Artificial intelligence will likely help radiologists diagnose patients more rapidly, but there are still challenges to overcome, according to The Wall Street Journal.
A typical radiologist might review hundreds of X-ray images before detecting a medical concern. However, advanced AI algorithms could be used to identify and analyze images on their own, and flag issues for review by a human radiologist.
Still, AI algorithms are new systems in need of improvement. Here are four key challenges to developing AI for diagnosis, according to The Wall Street Journal.
1. A developer trains an AI system on how to identify medical issues such as tumors or fractures using images of relevant examples. These example images must be labeled by a human being, so that the system identifies what to look for in the future.
2. Healthcare companies often struggle to recruit top data scientists. GE Healthcare is working to address this issue by partnering with colleges to advance science, technology, engineering and math coursework.
3. Some healthcare stakeholders also worry about patient privacy, since an AI system requires access to patient data to reach a diagnosis. In particular, hospital leaders might worry about whether companies accessing their medical records have proper information security protocols in place.
4. Another challenge to deploying AI in healthcare involves perception, as physicians worry the technology might one day replace them. However, health IT companies have critiqued this view of AI.
"What we're developing is a suite of applications to make [people] more effective" at diagnosing and treating patients, Charles Koontz, chief digital officer of GE Healthcare, told The Wall Street Journal.