Probabilistic Robustness in Medical Image Classification

2026-07-04

AI Summary

This work addresses the lack of effective reliability evaluation for deep learning models in medical image classification under realistic perturbations, noting that conventional adversarial robustness methods overly emphasize worst-case scenarios and poorly reflect clinical reality. For the first time, probabilistic robustness is introduced to the domain of medical image classification, establishing an evaluation framework tailored to natural perturbations. The authors systematically assess mainstream models on the MedMNIST v2 benchmark using simulated natural image corruptions and probabilistic robustness analysis. Their findings reveal consistent patterns of performance degradation under realistic perturbations, offering a statistically principled and clinically relevant evaluation paradigm that provides empirical grounding for the trustworthy deployment of AI systems in medical imaging.

Abstract

Deep learning (DL) has shown strong performance in medical image classification, but its trustworthy deployment remains challenging in safety-critical clinical settings, where prediction errors under perturbations may lead to severe consequences. Existing studies mainly focus on adversarial robustness (AR) from a worst-case perspective; however, such settings may be less representative of real medical applications. In this work, we investigate probabilistic robustness (PR) as a more practical measure of model trustworthiness. To this end, we construct a set of natural corruption settings for medical image classification and systematically evaluate commonly used DL models on MedMNIST v2 dataset. Our study provides a statistically grounded perspective on assessing the trustworthiness of DL models, thereby supporting their more trustworthy deployment in medical imaging applications.

Problem

• probabilistic robustness
• medical image classification
• deep learning trustworthiness
• natural corruptions
• model reliability

Innovation

• Probabilistic Robustness
• Medical Image Classification
• Natural Corruption
• Trustworthy AI
• Deep Learning