Where Is MRI Technology Heading in the Next Five Years?
In recent years, magnetic resonance imaging (MRI) has undergone a remarkable transformation. Once primarily a qualitative tool, MRI is now increasingly being used as a quantitative resource. Dr. Tim highlighted this shift, explaining that while MRI was initially developed to replace CT scans and X-rays, it has carved out its own niche as a unique imaging tool. Radiologists were traditionally trained to interpret MRI images qualitatively, but advancements in the field are unlocking its rich physics for quantitative analysis.
For example, MRI can now measure the rate of water molecule diffusion in the body, providing insights into cellular behaviour. Functional MRI (fMRI) has also emerged, enabling researchers to map brain activity by observing which areas light up during specific mental tasks. According to Dr. Tim, the quantitative side of MRI holds the most promise. By delving deeper into the wealth of physical information available, we stand to gain significant insights into various diseases. The pace of innovation in this field shows no signs of slowing, and the next five years promise even more groundbreaking developments.
AI and MRI: How Are Neural Networks Trained?
Artificial intelligence (AI) has become a key driver in advancing MRI capabilities. Neural networks, the foundation of AI models, mimic the structure of the human brain, learning by processing vast amounts of data. But how are these models trained to interpret complex MRI images?
Developed by computer scientists, these networks are inspired by the way the human brain works, using structures similar to synapses and neurons. Essentially, neural networks process information by taking in data and producing an output. What makes them fascinating—and somewhat mysterious—is how they learn from the examples we provide.
Dr. Tim explained the process of training and testing neural networks using the development of FerriSmart® as an example. FerriSmart is a software medical device that automatically evaluates liver iron concnetration from MR images. The training begins with feeding the AI tens of thousands of MRI datasets. Each dataset was acquired in a highly specific way, quality-checked by humans, and analysed manually using specialised software to produce results. These datasets were used to train the neural network by showing it examples and teaching it the correct answers. For instance, an MRI dataset would be provided along with the corresponding liver iron concentration. As the neural network processes each dataset, it adjusts its internal parameters—often called weights or biases—to improve its understanding.
Once the network has been exposed to enough data, the training shifts to a different phase. Instead of feeding it more known cases, new, unseen data is introduced to test whether it can accurately generate results—like predicting liver iron concentration. Remarkably, it works, though we don’t fully understand how it learns. It’s akin to teaching a six-year-old to catch a cricket ball: through training and practice, they eventually get it, even if the neurological processes behind it remain a mystery.
To ensure the model performs reliably, it is tested extensively on additional cases. For example, in the case of FerriSmart®,over 1,200 cases were evaluated to verify its accuracy. While the model generally performs well, it does struggle with certain groups, such as infants under one year old, likely due to limited training data for this age group. This highlights the importance of understanding the limitations of AI models and being cautious about their application, particularly in areas where the training dataset has limitations. Used judiciously, however, this approach can yield highly reliable results for most patients.
Affordability and Accessibility in Low- and Middle-Income Countries
One of the most pressing questions about MRI technology is its cost and feasibility for deployment in low- and middle-income countries. Dr. Tim emphasised that this was a key motivation for developing AI-driven analysis methods. By reducing the cost of data analysis, this technology becomes more accessible to countries with limited resources.
While MRI scanners themselves are a significant investment, often costing millions of dollars, they are typically already in use in hospitals for other purposes. The additional use of these scanners for advanced MRI analysis represents a cost-effective way to maximise their value. Countries like Lebanon, Turkey, and Greece have already leveraged this technology, demonstrating its potential for broader adoption in similar regions.
An Interesting Aside – Magnetics and the Human Brain!
One area of research that directly explores magnetics and the brain is Transcranial Magnetic Stimulation (TMS). TMS uses magnetic pulses to create tiny electric currents in the brain. These currents are believed to interact with the brain’s natural electrical activity, offering potential for therapeutic applications.
TMS differs significantly from Magnetic Resonance Imaging (MRI). While MRI involves mostly static magnetic fields with slight changes during the scan, TMS uses very rapidly changing magnetic fields, which are far more dynamic. These rapid changes generate electric currents in targeted areas of the brain, potentially stimulating or altering brain function.
TMS is being studied for its potential in treating conditions like epilepsy, post-traumatic stress disorder (PTSD), and other neurological or psychiatric disorders. While the exact mechanisms and applications are still under exploration, the technology holds promise for advancing our understanding of the brain and developing novel therapies.
So, do humans have the ability to sense magnetic fields?
Many animals, like homing pigeons and certain bacteria, are known to navigate using Earth’s magnetic field. However, studies on humans have yielded mixed results. Although some research suggests humans might have had this ability in the distant past, the consensus is that we don’t possess this capability today.
The connection between magnetics and the brain remains a rich area of exploration. From understanding how magnetic fields might influence our neurological processes to investigating potential therapeutic applications, researchers continue to uncover new insights.
Conclusion
From the transition to quantitative MRI to the integration of AI and its growing accessibility worldwide, it’s clear that this technology is on the cusp of even greater innovation. Whether it’s unravelling the mysteries of the brain or making cutting-edge diagnostics more affordable, the future of MRI holds immense promise. As Dr. Tim’s insights show, we’re only beginning to scratch the surface of what’s possible with this groundbreaking technology.
Contact us to know more here – https://www.resonancehealth.com/contact/
About Professor Tim St Pierre PhD, Director, Scientific Liaison
Professor St Pierre led the team which developed the original FerriScan®, FerriSmart®, and HepaFat-Scan® technologies at Resonance Health. He has collaborated with many international experts in the field of iron overload and fatty liver disease.
Prof. St Pierre has published over 200 peer reviewed research papers and regularly participates in international research collaborations. In 2010, he won the Clunies Ross Award from the Australian Academy of Technological Sciences and Engineering, in 2015 the Ernst and Young Western Region Entrepreneur of the Year Award and in 2016 the Panos Englezos Prize from the Thalassaemia International Federation for his contribution to the improvement of health for patients with thalassaemia. He currently holds an honorary Senior Research Fellowship at The University of Western Australia and is our Director of Scientific Liaison.
RHAS website – https://www.resonancehealth.com/board-leadership/
LinkedIn – https://www.linkedin.com/in/tim-st-pierre-48041030/