Researchers at University College London have developed a new artificial intelligence (AI) tool that can help interpret and assess how well treatments are working for patients with multiple sclerosis (MS). The tool, called MindGlide, can extract key information from existing brain MRI scan images acquired during the care of MS patients, such as measuring damaged areas of the brain and highlighting subtle changes such as brain shrinkage and plaques.
AI uses mathematical models to train computers using massive amounts of data to learn and solve problems in ways that can seem human, including performing complex tasks like image recognition. The MindGlide model was trained on more than 4,000 MRI scans from nearly 3,000 MS patients, and then externally validated on nearly 15,000 scans from another 1,000 patients. In tests, MindGlide was found to perform better at some types of assessment than two other state-of-the-art AI tools, SAMSEG and WMH-SynthSeg, when compared with expert clinical evaluation.
Philipp Goebl, at UCL Queen Square Institute of Neurology and UCL Hawkes Institute, explained, “Using MindGlide will enable us to use existing brain images in hospital archives to better understand multiple sclerosis and how treatment affects the brain. We hope that the tool will unlock valuable information from millions of untapped brain images that were previously difficult or impossible to understand, immediately leading to valuable insights into multiple sclerosis for researchers and, in the near future, to better understand a patient’s condition through AI in the clinic. We hope this will be possible in the next five to 10 years.”
Goebl is first author of the team’s published paper in Nature Communications, titled, “Enabling new insights from old scans by repurposing clinical MRI archives for multiple sclerosis research,” in which they concluded, “MindGlide uniquely enables quantitative analysis of archival single-contrast MRIs, unlocking insights from untapped hospital datasets.”
MS is a condition where the immune system attacks the brain and spinal cord. This causes problems in how a person moves, feels, or thinks. In the U.K., 130,000 people live with MS. As the authors further wrote, “Multiple sclerosis (MS) is a chronic disabling disease affecting over 2.8 million people worldwide, with a disproportionate impact on young populations.”
MRI markers are crucial for studying and testing treatments for MS. However, measuring these markers needs different types of specialized MRI scans, limiting the effectiveness of many routine hospital scans. “Magnetic resonance imaging (MRI) biomarkers are vital for multiple sclerosis (MS) clinical research and trials but quantifying them requires multi-contrast protocols and limits the use of abundant single-contrast hospital archives,” the team continued.
Recent studies have demonstrated the feasibility of deep-learning-based quantification of thalamic and lesion volumes in routine clinical MRI scans for MS patients, the investigators continued. However, such studies didn’t assess treatment effects or include the wide variety of clinical-grade two-dimensional scans of different contrasts (T1-weighted, T2-weighted, and T2-FLAIR contrasts) and scans from clinical archives. “This highlights the urgent need for solutions that extract MRI biomarkers, including lesion load and changes in brain volume, from the varying scans acquired in routine care for research repurposing and potential future clinical applications.”
The UCL-led researchers developed the deep learning model, MindGlide, to address limitations of existing approaches and extract brain region and white matter lesion volumes from any single MRI contrast, to identify damage and changes caused by MS. “We developed MindGlide, a 3D convolutional neural network (CNN), to segment specific brain structures and white matter lesions,” they wrote. They aimed to rapidly, and with no need for pre-processing by the user, quantify brain structures and lesions from single MRI contrast images, detect brain volume changes resulting from treatment effects by analyzing MRI scans that wouldn’t typically be used for these purposes, and also demonstrate the potential of routine MRI scans to detect new lesions and subtle loss of brain tissue even when ideal imaging contrasts aren’t available.
In developing MindGlide the scientists used an initial dataset of 4,247 brain MRI scans from 2,934 MS patients across 592 MRI scanners. During this process, MindGlide trained itself to identify markers of the disease. Through the newly reported study, the authors tested the effectiveness of MindGlide, against three separate databases of 14,952 images from 1,001 patients. This task had previously required expert neuro-radiologists to interpret years of complex scans manually—and the turnaround time for reporting these images is often weeks. “To test MindGlide’s generalizability across ages 14–64, we employed an external validation set of two progressive MS clinical trials, and a real-world cohort of pediatric relapsing-remitting MS patients,” they explained.
MindGlide was able to use AI to detect how different treatments affected disease progression in clinical trials and in routine care, using images that could not previously be analyzed and routine MRI scan images. The process took just five to 10 seconds per image.
MindGlide also performed better than two other AI tools—SAMSEG (a tool used to identify and outline different parts of the brain in MRI scans) and WMH-SynthSeg (a tool that detects and measures bright spots seen on certain MRI scans, that can be important for diagnosing and monitoring conditions like MS)—when compared to expert clinical analysis. “MindGlide demonstrates superior performance in multiple key areas compared to state-of-the-art,” the team noted. MindGlide was 60% better than SAMSEG and 20% better than WMH-SynthSeg for locating brain plaque (or lesion) abnormalities, or for monitoring treatment effects.
