For many years, studies in non-human primates have shown that theĚýprefrontal cortex (PFC)—a part of the brain involved in decision-making, planning, and complex thinking—contains a dense network ofĚýshort-range connections(local wiring between nearby brain regions). However, studying these fine connections in living humans has been difficult because non-invasive imaging methods likeĚýdiffusion tractographyĚý(a technique that estimates brain pathways by tracking water movement) can produce inaccurate results, including false connections.

In this study, researchers developed a new approach to map these local brain connections more reliably in living humans. They combined high-resolution tractography with prior knowledge fromĚýhistology (microscopic studies of brain tissue, considered a gold standard for anatomical detail) to guide their analysis. Using brain scans from over 1,000 individuals, they were able to map 91 specific short-range connections within and between five key regions of the PFC. Their method showed strong agreement with known anatomical data, achieving over 80% precision (correctly identified connections) and over 70% accuracy (overall correctness compared to histological findings). Importantly, the results captured not only general patterns of connectivity but also subtle details that had previously only been observed in invasive studies.

The study also found that these brain connections are highly consistent within the same person over time, yet vary meaningfully between individuals—suggesting each person has a stable but unique “wiring pattern” in their PFC. Overall, this work demonstrates that combining detailed anatomical knowledge with advanced imaging can significantly improve our ability to map the human brain’s internal connections. This opens new possibilities for understanding how local brain circuits support thinking and behavior, and how they may be altered in neurological or psychiatric conditions.

FIGURE 1

(a.) Schematic depicting the interconnections of the prefrontal cortex (PFC). Each dot represents a connection with histological precedence. Red = dl-PFC (dorsolateral prefrontal), blue = vl-PFC (ventrolateral prefrontal), orange = orbitofrontal, purple = F. Pole (frontal pole), green = ACC (anterior cingulate). (b.) Table overview of histologically supported connections. + = consistently shown in histological literature.

This study looks at how changes in brain structure are linked to the return of depression in older adults. Specifically, it focuses onĚýwhite matter hyperintensities (WMH)Ěý˛ą˛Ô»ĺĚýhypointensities (WMh)—areas in brain scans that appear unusually bright or dark and are thought to reflect small blood vessel damage and increased vascular (blood flow–related) risk. These markers are commonly seen in older individuals and have been associated with late-life depression (LLD), but it has been unclear whether changes in these brain features over time contribute to depression coming back after recovery.

To investigate this, researchers followed 223 older adults (average age about 67), including people whose depression had improved (remitted LLD) and a comparison group without depression. Brain scans were taken every 8 months over two years to track changes in WMh. During this period, about half of the participants who had recovered from depression experienced a relapse. The researchers found that people who relapsed already had higher levels of WMh at the start of the study compared to those without depression. However, theĚýrate at which these brain changes increased over time was not significantly different between groups. When looking more closely, individuals who started with high WMh levels and also showed faster accumulation over time had nearly three times the risk of relapse compared to those with low levels and slow changes.

Overall, the findings suggest that having a higher burden of these brain changes at baseline is an important risk factor for depression returning in older adults, while the speed of progression alone may be less informative. However, people with both high initial levels and rapid increases may be at especially high risk and could benefit from closer monitoring and care to help prevent relapse.

Figure. 1ĚýMixed effects model predictions. (A), (B): Results of the model comparing HC vs. remLLD. (C), (D): Results comparing HC vs. REM vs. EarlyREL vs. LateREL. EarlyREL and LateREL represent individuals that relapse within 250 days of baseline and after 250 days of baseline, respectively. Both models adjusted for time (days since baseline), vascular disease burden using CIRS-G, age at baseline, ICV at baseline, sex, education, race, study site, group, and time * group effects.

In large, multi-site Alzheimer’s disease studies, brain scans known as amyloid positron emission tomography (PET) are usually interpreted by a small group of central experts. However, expanding the number of trained readers across research centers could make this process more scalable and improve how well the results apply across different hospitals and imaging systems. In this study, ten neuroimaging experts from eight Alzheimer’s Disease Research Centers (ADRCs) reviewed 180 amyloid PET scans. The scans included four different amyloid tracers and were collected from a wide range of imaging systems. Each expert analyzed 30 scans and 15 duplicate scans using their preferred viewing software, without the use of anatomical images or automated quantitative tools. Scans were classified as either elevated or non-elevated in amyloid according to tracer-specific reading criteria. The level of agreement between readers (inter-rater agreement) and within each reader’s repeated assessments (intra-rater agreement) was then evaluated.

The study found substantial agreement among experts, with a Fleiss’ κ of 0.78, and complete consensus on 69% of scans. Reliability across the four amyloid tracers ranged from substantial to perfect (Fleiss’ κ = 0.70–0.87), while intra-rater consistency ranged from 0.79 to 1.0 (Cohen’s κ), indicating that readers were highly consistent with themselves. Agreement was lower for scans showing intermediate amyloid levels, corresponding to 10–40 Centiloids, a range that is more difficult to interpret.

