Doctors are better at detecting breast cancer when they get help from AI, large study finds
andresr/Getty Images
- A new study found that AI was slightly better at detecting breast cancer than radiologists.
- AI made the same amount of errors as humans and reduced workload by about 44%.
- This could help with burnout in the medical field and lead to quicker cancer diagnoses.
Much of the conversation around AI is (rightfully) ominous, especially when it comes to job replacement.
But an area where AI could positively impact the world is in medical breakthroughs or quicker diagnoses, especially if it can reduce burnout in healthcare professionals or scientists.
A new Swedish study published Tuesday in the Lancet Oncology Journal put that theory to the test and found very promising results. Participants had their mammograms screened by either AI or two human radiologists — and AI was just as efficient (if not slightly more) at identifying signs of breast cancer.
40,000 women were randomly assigned to AI screenings
The study used over 80,000 women around the age of 54. A little over half were assigned to an AI-assisted screening, which involved at least one radiologist interpreting the results with the help of AI.
The other half were assigned to a standard mammogram screening, with one to two radiologists reading the scans themselves.
AI detected slightly higher rates of cancer with few errors
The AI-assisted doctors accurately detected breast cancer in 28% of patients, while the doctors who did not use AI only spotted cancer in 25%. This meant AI assistance led to 41 more cancer diagnoses overall, including invasive and early-stage breast cancers.
Additionally, AI had about the same error rate as the radiologists: The false-positive rate was 1.5% in both groups.
AI reduced radiologists' workload by nearly half
With the help of AI, the study concluded that fewer screen readings were needed and estimated a 44% reduction in workload.
According to a 2022 Medscape survey, radiologists ranked as some of the most burned-out medical professionals. Offloading some of the work to AI can potentially lead to earlier diagnoses of one of the deadliest cancers in the US.