A Deep Learning Framework for Predicting Infant Brain Age from Myelin Maturation in MRI
Poster No:
1023
Submission Type:
Abstract Submission
Authors:
Anand Joshi1, Jessica Wisnowski2, Sachin Salim1, Richard Leahy1, Vidya Rajagopalan2, David Shattuck3
Institutions:
1University of Southern California, Los Angeles, CA, 2Children's Hospital, Los Angeles, Los Angeles, CA, 3University of California, Los Angeles, Los Angeles, CA
First Author:
Co-Author(s):
Introduction:
Infants undergo a critical period of brain maturation during the first year of life, characterized by rapid growth in cortical and subcortical regions, synaptogenesis, and myelination. During this time, gray matter volume nearly doubles, while white matter develops through progressive myelination, which enhances neural connectivity and signal efficiency [2]. These processes are reflected in evolving T1- and T2-weighted MRI signal intensities, corresponding to shifts in water and lipid content. While this maturation follows a typical trajectory, deviations have been linked to neurodevelopmental disorders such as autism and cerebral palsy, underscoring the need for automated, robust methods to analyze developmental changes and detect deviations early. However, analyzing infant brain MRI poses unique challenges, including small brain size, underdeveloped cortical folding, and contrast inversion, where unmyelinated white matter appears hypointense on T1-weighted images and hyperintense on T2-weighted images. We propose a novel deep learning framework that analyzes T1- and T2-weighted MRI data to predict brain age and to monitor developmental changes in infants.
Methods:
We used three datasets in the development and evaluation of our method, which included MRI from a total of N=1548 subjects. These included: (1) the Cambridge Centre for Aging and Neuroscience (CamCAN) project (N3=652) [4] (2) the Baby Open Brains Repository (N1=63) [3]; and (3) Pediatric Imaging dataset (N2=833) [1]. The data from each of these three datasets were subdivided into a training set (80%) and a validation set (20%).
Our approach employed a U-Net-based neural network with a fully-connected layer in the latent space to classify tissue types and predict brain age. To enhance generalizability across varying MRI signal characteristics and achieve robustness to contrast inversion, we applied data augmentation techniques, including simulated contrast inversion on the CamCAN dataset. The model was trained and validated on multiple infant brain MRI datasets to encompass a wide spectrum of developmental features, ensuring robustness across diverse populations (see Fig. 1). Expert-estimated brain ages served as the ground truth for our evaluations. Model performance was assessed using metrics such as accuracy, Dice coefficient scores, and correlations with ground-truth brain ages.
Our approach employed a U-Net-based neural network with a fully-connected layer in the latent space to classify tissue types and predict brain age. To enhance generalizability across varying MRI signal characteristics and achieve robustness to contrast inversion, we applied data augmentation techniques, including simulated contrast inversion on the CamCAN dataset. The model was trained and validated on multiple infant brain MRI datasets to encompass a wide spectrum of developmental features, ensuring robustness across diverse populations (see Fig. 1). Expert-estimated brain ages served as the ground truth for our evaluations. Model performance was assessed using metrics such as accuracy, Dice coefficient scores, and correlations with ground-truth brain ages.
Results:
Our deep learning framework achieved high segmentation accuracy across diverse infant brain MRI datasets, with average Dice coefficients exceeding 0.85 for gray and white matter tissue classification. Brain age predictions demonstrated a correlation (r > 0.9) with corrected age, confirming the model's ability to capture developmental milestones. Augmentation strategies effectively addressed MRI contrast inversion, enabling accurate predictions across different developmental stages. The model predicted brain age with mean absolute error (MAE) of 1.7 ±.0.9 weeks. Figure 2 shows a comparison of ground truth and automatic segmentation for one example subject and violin plots of the differences in age prediction results from expert-determined age and chronological age. This highlights the utility of incorporating regression capabilities into the model for predicting developmental biomarkers.
Conclusions:
We introduced a deep learning framework designed to analyze infant brain MRI for tissue segmentation and estimation of physiologic brain age. The model showed robust performance in tissue classification and brain age prediction despite challenges unique to infant imaging such as contrast inversion, limited anatomical differentiation, and small training sets. Early identification of deviations from normative brain development patterns in infants holds significant potential for detecting neurodevelopmental disorders and informing personalized interventions during critical developmental windows.
Lifespan Development:
Early life, Adolescence, Aging
Normal Brain Development: Fetus to Adolescence 1
Modeling and Analysis Methods:
Classification and Predictive Modeling
Methods Development 2
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Normal Development
Keywords:
Machine Learning
Modeling
MRI
Myelin
PEDIATRIC
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
[2] Dubois, J., et al. (2021). MRI of the neonatal brain: a review of methodological challenges and neuroscientific advances. Journal of Magnetic Resonance Imaging, 53(5), 1318-1343.
[3] Feczko, E., Stoyell,et al. (2024). Baby Open Brains: An Open-Source Repository of Infant Brain Segmentations. bioRxiv, 2024-10.
[4] Taylor, J. R., et al. (2017). The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: Structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample. neuroimage, 144, 262-269.