Overall, the findings demonstrate that a distributed network of experts across multiple centers can reliably classify amyloid PET scans, even when using different tracers and reading software. The set of scans used in this project will serve as a valuable reference for reader training and quality assurance in future multicenter Alzheimer’s disease studies.

FIGURE 1

Amyloid PET scans examples and Centiloid distribution. A, Examples of amyloid PET scans for each radiotracer that span the Centiloid continuum. B, Centiloid distribution of amyloid PET scans by tracer. Red dotted lines represent the boundaries of the 10–40 Centiloid intermediate range. FBB, [¹⁸F]florbetaben; FBP, [¹⁸F]florbetapir; FFN, [¹⁸F]flutafuranol; PET, positron emission tomography; PIB, [¹¹C]Pittsburgh compound B.

Magnetic resonance images of healthy brains were analyzed to better understand how resting-state BOLD signals in white matter are related to neural activity in the cortex (the outer layer of the brain). We measured how much spontaneous activity in the cortex—seen as low-frequency fluctuations in BOLD signals from gray matter—affects the resting-state BOLD signals in white matter. We found that the similarity between BOLD signals from cortical regions and white matter areas was directly linked to the strength of the cortical BOLD signal.

From these measurements, we observed that cortical networks involved in more basic functions tend to contribute more to the fluctuations in white matter than those involved in higher-level functions. We also discovered that each cortical network has its own unique spatial pattern of influence on white matter BOLD signals, and the strength of these effects is closely related to how much myelin (the protective coating around nerve fibers) the cortical network has.

Overall, our findings show that resting-state BOLD signals in white matter reflect the spontaneous activity of specific cortical networks and are shaped by the structure and myelination of the cortex.

Fig 1

(a) Relationship between subject-averaged fALFF of cortical BOLD signals and their subject-averaged mean white matter projection. Each data point represents subject-averaged measures for an ROI in the cortex. (b) Mean white matter projection of BOLD signals in the cortical functional networks analyzed. The vertical line at the top of each bar represents standard error across the 120 subjects studied. Abbreviations: prim = primary, DMN = default mode network. LECN = left executive control network. RECN = right executive control network.

The perivascular space (PVS) plays an important role in helping the brain clear out waste by allowing fluid to flow around blood vessels. A brain imaging method called DTI-ALPS was suggested as a way to measure how fluid moves in these spaces without surgery. However, it’s not clear how accurate or specific this method is. The DTI-ALPS method assumes certain patterns in brain tissue called “radial symmetrby” and interprets when these patterns are uneven (called “radial asymmetry”) as a sign of fluid movement in the PVS. But other factors in the brain’s structure might affect these measurements.

In this study, we carefully examined these possible influences using detailed brain scans from the Human Connectome Project and high-resolution imaging. We looked at how common radial asymmetry is in brain white matter, how crossing nerve fibers affect the measurements, how nerve fibers’ twisting and spreading impact results, and how blood vessels are oriented in these brain areas. We found that radial asymmetry happens widely in white matter and is mostly caused by the shape and arrangement of nerve fibers—not just fluid in the PVS. Crossing fibers made the measurements seem larger, and twisting or spreading of fibers also caused asymmetry, regardless of fluid flow. Additionally, blood vessels were not always aligned in the way the method assumes.

Overall, the DTI-ALPS measurements are strongly influenced by the brain’s nerve fiber structure rather than just fluid movement in the perivascular space. This means that using DTI-ALPS as a direct marker of the brain’s waste clearance system might be misleading unless these structural factors are considered. Future research should use more advanced methods to separate the effects of fluid flow from the complex structure of brain tissue.

Fig 1

Radial asymmetry is widespread throughout white matter. Sagittal, coronal, and axial slices of an example HCP subject show radial asymmetry at all diffusion weightings, and throughout white matter, with most regions exhibiting average asymmetry values ~1.3–1.8, with many voxels > 2.

Brain scans can be used to estimate a person’s “brain age,” which may be older or younger than their actual age. A larger difference between brain age and actual age can provide early warning signs of neurodegenerative diseases like Alzheimer’s, potentially allowing for earlier diagnosis and prevention. One type of brain scan, called diffusion MRI (dMRI), is especially useful for this because it can detect very subtle changes in the brain’s structure that may happen before more obvious signs appear. However, dMRI captures both large-scale (macrostructural) and small-scale (microstructural) features of the brain, and it’s unclear whether current models for estimating brain age from dMRI are focusing on the small-scale changes that matter most for early detection.

