SCHOLARLY PUBLICATION ✓ Open Access

AI for radiographic COVID-19 detection selects shortcuts over signal

Alex J. DeGrave, Joseph D. Janizek, Su‐In Lee

📖 Nature Machine Intelligence → 📅 2021-05-31 🔗 DOI: 10.1038/s42256-021-00338-7

View Journal Profile → ⬇ Download PDF View Official Paper ↗

📄 Abstract

Abstract Artificial intelligence (AI) researchers and radiologists have recently reported AI systems that accurately detect COVID-19 in chest radiographs. However, the robustness of these systems remains unclear. Using state-of-the-art techniques in explainable AI, we demonstrate that recent deep learning systems to detect COVID-19 from chest radiographs rely on confounding factors rather than medical pathology, creating an alarming situation in which the systems appear accurate, but fail when tested in new hospitals.

📄 Abstract

📤 Share this page

🔬 Related Research Papers

Geometry-enhanced molecular representation learning for property prediction

Three types of incremental learning

Augmenting large language models with chemistry tools

Predicting the state of charge and health of batteries using data-driven machine learning

Molecular contrastive learning of representations via graph neural networks

CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling

An interpretable mortality prediction model for COVID-19 patients

Parameter-efficient fine-tuning of large-scale pre-trained language models

Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans

Drug discovery with explainable artificial intelligence

An open source machine learning framework for efficient and transparent systematic reviews

Secure, privacy-preserving and federated machine learning in medical imaging