The results from the study show that it is possible to use MindGlide to accurately identify and measure important brain tissues and lesions even with limited MRI data and single types of scans that aren’t usually used for this purpose—such as T2-weighted MRI without FLAIR (a type of scan that highlights fluids in the body but still contains bright signals, making it harder to see plaques). “Our results demonstrate that clinically meaningful tissue segmentation and lesion quantification are achievable even with limited MRI data and single contrasts not typically used for these tasks (e.g., T2-weighted MRI without FLAIR),” they stated. “Importantly, our findings generalized across datasets and MRI contrasts. Our training used only FLAIR and T1 images, yet the model successfully processed new contrasts (like PD and T2) from different scanners and periods encountered during external validation.”
As well as performing better at detecting changes in the brain’s outer layer, MindGlide also performed well in deeper brain areas. The findings were valid and reliable both at one point in time and over longer periods (i.e., at annual scans attended by patients). Additionally, MindGlide was able to corroborate previous high-quality research regarding which treatments were most effective. “In clinical trials, MindGlide detected treatment effects on T2-lesion accrual and cortical and deep grey matter volume loss. In routine-care data, T2-lesion volume increased with moderate-efficacy treatment but remained stable with high-efficacy treatment,” the investigators wrote in summary.
The researchers now hope that MindGlide can be used to evaluate MS treatments in real-world settings, overcoming previous limitations of relying solely on high-quality clinical trial data, which often did not capture the full diversity of people with MS.
Principle investigator Arman Eshaghi MD, PhD, at UCL Queen Square Institute of Neurology and UCL Hawkes Institute, and lead of the MS-PINPOINT group, commented, “We were not previously analyzing the bulk of clinical brain images due to their lower quality. AI will unlock the untapped potential of the treasure trove of hospital information to provide unprecedented insights into MS progression and how treatments work and affect the brain.”
The authors noted limitations of their study, and commented that the current implementation of MindGlide is limited to brain scans and does not include spinal cord imaging, which is important for assessing disability in people with MS. They suggest that future research will need to develop a more comprehensive assessment of the whole neural system to encompass both the brain and the spinal cord. “Spinal cord MRI is also widely available in routine care settings and is strongly associated with disability. Future work should expand our approach to the entire central nervous system.”
The post New AI Tool Aids Multiple Sclerosis Research and Therapy Monitoring by Analyzing Existing MRI Scans appeared first on GEN - Genetic Engineering and Biotechnology News.
AI uses mathematical models to train computers using massive amounts of data to learn and solve problems in ways that can seem human, including performing complex tasks like image recognition. The MindGlide model was trained on more than 4,000 MRI scans from nearly 3,000 MS patients, and then externally validated on nearly 15,000 scans from another 1,000 patients. In tests, MindGlide was found to perform better at some types of assessment than two other state-of-the-art AI tools, SAMSEG and WMH-SynthSeg, when compared with expert clinical evaluation.
Philipp Goebl, at UCL Queen Square Institute of Neurology and UCL Hawkes Institute, explained, “Using MindGlide will enable us to use existing brain images in hospital archives to better understand multiple sclerosis and how treatment affects the brain. We hope that the tool will unlock valuable information from millions of untapped brain images that were previously difficult or impossible to understand, immediately leading to valuable insights into multiple sclerosis for researchers and, in the near future, to better understand a patient’s condition through AI in the clinic. We hope this will be possible in the next five to 10 years.”
Goebl is first author of the team’s published paper in Nature Communications, titled, “Enabling new insights from old scans by repurposing clinical MRI archives for multiple sclerosis research,” in which they concluded, “MindGlide uniquely enables quantitative analysis of archival single-contrast MRIs, unlocking insights from untapped hospital datasets.”
MS is a condition where the immune system attacks the brain and spinal cord. This causes problems in how a person moves, feels, or thinks. In the U.K., 130,000 people live with MS. As the authors further wrote, “Multiple sclerosis (MS) is a chronic disabling disease affecting over 2.8 million people worldwide, with a disproportionate impact on young populations.”
MRI markers are crucial for studying and testing treatments for MS. However, measuring these markers needs different types of specialized MRI scans, limiting the effectiveness of many routine hospital scans. “Magnetic resonance imaging (MRI) biomarkers are vital for multiple sclerosis (MS) clinical research and trials but quantifying them requires multi-contrast protocols and limits the use of abundant single-contrast hospital archives,” the team continued.
