Magnetic resonance imaging, or MRI, has been one of the most transformative medical imaging tools in the last few decades. While powerful and insightful, MRI scans are a slow technology that can frequently cause patient discomfort. A typical scan lasting 40 to 60 minutes can be overwhelming for a patient who is already in pain.
Researchers recently investigated how AI techniques could speed up MRI scans. They discovered that AI techniques are at least as reliable as existing non-technical methods and outperform them when clarifying small details in an image.
There are numerous practical advantages to using AI for MRI reconstruction:
Patients can have much faster imaging procedures.
The images are less likely to have artifacts due to patient motion.
Hospitals can accommodate more patients with shorter wait times.
Radiologists can still make accurate diagnoses for their patients.
The team compared the most promising two AI techniques — trained and untrained neural networks — to non-AI-based image reconstruction methods currently in clinical use to assess AI’s potential. Trained networks have been studied extensively in recent years, and they rely on high-quality examples to train against in a supervised manner. Untrained networks, on the other hand, represent cutting-edge advances in unsupervised AI that do not require any training data at all.
Related Content: Miniaturizing Medical Imaging, Sensing Technology