Adversarial artificial intelligence in radiology: Attacks, defenses, and future considerations DOI Creative Commons

Nicholas Dietrich,

Bo Gong, Michael N. Patlas

et al.

Diagnostic and Interventional Imaging, Journal Year: 2025, Volume and Issue: unknown

Published: May 1, 2025

Artificial intelligence (AI) is rapidly transforming radiology, with applications spanning disease detection, lesion segmentation, workflow optimization, and report generation. As these tools become more integrated into clinical practice, new concerns have emerged regarding their vulnerability to adversarial attacks. This review provides an in-depth overview of AI in a topic growing relevance both research domains. It begins by outlining the foundational concepts model characteristics that make machine learning systems particularly susceptible manipulation. A structured taxonomy attack types presented, including distinctions based on attacker knowledge, goals, timing, computational frequency. The implications attacks are then examined across key radiology tasks, literature highlighting risks classification, image segmentation reconstruction, Potential downstream consequences such as patient harm, operational disruption, loss trust discussed. Current mitigation strategies reviewed, input-level defenses, training modifications, certified robustness approaches. In parallel, role broader lifecycle safeguard considered. By consolidating current knowledge technical domains, this helps identify gaps, inform future priorities, guide development robust, trustworthy radiology.

Language: Английский

Adversarial artificial intelligence in radiology: Attacks, defenses, and future considerations DOI Creative Commons

Nicholas Dietrich,

Bo Gong, Michael N. Patlas

et al.

Diagnostic and Interventional Imaging, Journal Year: 2025, Volume and Issue: unknown

Published: May 1, 2025

Artificial intelligence (AI) is rapidly transforming radiology, with applications spanning disease detection, lesion segmentation, workflow optimization, and report generation. As these tools become more integrated into clinical practice, new concerns have emerged regarding their vulnerability to adversarial attacks. This review provides an in-depth overview of AI in a topic growing relevance both research domains. It begins by outlining the foundational concepts model characteristics that make machine learning systems particularly susceptible manipulation. A structured taxonomy attack types presented, including distinctions based on attacker knowledge, goals, timing, computational frequency. The implications attacks are then examined across key radiology tasks, literature highlighting risks classification, image segmentation reconstruction, Potential downstream consequences such as patient harm, operational disruption, loss trust discussed. Current mitigation strategies reviewed, input-level defenses, training modifications, certified robustness approaches. In parallel, role broader lifecycle safeguard considered. By consolidating current knowledge technical domains, this helps identify gaps, inform future priorities, guide development robust, trustworthy radiology.

Language: Английский

Citations

0