Artificial Intelligence (AI) is in the momentum of its application options. The machine learning algorithms can be utilized in a myriad of different industries and improve processes that have been working in the same way for the past decades. One of the uses of AI that has developed abruptly is MRI technology. Magnetic resonance imaging (MRI) is a leading technology for assessing and diagnosing musculoskeletal disorders or injuries. Having said that, there are several drawbacks and challenges that the medical field has to face. Issues with undersampled MRI data, low resolution, and the lengthy process of an MRI scan are only a few. Thus, Medtech has been focusing on improving the technique for years. It is only in the last three years that tremendous progress has been made in combining the strength of MRI and AI in order to make the process more efficient.
AI Improving The Signal-To-Noise Ratio In MRI Scans
A Harvard/MIT research released recently showed that machine learning shows great opportunities regarding improving MRI scans, having less signal-to-noise ratio than the traditional MRI. This is consistent with the 2020 study on improving the MRI speed with AI showed that in the past 3 years, the field has progressed amazingly in past 3 years, not only improving the resolution of the data and allowing for reconstruction of undersampled MRI data.
Unbelievable MedTech Progress In The Past 3 Years
It can also significantly improve the speed which leads to more patients being scanned in one day. Doctors continue to mostly rely on MRI scans and other bioimaging technology to examine the inside of one’s body and detect any abnormal tissue. Thus, utilizing technology to prevent noise or artifacts from MRI scans is crucial.
AUTOMAP Upgrading The MRI Scans With AI
The researchers developed an automated reconstruction process, AUTOMAP, and trained the neural network using 50 thousand MRI brain images. Subsequently, they tested how accurately the AUTOMAP system could reconstruct data using a real-world MRI machine. The results were encouraging and showed that AUTOMAP creates higher-quality images, with a signal-to-noise ratio of 21.6 (an improvement from the traditional 17.6). It was also faster than the manual tweaking that is now done by MRI experts.
The Instantaneous Image Reconstruction
“Since AUTOMAP is implemented as a feedforward neural network, the speed of image reconstruction is almost instantaneous—just tens of milliseconds,” Dr. Matthew S. Rosen of Massachusetts General Hospital who led the research said. “Some types of scans currently require time-consuming computational processing to reconstruct the images. In those cases, immediate feedback is not available during initial imaging, and a repeat study may be required to better identify a suspected abnormality. AUTOMAP would provide instant image reconstruction to inform the decision-making process during scanning and could prevent the need for additional visits.”
Facebook AI Research Cuts MRI Scan Time In Half
Another ongoing research, conducted by the NYU Langone Health in partnership with the Facebook Artificial Intelligence Research, showed that with the help of AI, an MRI machine can spend only half of the current time on taking the scan (thus, collecting less data than a normal image) but still produce the images of equal quality.
Possibility For Faster Diagnosis
That way, patients could spend much less time in the MRI tube and hospitals could diagnose many more patients. The researchers working on the research reported that the radiologists that were given the images done with the conventional MRI and the ones done in a shorter time using AI could not see the difference. The clinical study that was published in the American Journal of Roentgenology in January of 2021 showed that the fast MRI assisted by AI would produce diagnostically interchangeable images using 75% less raw data.
“There is no difference in how people read the scans, whether they’re reading the accelerated or the clinical [traditional] sequences,” Dr. Michael Recht, the chair of the radiology department at NYU Langone Health says. “They’re able to make the diagnosis equally well on either of the scans.”
“The sequences really are interchangeable, and I’m very, very comfortable using those sequences to make a diagnosis,” he says. “Of the six radiologists in the study, only one of them was able to discern whether the scans were made the normal way or with AI.”
Canon Medical Expanding Their AI Programs To Improve MRI Image Quality
Canon Medical announced today that they are expanding their research on AI programs aiming at improving the MRI image quality. After its initial research, the company says that the artificial intelligence programs can be used in more than 95% of scanning procedures.
The Advanced intelligence Clear-IQ Engine (AiCE) sharpens the scans significantly and while it was already approved by FDA for some brain and knee scans, it can now cover all joints, as well as the scans focusing on the pelvic and abdomen areas, as well as spine and cardiac scans.
Making MRI Images Easy To Acquire
“In today’s environment, making images easy to read and acquire is more important than ever.” Jonathan Furuyama, the managing director of the MR unit of Canon Medical said. “This is the latest demonstration of our commitment to offering accessible AI that clinicians can use to make the greatest impact on patient care.”
Using AI Classifiers Combined With MRI To Diagnose AD
AI classifiers combined with MRI was also used for the diagnosis of Alzheimer’s Disease and separated patients suffering from mild AD and patients who needed treatment. The diagnosis is made possible by using the structural T1-weighted AI and an AI classifier that is based on some extremely specific neuropsychological measures.
The method’s success rate is currently estimated at more than 80% accuracy and 87% specificity. That is much higher than the success rate of MRI data alone that has approximately 72% accuracy and 75% specificity. The combination of AI and MRI can lead to a diagnosis provided up to 24 months earlier, allowing for a much faster reaction.
To find out more about AI and how it improves numerous fields, check out: