated. While many methods perform adequately on public datasets, we observe a concerning frequency of hallucinations when these same models are retrained and evaluated on real-world clinical data. To address this, we propose a novel image comparison measure designed to automatically detect hallucinations in normalized outputs. Using this measure, we systematically evaluate several well-cited normalization methods retrained on real-world data, revealing significant inconsistencies and failures that are not captured by conventional metrics. Our findings underscore the need for more robust, interpretable normalization techniques and stricter validation protocols in clinical deployment.

Problem

• AI normalization in pathology risks hallucinating realistic artifacts
• Current methods fail to detect hallucinations in clinical data
• Need robust techniques and validation to prevent diagnostic errors

Innovation

• Novel image comparison measure detects hallucinations
• Systematic evaluation reveals normalization method failures
• Proposes robust interpretable techniques for clinical deployment