Recent critical commentaries unfavorably compare deep learning (DL) with standard machine learning (SML) approaches for brain imaging data analysis. However, their conclusions are often based on pre-engineered features depriving DL of its main advantage — representation learning. We conduct a large-scale systematic comparison profiled in multiple classification and regression tasks on structural MRI images and show the importance of representation learning for DL. Results show that if trained following prevalent DL practices, DL methods have the potential to scale particularly well and substantially improve compared to SML methods, while also presenting a lower asymptotic complexity in relative computational time, despite being more complex. We also demonstrate that DL embeddings span comprehensible task-specific projection spectra and that DL consistently localizes task-discriminative brain biomarkers. Our findings highlight the presence of nonlinearities in neuroimaging data that DL can exploit to generate superior task-discriminative representations for characterizing the human brain.

The past few years have seen an emergence of approaches that leverage temporal changes in whole-brain patterns of functional connectivity (the chronnectome). In this chronnectome study, we investigate the replicability of the human brain's inter-regional coupling dynamics during rest by evaluating two different dynamic functional network connectivity (dFNC) analysis frameworks using 7 500 functional magnetic resonance imaging (fMRI) datasets. To quantify the extent to which the emergent functional connectivity (FC) patterns are reproducible, we characterize the temporal dynamics by deriving several summary measures across multiple large, independent age-matched samples. Reproducibility was demonstrated through the existence of basic connectivity patterns (FC states) amidst an ensemble of inter-regional connections. Furthermore, application of the methods to conservatively configured (statistically stationary, linear and Gaussian) surrogate datasets revealed that some of the studied state summary measures were indeed statistically significant and also suggested that this class of null model did not explain the fMRI data fully. This extensive testing of reproducibility of similarity statistics also suggests that the estimated FC states are robust against variation in data quality, analysis, grouping, and decomposition methods. We conclude that future investigations probing the functional and neurophysiological relevance of time-varying connectivity assume critical importance.

Exploring brain changes across the human lifespan is becoming an important topic in neuroscience. Though there are multiple studies which investigated the relationship between age and brain imaging, the results are heterogeneous due to small sample sizes and relatively narrow age ranges. Here, based on year‐wise estimation of 5,967 subjects from 13 to 72 years old, we aimed to provide a more precise description of adult lifespan variation trajectories of gray matter volume (GMV), structural network correlation (SNC), and functional network connectivity (FNC) using independent component analysis and multivariate linear regression model. Our results revealed the following relationships: (a) GMV linearly declined with age in most regions, while parahippocampus showed an inverted U‐shape quadratic relationship with age; SNC presented a U‐shape quadratic relationship with age within cerebellum, and inverted U‐shape relationship primarily in the default mode network (DMN) and frontoparietal (FP) related correlation. (b) FNC tended to linearly decrease within resting‐state networks (RSNs), especially in the visual network and DMN. Early increase was revealed between RSNs, primarily in FP and DMN, which experienced a decrease at older ages. U‐shape relationship was also revealed to compensate for the cognition deficit in attention and subcortical related connectivity at late years. (c) The link between middle occipital gyrus and insula, as well as precuneus and cerebellum, exhibited similar changing trends between SNC and FNC across the adult lifespan. Collectively, these results highlight the benefit of lifespan study and provide a precise description of age‐related regional variation and SNC/FNC changes based on a large dataset.

Multimodal neuroimaging is an emerging field that leverages multiple sources of information to diagnose specific brain disorders, especially when deep learning‐based AI algorithms are applied. The successful combination of different brain imaging modalities using deep learning remains a challenging yet crucial research topic. The integration of structural and functional modalities is particularly important for the diagnosis of various brain disorders, where structural information plays a crucial role in diseases such as Alzheimer's, while functional imaging is more critical for disorders such as schizophrenia. However, the combination of functional and structural imaging modalities can provide a more comprehensive diagnosis. In this work, we present MultiViT, a novel diagnostic deep learning model that utilizes vision transformers and cross‐attention mechanisms to effectively fuse information from 3D gray matter maps derived from structural MRI with functional network connectivity matrices obtained from functional MRI using the ICA algorithm. MultiViT achieves an AUC of 0.833, outperforming both our unimodal and multimodal baselines, enabling more accurate classification and diagnosis of schizophrenia. In addition, using vision transformer's unique attentional maps in combination with cross‐attentional mechanisms and brain function information, we identify critical brain regions in 3D gray matter space associated with the characteristics of schizophrenia. Our research not only significantly improves the accuracy of AI‐based automated imaging diagnostics for schizophrenia, but also pioneers a rational and advanced data fusion approach by replacing complex, high‐dimensional fMRI information with functional network connectivity, integrating it with representative structural data from 3D gray matter images, and further providing interpretative biomarker localization in a 3D structural space.