AI can help make MRI scans more reliable. Thanks to this technology, the scan can be read much faster than with the human eye.
Most errors made in MRI scans are still human errors. Artificial intelligence can change this just like that, according to a study. AI can also read scans much faster than humans. Analysing a heart function from an MRI scan, for example, takes a human doctor about 13 minutes. With AI, the same scan can be analysed in as little as four seconds.
"Cardiovascular MRI offers unparalleled image quality for assessing the structure and function of a heart, but many manual analyses are very standard and outdated," said Charlotte Manisty, author of the study. "Automated machine learning techniques offer the potential to change this and dramatically improve efficiency. We therefore look forward to further research that can substantiate that AI is superior to human analytics.
Precision
In the UK, where the study was carried out, about 150,000 cardiovascular MRI scans are made each year. The researchers believe that with the use of AI, it is possible to save 54 days of doctors' deployment in the health centres where the scans are made. In addition, AI could eliminate human errors.
"Our research shows that automated techniques are at least as good as humans and have the potential to get even better. This is going to provide a lot of precision in clinical measurements," Manisty said.
For their research, the scientists taught a machine learning neural network AI to read the MRI scans of nearly six hundred patients. They then compared the AI's precision with that of an expert and trainee of 110 different patients from different centres. Despite AI analysing many more scans, the precision was almost the same.
Source: Circulation: Cardiovascular Imaging.
