Artificial Intelligence has repeatedly proven its worth in evaluating cardiac MRI images – how could this help the doctor, the patient and the researcher?
AI enabled heart failure gene studies
- Certain gene mutations have long been known to cause non-ischemic heart failure, but a lot more genes could also play a role. UK researchers set out to find out genes linked to heart failure. Their databank was big. At 17,000 DNA samples and cardiac MRI scans, one could say – too big!
- With the help of a deep learning algorithm, they only had to review 2,600 scans manually – a 75% decrease in workload!
- The neural network performed admirably, as it seemed to agree with human reviewers at a rate of 88-98%.
Lightning-fast AI reviews cardiac MRI like an expert would
- Another team of British researchers trained a neural network to analyze cardiac parameters over 600 MRI scans.
- It was then pitted against experts. The challenge was to measure differences between scans of patients (scan-rescan).
- The experts performed slightly better, but it took them 13 minutes per study – the AI only looked at each study for only 4 seconds!
Are radiologists going the way of the dodo?
- Most definitely not! At this moment, all results coming from neural networks must be reviewed by clinicians.
- The ability to automate previously lengthy and tedious parts of an MRI study yields multiple benefits.
- Doctors can invest their new-found time in giving more attention to challenging cases.
- Research becomes easier as interobserver variability is nullified when an AI is doing the observation.
- Patients enjoy better and faster healthcare. And who knows? One day, it might take less than a week to get the MRI’s results back!
Leftovers
- Each MRI had to be carefully examined, in order to quantify each heart’s dimensions and left ventricular ejection fraction.
- Thanks to a deep learning algorithm, the authors had to manually review only 4,200 scans.
- These may sound like a lot, but that amounts to a
- Heart failure is a major problem worldwide – for both patients and healthcare systems.
- A team of researchers from the UK set out to find out just that – with a publicly available database of almost 17,000 MRI cardiac images and DNA samples from the UK Biobank.