It has been hypothesized that the ventral stream processing for object recognition is based on a mechanism called cortically local subspace untangling. A mathematical abstraction of object recognition by the visual cortex is how to untangle the manifolds associated with different object categories. Such a manifold untangling problem is closely related to the celebrated kernel trick in metric space. In this paper, we conjecture that there is a more general solution to manifold untangling in the topological space without artificially defining any distance metric. Geometrically, we can either embed a manifold in a higher-dimensional space to promote selectivity or flatten a manifold to promote tolerance. General strategies of both global manifold embedding and local manifold flattening are presented and connected with existing work on the untangling of image, audio, and language data. We also discuss the implications of untangling the manifold into motor control and internal representations.

PurposeBrain glymphatic dysfunction is involved in the pathologic process of acute ischemic stroke (IS). The relationship between brain glymphatic activity and dysfunction in subacute IS has not been fully elucidated. Diffusion tensor image analysis along the perivascular space (DTI-ALPS) index was used in this study to explore whether glymphatic activity was related to motor dysfunction in subacute IS patients.MethodsTwenty-six subacute IS patients with a single lesion in the left subcortical region and 32 healthy controls (HCs) were recruited in this study. The DTI-ALPS index and DTI metrics (fractional anisotropy, FA, and mean diffusivity, MD) were compared within and between groups. Spearman's and Pearson's partial correlation analyses were performed to analyze the relationships of the DTI-ALPS index with Fugl-Meyer assessment (FMA) scores and with corticospinal tract (CST) integrity in the IS group, respectively.ResultsSix IS patients and two HCs were excluded. The left DTI-ALPS index of the IS group was significantly lower than that of the HC group (t = −3.02, p = 0.004). In the IS group, a positive correlation between the left DTI-ALPS index and the simple Fugl-Meyer motor function score (ρ = 0.52, p = 0.019) and a significant negative correlation between the left DTI-ALPS index and the FA (R = −0.55, p = 0.023) and MD (R = −0.48, p = 0.032) values of the right CST were found.ConclusionsGlymphatic dysfunction is involved in subacute IS. DTI-ALPS could be a potential magnetic resonance (MR) biomarker of motor dysfunction in subacute IS patients. These findings contribute to a better understanding of the pathophysiological mechanisms of IS and provide a new target for alternative treatments for IS.

AimStructural and functional changes in the brain have been identified in individuals with medication-overuse headache (MOH) using MRI. However, it has not been clearly established whether neurovascular dysfunction occurs in MOH, which could be elucidated by examining neurovascular coupling (NVC) from the viewpoints of neuronal activity and cerebral blood flow. The aim of this study was to investigate potential alterations in NVC function of the brain in individuals with MOH using resting-state functional MRI (rs-fMRI) and 3D pseudo-continuous arterial spin labeling (3D PCASL) imaging techniques.MethodsA total of 40 patients with MOH and 32 normal controls (NCs) were recruited, and rs-fMRI and 3D PCASL data were obtained using a 3.0 T MR scanner. Standard preprocessing of the rs-fMRI data was performed to generate images representing regional homogeneity (ReHo), fractional amplitude of low-frequency fluctuation (fALFF), and degree centrality (DC); cerebral blood flow (CBF) images were generated using 3D PCASL sequence data. These functional maps were all normalized into Montreal Neurological Institute (MNI) space, and NVC was subsequently determined on the basis of Pearson correlation coefficients between the rs-fMRI maps (ReHo, fALFF, and DC) and CBF maps. The statistical significance of differences between the MOH and NC groups in terms of NVC in different brain regions was established via Z-test. Further analysis was performed to examine correlations between NVC in the brain regions with NVC dysfunction and clinical variables among patients with MOH.ResultsNVC mainly presented a negative correlation in patients with MOH and NCs. No significant difference between the two groups was detected in terms of average NVC over the entire gray matter area. However, several brain regions with significantly decreased NVC in patients with MOH compared to NCs were identified: the left orbital region of the superior frontal gyrus, the bilateral gyrus rectus, and the olfactory cortex (P < 0.05). A correlation analysis revealed that the DC of the brain regions with NVC dysfunction was significantly positively correlated with disease duration (r = 0.323, P = 0.042), and DC–CBF connectivity was negatively correlated with VAS score (r = −0.424, P = 0.035).ConclusionThe current study demonstrated that cerebral NVC dysfunction occurs in patients with MOH, and the NVC technique could function as a new imaging biomarker in headache research.

This paper presents a theoretical perspective on modeling ventral stream processing by revisiting the computational abstraction of simple and complex cells. In parallel to David Marr's vision theory, we organize the new perspective into three levels. At the computational level, we abstract simple and complex cells into space partitioning and composition in a topological space based on the redundancy exploitation hypothesis of Horace Barlow. At the algorithmic level, we present a hierarchical extension of sparse coding by exploiting the manifold constraint in high-dimensional space (i.e., the blessing of dimensionality). The resulting over-parameterized models for object recognition differ from existing hierarchical models by disentangling the objectives of selectivity and invariance computation. It is possible to interpret our hierarchical construction as a computational implementation of cortically local subspace untangling for object recognition and face representation, which are closely related to exemplar-based and axis-based coding in the medial temporal lobe. At the implementation level, we briefly discuss two possible implementations based on asymmetric sparse autoencoders and divergent spiking neural networks.