STUDY OBJECTIVES: To use relatively noisy routinely collected clinical data (brain magnetic resonance imaging (MRI) data, clinical polysomnography (PSG) recordings, and neuropsychological testing), to investigate hypothesis-driven and data-driven relationships between brain physiology, structure, and cognition.

METHODS: We analyzed data from patients with clinical PSG, brain MRI, and neuropsychological evaluations. SynthSeg, a neural network-based tool, provided high-quality segmentations despite noise. A priori hypotheses explored associations between brain function (measured by PSG) and brain structure (measured by MRI). Associations with cognitive scores and dementia status were studied. An exploratory data-driven approach investigated age-structure-physiology-cognition links.

RESULTS: Six hundred and twenty-three patients with sleep PSG and brain MRI data were included in this study; 160 with cognitive evaluations. Three hundred and forty-two participants (55%) were female, and age interquartile range was 52 to 69 years. Thirty-six individuals were diagnosed with dementia, 71 with mild cognitive impairment, and 326 with major depression. One hundred and fifteen individuals were evaluated for insomnia and 138 participants had an apnea-hypopnea index equal to or greater than 15. Total PSG delta power correlated positively with frontal lobe/thalamic volumes, and sleep spindle density with thalamic volume. rapid eye movement (REM) duration and amygdala volume were positively associated with cognition. Patients with dementia showed significant differences in five brain structure volumes. REM duration, spindle, and slow-oscillation features had strong associations with cognition and brain structure volumes. PSG and MRI features in combination predicted chronological age (R2 = 0.67) and cognition (R2 = 0.40).

CONCLUSIONS: Routine clinical data holds extended value in understanding and even clinically using brain-sleep-cognition relationships.

Image-to-image translation has seen major advances in computer vision but can be difficult to apply to medical images, where imaging artifacts and data scarcity degrade the performance of conditional generative adversarial networks. We develop the spatial-intensity transform (SIT) to improve output image quality while closely matching the target domain. SIT constrains the generator to a smooth spatial transform (diffeomorphism) composed with sparse intensity changes. SIT is a lightweight, modular network component that is effective on various architectures and training schemes. Relative to unconstrained baselines, this technique significantly improves image fidelity, and our models generalize robustly to different scanners. Additionally, SIT provides a disentangled view of anatomical and textural changes for each translation, making it easier to interpret the model's predictions in terms of physiological phenomena. We demonstrate SIT on two tasks: predicting longitudinal brain MRIs in patients with various stages of neurodegeneration, and visualizing changes with age and stroke severity in clinical brain scans of stroke patients. On the first task, our model accurately forecasts brain aging trajectories without supervised training on paired scans. On the second task, it captures associations between ventricle expansion and aging, as well as between white matter hyperintensities and stroke severity. As conditional generative models become increasingly versatile tools for visualization and forecasting, our approach demonstrates a simple and powerful technique for improving robustness, which is critical for translation to clinical settings. Source code is available at github.com/clintonjwang/spatial-intensity-transforms.

The standard of care for brain tumors is maximal safe surgical resection as the first step. Neuronavigation augments the surgeon's ability to achieve this but loses validity due to brain shift as surgery progresses. Moreover, many gliomas are difficult to distinguish from adjacent healthy brain tissue. Intraoperative MRI (iMRI) is a useful surgical adjunct that can be used to visualize the residual tumor and brain shift. Intraoperative ultrasound (iUS) serves a similar purpose, while also being faster and easier to incorporate into the workflow. However, it provides lower contrast between tumor tissue and normal brain tissue as compared to intraoperative MRI. With the success of data-hungry Artificial Intelligence (AI)/Machine Learning (ML) algorithms in advancing the state of the art in medical image analysis, the benefits of sharing well-curated data can not be overstated. To this end, we provide here the largest publicly-available MRI and intraoperative ultrasound imaging database of surgically treated brain tumors, including gliomas (n=92), metastases (n=11), and others (n=11). This collection contains 369 preoperative MRI series, 320 3D intraoperative ultrasound series, 301 intraoperative MRI series, and 356 segmentations collected from 114 consecutive patients at a single institution. We expect this data to be a resource for computational research in brain shift and image analysis as well as for neurosurgical training in the interpretation of intraoperative ultrasound and iMRI.

