How can Machine Learning advance MRI scans?
Understanding the Benefits of Machine Learning in MRI scans
Applications of machine learning are so common nowadays that people don’t even pay much attention. But did you know machine learning can also help detect health anomalies and conditions, which can help doctors to a great extent? In the past few decades, technology has transformed the healthcare sector. MRI (magnetic resonance image), CT (computed tomography) scan, and X-ray are excellent techniques employed for quick diagnosis of diseases. These methods are better than existing physical tests or objective techniques used to determine what ails us. Further, machine learning is based on data acquisition and monitoring without prior programming and self-training. This implies that machine learning can be developed to facilitate the detection, segmentation, and classification of images and lesions.
MRI offers a wide variety of imaging techniques. This is opposed to functional MR images (fMRI) or positron emission tomography (PET) scans, which image blood flow activity and metabolic activity, respectively. However, it requires a vast data to be created as per examination, which needs to be checked for sufficient quality to derive a meaningful diagnosis. This is generally a manual process and therefore costs much time and money. Another problem is that any imaging artifacts originating from scanner hardware, signal processing, or induced by the patient may reduce the image quality and complicate the diagnosis or image post-processing. Therefore, we need to automate the whole process in a manner that does not compromise the images’ quality. Hence, machine learning is viewed as a potential solution to this expensive procedure.
Recently, Facebook’s AI research team (FAIR) and radiologists at NYU Langone Health collaborated in, the FastMRI initiative project. The scientists trained a machine learning model on pairs of low-resolution and high-resolution MRI scans, using this model to “predict” what final MRI scans look like from just a quarter of the usual input data. They created a check system for the neural network based on the physics of MRI scans. For this process, the data set of 108 MRI scans with different diseases and conditions were analyzed. In each scan, three-fourth of the data was removed and fed to the AI model. It was observed that with the use of this AI-model, the scans that were generated were four times faster than the traditional scans. The image generated with this model was also distinguishable from that of the older scanning models.
In a recent study, researchers at the Indian Institute of Technology (IIT) Roorkee, India, and Kyoto University, Japan, have had designed a machine-learning algorithm to identify the grade of glioma with high accuracy. Glioma is a fatal brain cancer resulting due to the abnormal growth of the glial cells. Another group of scientists from Stanford University Medical Center has developed a deep learning model (MRNet) for detecting general abnormalities and specific diagnoses (anterior cruciate ligament [ACL] tears and meniscal tears) on knee MRI exams.
Researchers at Mass Eye and Ear have developed a unique diagnostic tool that can detect dystonia from MRI scans, the first technology of its kind to provide an objective diagnosis of the disorder. Dystonia is a potentially disabling neurological condition that causes involuntary muscle contractions, leading to abnormal movements and postures. It is often misdiagnosed and can take people up to 10 years to get a correct diagnosis. DystoniaNet utilizes deep learning, a particular AI algorithm, to analyze data from individual MRI and identify subtler differences in brain structure.
Another machine learning group has developed a deep transfer learning computer-aided diagnosis (CADx) methodology to diagnose breast cancer using mpMRI (Multiparametric magnetic resonance imaging).
At the same time, a research team led by Dr. Shinjini Kundu of Carnegie Mellon University and the University of Pittsburgh and Dr. Gustavo Rohde of the University of Virginia investigated whether artificial intelligence could be used to analyze MRI images for early signs of osteoarthritis and predict who will develop the disease. They used MRI scans from 86 people who had no initial symptoms or visual signs of disease. The team achieved 78% accuracy in predicting future osteoarthritis cases. The team used a technique they developed called three-dimensional transport-based morphometry (3D TBM) to identify biochemical changes, such as how much water is present in cartilage using MRI scans. Using 3D TBM, they analyzed the baseline “cartilage maps” of the participants’ knees. After three years, they compared the cartilage maps for the participants who were eventually diagnosed with osteoarthritis with those who were not. The findings were published on September 21, 2020, in the Proceedings of the National Academy of Sciences.