A new study has found that artificial intelligence is on par with human medical experts when it comes to making medical diagnoses based on images.
The potential for AI in healthcare could free up time for doctor-patient interactions and tailored treatments.
The study appeared in the Lancet Digital Health and was a comprehensive review of published studies on the issues. The authors caution those who are taken in by the hype surrounding AI and medicine, saying that although the results are encouraging they only show that at best, AI is equivalent to humans.
Meanwhile, other medical experts have warned that these latest findings are only based on a small number of studies and that there is a lot of poor-quality research littering the field.
Pooling Promising Data
The research team from the University Hospitals Birmingham NHS foundation trust discovered more than 20,000 relevant studies but only 14 of them reported good quality data.
They pooled the most promising data from those 14 studies which revealed that deep learning systems correctly detected a disease state 87% of the time (compared with 86% for healthcare professionals) and correctly gave the all-clear 93% of the time. This was compared to 91% for human experts.
Although this isn’t a great deal of data and additional high-quality studies will need to be carried out, the study’s authors contend that AI could act as a diagnostic tool to help clear backlogs of scans and images. It could also be useful in medical centers where there are shortages of experts to interpret images.
To read the latest study, click here.