AI Can Predict the Recurrence of Brain Tumors in Kids, Study Finds
A team of researchers has leveraged a technique called temporal learning to train an AI system to predict the likelihood of brain cancer recurring in kids diagnosed with gliomas. This AI model uses the magnetic resonance images periodically captured after the kids have received treatment for the gliomas.
The study’s findings, based on analyzing multiple images, indicate that the AI model can produce predictions that are 89% accurate when compared to predictions made on the basis of just one image that have an accuracy rate of 50%., no different from flipping a coin.
This approach, which looks for subtle changes in brain structure over time, could reduce how frequently kids who have undergone glioma treatment may need follow-up MR imaging to detect tumor recurrence. It could also allow treatment teams to initiate treatment protocols early for those kids found to be likely to have a tumor recurrence.
Gliomas can be treated, but they can recur and the chances of recurrence vary from one patient to another. This is the reason why periodic MRIs are needed to identify which patients have suffered a recurrence and need additional treatment. This regular imaging can be costly and stressful for patients and their caregivers, which is why researchers are constantly looking for better ways to predict who could be at risk of suffering a recurrence of the tumors after the initial treatment.
The new model could help with this since it needs only 4-6 images to make an accurate prediction. This could reduce how often low-risk patients have to undergo follow-up imaging and it can flag those at risk for extra monitoring and early retreatment.
A joint team of researchers at Mass General Brigham, Boston Cancer & Blood Disorder Center and Boston Children’s Hospital conducted the study. They collected almost 4,000 MRI scans of 715 kids diagnosed with gliomas. They leveraged temporal learning to train their AI model to track subtle changes within the brain to predict the likelihood of glioma recurrence.
This approach hadn’t previously been deployed in studying the use of AI in medical imaging. The images used were captured periodically for several months after a pediatric patient received treatment.
The data from the study shows that the model could provide a recurrence prediction 12 months before it actually happened, which was superior to the AI models that relied on only one MRI image.
The researchers say this model still requires further validation and clinical study before its efficacy can be confirmed.
Its success could enable patients to start treatment early before recurrent gliomas progress. Catching the recurrence early would also increase the success odds of brain tumor treatments that companies like CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are working to bring to market.
NOTE TO INVESTORS: The latest news and updates relating to CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are available in the company’s newsroom at https://ibn.fm/CNSP
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
A team of researchers has leveraged a technique called temporal learning to train an AI system to predict the likelihood of brain cancer recurring in kids diagnosed with gliomas. This AI model uses the magnetic resonance images periodically captured after the kids have received treatment for the gliomas.
The study’s findings, based on analyzing multiple images, indicate that the AI model can produce predictions that are 89% accurate when compared to predictions made on the basis of just one image that have an accuracy rate of 50%., no different from flipping a coin.
This approach, which looks for subtle changes in brain structure over time, could reduce how frequently kids who have undergone glioma treatment may need follow-up MR imaging to detect tumor recurrence. It could also allow treatment teams to initiate treatment protocols early for those kids found to be likely to have a tumor recurrence.
Gliomas can be treated, but they can recur and the chances of recurrence vary from one patient to another. This is the reason why periodic MRIs are needed to identify which patients have suffered a recurrence and need additional treatment. This regular imaging can be costly and stressful for patients and their caregivers, which is why researchers are constantly looking for better ways to predict who could be at risk of suffering a recurrence of the tumors after the initial treatment.
The new model could help with this since it needs only 4-6 images to make an accurate prediction. This could reduce how often low-risk patients have to undergo follow-up imaging and it can flag those at risk for extra monitoring and early retreatment.
A joint team of researchers at Mass General Brigham, Boston Cancer & Blood Disorder Center and Boston Children’s Hospital conducted the study. They collected almost 4,000 MRI scans of 715 kids diagnosed with gliomas. They leveraged temporal learning to train their AI model to track subtle changes within the brain to predict the likelihood of glioma recurrence.
This approach hadn’t previously been deployed in studying the use of AI in medical imaging. The images used were captured periodically for several months after a pediatric patient received treatment.
The data from the study shows that the model could provide a recurrence prediction 12 months before it actually happened, which was superior to the AI models that relied on only one MRI image.
The researchers say this model still requires further validation and clinical study before its efficacy can be confirmed.
Its success could enable patients to start treatment early before recurrent gliomas progress. Catching the recurrence early would also increase the success odds of brain tumor treatments that companies like CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are working to bring to market.
NOTE TO INVESTORS: The latest news and updates relating to CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are available in the company’s newsroom at https://ibn.fm/CNSP
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
