Dec 14 2021: Algorithms that detect cancer can be fooled by hacked images

Back in Dec 14 2021: Algorithms that detect cancer can be fooled by hacked images

Cyberattacks can trick artificial intelligence (AI) programs that analyze medical images for cancer, potentially leading to incorrect diagnoses. Researchers demonstrated that a program could manipulate mammograms to either add or remove evidence of cancer, successfully fooling an AI tool and human radiologists.

Although such hacks have not yet occurred in real-world scenarios, the study highlights the need for healthcare organizations to prepare for this threat. Hackers are increasingly targeting healthcare institutions, and the potential for direct attacks on people's health is a growing concern. Training AI models with manipulated images can help detect and prevent these types of attacks.

Fast forward to Aug 10 2023: Hospital bosses love AI. Doctors and nurses are worried.

Mount Sinai Hospital is using artificial intelligence (AI) to improve patient care, with AI software generating scores that help doctors assess patient risk. AI has the potential to transform healthcare, as demonstrated by recent studies showing its ability to detect more breast cancer cases than radiologists.

Mount Sinai is investing heavily in AI software and education, using AI specialists to develop medical tools. However, there are concerns among healthcare workers about AI making wrong diagnoses, compromising patient privacy, and leading to staff cuts. While AI has its benefits, it cannot replace the work of human doctors and nurses.

What are the potential costs and concerns associated with the use of AI in healthcare?