To better isolate the microstructural information, this study developed a new approach that reduces the influence of macrostructural features by aligning all brain scans to a common reference template. The method was tested using imaging data from 13,398 people across 12 different datasets. The researchers compared this new microstructure-focused dMRI brain age model to several other models based on T1-weighted MRI, a common type of scan that primarily captures macrostructural features.

They found that the dMRI-based brain age and T1-based brain age showed different patterns depending on the stage of disease. For people who were transitioning from normal cognitive function to mild cognitive impairment (MCI), the dMRI brain age appeared older than the T1-based brain age. In contrast, for those already diagnosed with Alzheimer’s disease, the dMRI brain age appeared younger. Models based on T1-weighted MRI generally performed better at identifying who had Alzheimer’s, but the dMRI-based brain age may be more helpful in identifying early, subtle changes that happen before symptoms begin.

Fig 1

Brain age estimation frameworks have proven effective for using affinely aligned brain images to identify common patterns of aging, with deviations from these patterns likely indicating presence of abnormal neuropathologic processes. A common theme of existing brain age estimation methods has been using T1w MRI, denoted as “GM age” in the first row. Among them, there have been many innovations in network design, such as DeepBrainNet (DBN) (Bashyam et al., 2020) and the 3D convolutional neural network of TSAN (Cheng et al., 2021). T1w MRI lacks detail in white matter (WM). Here, we take the two most commonly used modalities for characterizing WM microstructure, fractional anisotropy (FA), and mean diffusivity (MD), and we evaluate brain age estimation in two contexts. First, we examine the direct substitution of FA and MD for T1w image, which we denote as “WM age affine” in the second row. A substantial amount of macrostructural differences is still present in WM age affine, notably ventricle enlargement. To isolate the microstructural changes, we apply non-rigid (deformable) registration into template space to mitigate the macrostructural changes and produce the “WM age nonrigid” in the third row. We explore the relative timing of changes in these brain age variants and their relative explainability in the context of mild cognitive impairment. Throughout the paper, we adhere to a consistent color scheme when visualizing results from different brain age estimates within the same plot to facilitate easier visual inspection. Specifically, we use red to represent GM ages, blue for WM age nonrigid, and purple for WM age affine.

Vigilance—our level of alertness or attention—naturally rises and falls over time. These shifts are known to affect signals seen in brain scans, such as those from functional magnetic resonance imaging (fMRI), though the exact cause of these changes isn’t fully understood. Separate studies have connected changes in vigilance not only to brain signal patterns but also to changes in physical responses controlled by the autonomic nervous system, such as breathing and heart rate. This raises the question: could some of the brain signal changes actually be caused by these bodily responses?Ěý

To explore this, we recorded fMRI scans alongside measures of brain activity (EEG), breathing, and blood oxygen levels, while people were either resting or doing a task that required attention. We found that the link between the body’s automatic functions (like pulse and respiration) and brain signals became stronger as people became less alert. These body-related signals first showed quick positive connections with brain activity, then slower negative ones, with some later positive responses in fluid-filled spaces of the brain.Ěý

We also saw that fluctuations in EEG (a measure of brainwave activity) depended on alertness level and were related to both brain and body signals. Additionally, the strength of communication between different brain regions (called functional connectivity) increased when people were less alert—especially during rest. But when we removed the influence of body signals from the fMRI data, this increase mostly disappeared.Ěý

Overall, our results show that changes in alertness affect not just brain activity but also how the body and brain interact. This understanding helps scientists more accurately interpret fMRI data by highlighting the important role of physiological changes.Ěý

Ěý

Fig 1. Ěý

Comparing EEG, fMRI, autonomic, and behavioral measures across time windows. (A) Simultaneous EEG, fMRI, and autonomic data were divided into non-overlapping windows of 126 s each. This panel shows three representative, contiguous windows (the fourth, fifth, and sixth windows from rest participant 3), including their “fast” (i.e., seconds-level) and baseline (i.e., window-averaged) EEG alpha/theta ratios, for a participant in the process of falling asleep. (B) For the task scans, we compared the mean of the EEG alpha/theta power ratio within each window (which we define as “baseline vigilance”) to the mean reaction time in each window with Spearman’s rank correlations for non-normal distributions. Significant negative correlations, whether excluding trials without responses (“Responses only”) or including them as indicating arbitrarily long reaction times of 4 s (“All trials”), support the use of an EEG alpha/theta ratio as a measure of vigilance in this study. (C) The temporal variance of the percent signal change in the fMRI global signal also exhibited a negative relationship with baseline vigilance levels (shifted by 4.2 s in this case to accommodate the hemodynamic delay of the fMRI signal). This effect was significant for both resting-state and task data, indicating greater global fMRI variability as baseline vigilance decreases. Although the correlation values shown in (B–C) are based on non-parametric statistics, we include least-squares trend lines for visualization. RV = respiratory volume, HR = heart rate, PWA = pulse wave amplitude.Ěý

Ěý