Recent studies have demonstrated the feasibility of deep-learning-based quantification of thalamic and lesion volumes in routine clinical MRI scans for MS patients, the investigators continued. However, such studies didn’t assess treatment effects or include the wide variety of clinical-grade two-dimensional scans of different contrasts (T1-weighted, T2-weighted, and T2-FLAIR contrasts) and scans from clinical archives. “This highlights the urgent need for solutions that extract MRI biomarkers, including lesion load and changes in brain volume, from the varying scans acquired in routine care for research repurposing and potential future clinical applications.”
The UCL-led researchers developed the deep learning model, MindGlide, to address limitations of existing approaches and extract brain region and white matter lesion volumes from any single MRI contrast, to identify damage and changes caused by MS. “We developed MindGlide, a 3D convolutional neural network (CNN), to segment specific brain structures and white matter lesions,” they wrote. They aimed to rapidly, and with no need for pre-processing by the user, quantify brain structures and lesions from single MRI contrast images, detect brain volume changes resulting from treatment effects by analyzing MRI scans that wouldn’t typically be used for these purposes, and also demonstrate the potential of routine MRI scans to detect new lesions and subtle loss of brain tissue even when ideal imaging contrasts aren’t available.
In developing MindGlide the scientists used an initial dataset of 4,247 brain MRI scans from 2,934 MS patients across 592 MRI scanners. During this process, MindGlide trained itself to identify markers of the disease. Through the newly reported study, the authors tested the effectiveness of MindGlide, against three separate databases of 14,952 images from 1,001 patients. This task had previously required expert neuro-radiologists to interpret years of complex scans manually—and the turnaround time for reporting these images is often weeks. “To test MindGlide’s generalizability across ages 14–64, we employed an external validation set of two progressive MS clinical trials, and a real-world cohort of pediatric relapsing-remitting MS patients,” they explained.
MindGlide was able to use AI to detect how different treatments affected disease progression in clinical trials and in routine care, using images that could not previously be analyzed and routine MRI scan images. The process took just five to 10 seconds per image.
MindGlide also performed better than two other AI tools—SAMSEG (a tool used to identify and outline different parts of the brain in MRI scans) and WMH-SynthSeg (a tool that detects and measures bright spots seen on certain MRI scans, that can be important for diagnosing and monitoring conditions like MS)—when compared to expert clinical analysis. “MindGlide demonstrates superior performance in multiple key areas compared to state-of-the-art,” the team noted. MindGlide was 60% better than SAMSEG and 20% better than WMH-SynthSeg for locating brain plaque (or lesion) abnormalities, or for monitoring treatment effects.
The results from the study show that it is possible to use MindGlide to accurately identify and measure important brain tissues and lesions even with limited MRI data and single types of scans that aren’t usually used for this purpose—such as T2-weighted MRI without FLAIR (a type of scan that highlights fluids in the body but still contains bright signals, making it harder to see plaques). “Our results demonstrate that clinically meaningful tissue segmentation and lesion quantification are achievable even with limited MRI data and single contrasts not typically used for these tasks (e.g., T2-weighted MRI without FLAIR),” they stated. “Importantly, our findings generalized across datasets and MRI contrasts. Our training used only FLAIR and T1 images, yet the model successfully processed new contrasts (like PD and T2) from different scanners and periods encountered during external validation.”
As well as performing better at detecting changes in the brain’s outer layer, MindGlide also performed well in deeper brain areas. The findings were valid and reliable both at one point in time and over longer periods (i.e., at annual scans attended by patients). Additionally, MindGlide was able to corroborate previous high-quality research regarding which treatments were most effective. “In clinical trials, MindGlide detected treatment effects on T2-lesion accrual and cortical and deep grey matter volume loss. In routine-care data, T2-lesion volume increased with moderate-efficacy treatment but remained stable with high-efficacy treatment,” the investigators wrote in summary.
The researchers now hope that MindGlide can be used to evaluate MS treatments in real-world settings, overcoming previous limitations of relying solely on high-quality clinical trial data, which often did not capture the full diversity of people with MS.
Principle investigator Arman Eshaghi MD, PhD, at UCL Queen Square Institute of Neurology and UCL Hawkes Institute, and lead of the MS-PINPOINT group, commented, “We were not previously analyzing the bulk of clinical brain images due to their lower quality. AI will unlock the untapped potential of the treasure trove of hospital information to provide unprecedented insights into MS progression and how treatments work and affect the brain.”
The authors noted limitations of their study, and commented that the current implementation of MindGlide is limited to brain scans and does not include spinal cord imaging, which is important for assessing disability in people with MS. They suggest that future research will need to develop a more comprehensive assessment of the whole neural system to encompass both the brain and the spinal cord. “Spinal cord MRI is also widely available in routine care settings and is strongly associated with disability. Future work should expand our approach to the entire central nervous system.”
The post New AI Tool Aids Multiple Sclerosis Research and Therapy Monitoring by Analyzing Existing MRI Scans appeared first on GEN - Genetic Engineering and Biotechnology News.