Different task-based and resting-state imaging datasets provide complementary information about the organization of the human brain. Brain parcellations based on single datasets will, therefore, be biased toward the particular type of information present in each dataset. To overcome this limitation, we propose here a hierarchical Bayesian framework that can learn a probabilistic brain parcellation across numerous task-based and resting-state datasets, exploiting their combined strengths. The framework is partitioned into a spatial arrangement model that defines the probability of each voxel belonging to a specific parcel (the probabilistic group atlas), and a set of dataset-specific emission models that define the probability of the observed data given the parcel of the voxel. Using the human cerebellum as an example, we show that the framework optimally combines information from different datasets to achieve a new population-based atlas that outperforms atlases based on single datasets. Furthermore, we demonstrate that using only 10 min of individual data, the framework is able to generate individual brain parcellations that outperform group atlases.

Individualized phenotypic prediction based on structural magnetic resonance imaging (MRI) is an important goal in neuroscience. Prediction performance increases with larger samples, but small-scale datasets with fewer than 200 participants are often unavoidable. We have previously proposed a “meta-matching” framework to translate models trained from large datasets to improve the prediction of new unseen phenotypes in small collection efforts. Meta-matching exploits correlations between phenotypes, yielding large improvement over classical machine learning when applied to prediction models using resting-state functional connectivity as input features. Here, we adapt the two best performing meta-matching variants (“meta-matching finetune” and “meta-matching stacking”) from our previous study to work with T1-weighted MRI data by changing the base neural network architecture to a 3D convolution neural network. We compare the two meta-matching variants with elastic net and classical transfer learning using the UK Biobank (N = 36,461), the Human Connectome Project Young Adults (HCP-YA) dataset (N = 1,017), and the HCP-Aging dataset (N = 656). We find that meta-matching outperforms elastic net and classical transfer learning by a large margin, both when translating models within the same dataset and when translating models across datasets with different MRI scanners, acquisition protocols, and demographics. For example, when translating a UK Biobank model to 100 HCP-YA participants, meta-matching finetune yielded a 136% improvement in variance explained over transfer learning, with an average absolute gain of 2.6% (minimum = –0.9%, maximum = 17.6%) across 35 phenotypes. Overall, our results highlight the versatility of the meta-matching framework.

Resting-state functional connectivity (RSFC) is widely used to predict phenotypic traits in individuals. Large sample sizes can significantly improve prediction accuracies. However, for studies of certain clinical populations or focused neuroscience inquiries, small-scale datasets often remain a necessity. We have previously proposed a “meta-matching” approach to translate prediction models from large datasets to predict new phenotypes in small datasets. We demonstrated a large improvement over classical kernel ridge regression (KRR) when translating models from a single source dataset (UK Biobank) to the Human Connectome Project Young Adults (HCP-YA) dataset. In the current study, we propose two meta-matching variants (“meta-matching with dataset stacking” and “multilayer meta-matching”) to translate models from multiple source datasets across disparate sample sizes to predict new phenotypes in small target datasets. We evaluate both approaches by translating models trained from five source datasets (with sample sizes ranging from 862 participants to 36,834 participants) to predict phenotypes in the HCP-YA and HCP-Aging datasets. We find that multilayer meta-matching modestly outperforms meta-matching with dataset stacking. Both meta-matching variants perform better than the original “meta-matching with stacking” approach trained only on the UK Biobank. All meta-matching variants outperform classical KRR and transfer learning by a large margin. In fact, KRR is better than classical transfer learning when less than 50 participants are available for finetuning, suggesting the difficulty of classical transfer learning in the very small sample regime. The multilayer meta-matching model is publicly available at https://github.com/ThomasYeoLab/Meta_matching_models/tree/main/rs-fMRI/v2.0 .

The global signal (GS) in resting-state functional MRI (fMRI), known to contain artifacts and non-neuronal physiological signals, also contains important neural information related to individual state and trait characteristics. Here, we show distinct linear and curvilinear relationships between GS topography and age in a cross-sectional sample of individuals (6-85 years old) representing a significant portion of the lifespan. Subcortical brain regions such as the thalamus and putamen show linear associations with the GS across age. The thalamus has stronger contributions to the GS in older-age individuals compared with younger-aged individuals, while the putamen has stronger contributions in younger individuals compared with older individuals. The subcortical nucleus basalis of Meynert shows a u-shaped pattern similar to cortical regions within the lateral frontoparietal network and dorsal attention network, where contributions of the GS are stronger at early and old age, and weaker in middle age. This differentiation between subcortical and cortical brain activity across age supports a dual-layer model of GS composition, where subcortical aspects of the GS are differentiated from cortical aspects of the GS. We find that these subcortical-cortical contributions to the GS depend strongly on age across the lifespan of human development. Our findings demonstrate how neurobiological information within the GS differs across development and highlight the need to carefully consider whether or not to remove this signal when investigating age-related functional differences in the brain.