Summary: A new artificial intelligence algorithm analyses MRI data to accurately predict the age of a brain.
Source: King’s College London
Researchers from the School of Biomedical Engineering & Imaging Sciences have developed a new machine learning tool that analyses brain MRIs and predicts the age of a brain compared to the rest of the population. Essentially a screening tool, it automatically detects older-appearing brains in real-time using routine clinical scans.
Published in Neuroimage, the research shows that as part of a natural process, brains lose volume with age and as long as the volume loss is appropriate for the patient’s age, the new tool will predict the correct age of the patient.
But if a patient has a brain which is diseased and has lost a disproportionate amount of volume, such as in dementia, the tool will show the mismatch between the real age and the predicted age thereby alerting clinicians to this important discrepancy and flag that the brain is abnormal for age.
“We have shown it is possible to close that gap from point of scan to expert review, if the center is lucky enough to have experts, by automating that process.”
Using a deep learning based neuroradiology report classifier, the researchers generated a dataset of 23, 302 ‘radiologically normal for age’ head MRI examinations from two large UK hospitals, namely, Guy’s and St Thomas’ NHS Foundation Trust and King’s College Hospital using the pre-existing neuroradiology reports.
Then using an unusual approach whereby there is very little computational pre-processing of the scans, they applied another deep learning image algorithm to the large dataset of normal scans.
Further experiments used a variety of normal scan types from a third institute as well as an open-source dataset.
Their final model was tested using scans with a disproportionate amount of brain volume loss and then scrutinized their model findings by building heatmaps of the parts of the scans which the model predicted there was a disproportionate amount of brain volume loss.
First author Dr David Wood, researcher at the School of Biomedical Engineering and Imaging Sciences, said a key aspect of this study was the use of a large, clinically-representative dataset for model training.
The researchers say this framework could have important implications for patient care, drug development, and optimizing MRI data collection.
“Currently abnormal older-appearing brains are detected sometime after the scan at the time of reporting. The most accurate reports will be in centers where there are neuroradiologists however few centers have neuroradiologists,” Dr Booth said.
“Automatically detecting volume loss in real time helps screen for the common problem of neurodegeneration during scans obtained for all reasons. A subsequent diagnosis of, for example early-stage Alzheimer’s disease, could potentially improve patient care through implementing early medical and social interventions. Similarly, patients could potentially be recruited into drug trials at an earlier stage.”
Dr Booth said the framework could also be used to leverage the wealth of existing large hospital databases to provide powerful new resources for the training, testing and clinical validation of medical image analysis tools beyond brain-age such as abnormality detection.
About this AI and brain aging research news
Original Research: Open access.
“Accurate brain‐age models for routine clinical MRI examinations” by David A. Wood et al. NeuroImage
Accurate brain‐age models for routine clinical MRI examinations
Convolutional neural networks (CNN) can accurately predict chronological age in healthy individuals from structural MRI brain scans. Potentially, these models could be applied during routine clinical examinations to detect deviations from healthy ageing, including early-stage neurodegeneration. This could have important implications for patient care, drug development, and optimising MRI data collection. However, existing brain-age models are typically optimised for scans which are not part of routine examinations (e.g., volumetric T1-weighted scans), generalise poorly (e.g., to data from different scanner vendors and hospitals etc.), or rely on computationally expensive pre-processing steps which limit real-time clinical utility.
Here, we sought to develop a brain-age framework suitable for use during routine clinical head MRI examinations. Using a deep learning-based neuroradiology report classifier, we generated a dataset of 23,302 ‘radiologically normal for age’ head MRI examinations from two large UK hospitals for model training and testing (age range = 18–95 years), and demonstrate fast (< 5 s), accurate (mean absolute error [MAE] < 4 years) age prediction from clinical-grade, minimally processed axial T2-weighted and axial diffusion-weighted scans, with generalisability between hospitals and scanner vendors (Δ MAE < 1 year).
The clinical relevance of these brain-age predictions was tested using 228 patients whose MRIs were reported independently by neuroradiologists as showing atrophy ‘excessive for age’. These patients had systematically higher brain-predicted age than chronological age (mean predicted age difference = +5.89 years, ‘radiologically normal for age’ mean predicted age difference = +0.05 years, p < 0.0001).
Our brain-age framework demonstrates feasibility for use as a screening tool during routine hospital examinations to automatically detect older-appearing brains in real-time, with relevance for clinical decision-making and optimising patient pathways.