The corticospinal tract (CST) is a critically important white matter fiber tract in the human brain that enables control of voluntary movements of the body. The CST exhibits a somatotopic organization, which means that the motor neurons that control specific body parts are arranged in order within the CST. Diffusion magnetic resonance imaging (MRI) tractography is increasingly used to study the anatomy of the CST. However, despite many advances in tractography algorithms over the past decade, modern, state-of-the-art methods still face challenges. In this study, we compare the performance of six widely used tractography methods for reconstructing the CST and its somatotopic organization. These methods include constrained spherical deconvolution (CSD) based probabilistic (iFOD1) and deterministic (SD-Stream) methods, unscented Kalman filter (UKF) tractography methods including multi-fiber (UKF2T) and single-fiber (UKF1T) models, the generalized q-sampling imaging (GQI) based deterministic tractography method, and the TractSeg method. We investigate CST somatotopy by dividing the CST into four subdivisions per hemisphere that originate in the leg, trunk, hand, and face areas of the primary motor cortex. A quantitative and visual comparison is performed using diffusion MRI data (N = 100 subjects) from the Human Connectome Project. Quantitative evaluations include the reconstruction rate of the eight anatomical subdivisions, the percentage of streamlines in each subdivision, and the coverage of the white matter-gray matter (WM-GM) interface. CST somatotopy is further evaluated by comparing the percentage of streamlines in each subdivision to the cortical volumes for the leg, trunk, hand, and face areas. Overall, UKF2T has the highest reconstruction rate and cortical coverage. It is the only method with a significant positive correlation between the percentage of streamlines in each subdivision and the volume of the corresponding motor cortex. However, our experimental results show that all compared tractography methods are biased toward generating many trunk streamlines (ranging from 35.10% to 71.66% of total streamlines across methods). Furthermore, the coverage of the WM-GM interface in the largest motor area (face) is generally low (under 40%) for all compared tractography methods. Different tractography methods give conflicting results regarding the percentage of streamlines in each subdivision and the volume of the corresponding motor cortex, indicating that there is generally no clear relationship, and that reconstruction of CST somatotopy is still a large challenge. Overall, we conclude that while current tractography methods have made progress toward the well-known challenge of improving the reconstruction of the lateral projections of the CST, the overall problem of performing a comprehensive CST reconstruction, including clinically important projections in the lateral (hand and face areas) and medial portions (leg area), remains an important challenge for diffusion MRI tractography.

OBJECTIVE: To investigate how the presence/side of hippocampal sclerosis (HS) are related to the white matter structure of cingulum bundle (CB), arcuate fasciculus (AF), and inferior longitudinal fasciculus (ILF) in mesial temporal lobe epilepsy (MTLE).

METHODS: We acquired diffusion-weighted magnetic resonance imaging (MRI) from 86 healthy and 71 individuals with MTLE (22 righ-HS; right-HS, 34 left-HS; left-HS, and 15 nonlesional MTLE). We utilized two-tensor tractography and fiber clustering to compare fractional anisotropy (FA) of each side/tract between groups. Additionally, we examined the association between FA and nonverbal (WMS-R) and verbal (WMS-R, RAVLT codification) memory performance for MTLE individuals.

RESULTS: White matter abnormalities depended on the side and presence of HS. The left-HS demonstrated widespread abnormalities for all tracts, the right-HS showed lower FA for ipsilateral tracts and the nonlesional MTLE group did not differ from healthy individuals. Results indicate no differences in verbal/nonverbal memory performance between the groups, but trend-level associations between higher FA of visual memory and the left CB (r = 0.286, P = 0.018), verbal memory (RAVLT) and -left CB (r = 0.335, P = 0.005), -right CB (r = 0.286, P = 0.016), and -left AF (r = 0.287, P = 0.017).

SIGNIFICANCE: Our results highlight that the presence and side of HS are crucial to understand the pathophysiology of MTLE. Specifically, left-sided HS seems to be related to widespread bilateral white matter abnormalities. Future longitudinal studies should focus on developing diagnostic and treatment strategies dependent on HS's presence/side.