scMoMaT jointly performs single cell mosaic integration and multi-modal bio-marker detection.

Single cell data integration methods aim to integrate cells across data batches and modalities, and data integration tasks can be categorized into horizontal, vertical, diagonal, and mosaic integration, where mosaic integration is the most general and challenging case with few methods developed. We propose scMoMaT, a method that is able to integrate single cell multi-omics data under the mosaic integration scenario using matrix tri-factorization. During integration, scMoMaT is also able to uncover the cluster specific bio-markers across modalities. These multi-modal bio-markers are used to interpret and annotate the clusters to cell types. Moreover, scMoMaT can integrate cell batches with unequal cell type compositions. Applying scMoMaT to multiple real and simulated datasets demonstrated these features of scMoMaT and showed that scMoMaT has superior performance compared to existing methods. Specifically, we show that integrated cell embedding combined with learned bio-markers lead to cell type annotations of higher quality or resolution compared to their original annotations.

Comparison of embolization strategies for mixed plexiform and fistulous brain arteriovenous malformations: a computational model analysis of theoretical risks of nidus rupture.

BACKGROUND: High-flow fistulas related to plexiform nidi are found in 40% of large brain arteriovenous malformations (AVMs). Endovascular occlusion of intranidal fistulas before plexiform components is empirically considered safe, but potential ensuing dangerous re-routing of flow through plexiform vessels may in theory raise their rupture risk. It remains unclear whether it is safer to embolize plexiform or fistulous vessels initially. We used a novel biomathematical AVM model to compare theoretical hemodynamic changes and rupture risks on sequential embolizations of both types of nidus vessels. METHODS: We computationally modeled a theoretical AVM as an electrical circuit containing a nidus consisting of a massive stochastic network ensemble comprising 1000 vessels. We sampled and individually simulated 10 000 different nidus morphologies with a fistula angioarchitecturally isolated from its adjacent plexiform nidus. We used network analysis to calculate mean intravascular pressure (Pmean) and flow rate within each nidus vessel; and Monte Carlo analysis to assess overall risks of nidus rupture when simulating sequential occlusions of vessel types in all 10 000 nidi. RESULTS: We consistently observed lower nidus rupture risks with initial fistula occlusion in different network morphologies. Intranidal fistula occlusion simultaneously reduced Pmean and flow rate within draining veins. CONCLUSIONS: Initial occlusion of AVM fistulas theoretically reduces downstream draining vessel hypertension and lowers the risk of rupture of an adjoining plexiform nidus component. This mitigates the theoretical concern that fistula occlusion may cause dangerous redistribution of hemodynamic forces into plexiform nidus vessels, and supports a clinical strategy favoring AVM fistula occlusion before plexiform nidus embolization.

Three-Dimensional Angles of Confluence of Cortical Bridging Veins and the Superior Sagittal Sinus on MR Venography: Does Drainage of Adjacent Brain Arteriovenous Malformations Alter this Spatial Configuration?

Few neuroimaging anatomic studies to date have investigated in detail the point of entry of cortical bridging veins (CBVs) into the superior sagittal sinus (SSS). Although we know that most CBVs join the SSS at an acute angle opposite to the direction of SSS blood flow, the three-dimensional (3-D) spatial configuration of these venous confluences has not been studied previously. This anatomical information would be pertinent to several clinically applicable scenarios, such as in planning intracranial surgical approaches that preserve bridging veins; studying anatomical factors in the pathophysiology of SSS thrombosis; and when planning endovascular microcatheterization of pial veins to retrogradely embolize brain arteriovenous malformations (AVMs). We used the concept of Euclidean planes in 3-D space to calculate the arccosine of these CBV-SSS angles of confluence. To test the hypothesis that pial AVM draining veins may not be any more acutely angled or difficult to microcatheterize at the SSS than for normal CBVs, we measured 70 angles of confluence on magnetic resonance venography images of 11 normal, and nine AVM patients. There was no statistical difference between normal and AVM patients in the CBV-SSS angles projected in 3-D space (56.2° [SD = 22.4°], and 46.2° [SD = 22.3°], respectively; P > 0.05). Hence, participation of CBVs in drainage of pial AVMs should not confer any added difficulty to their microcatheterization across the SSS, when compared to the acute angles found in normal individuals. This has useful implications for potential choices of strategies requiring endovascular transvenous retrograde approaches to treat AVMs. Clin. Anat. 33:293-299, 2020. © 2019 Wiley Periodicals, Inc.

Computational Network Modeling of Intranidal Hemodynamic Compartmentalization in a Theoretical Three-Dimensional Brain Arteriovenous Malformation.

There are currently no in vivo techniques to accurately study dynamic equilibrium of blood flow within separate regions (compartments) of a large brain arteriovenous malformation (AVM) nidus. A greater understanding of this AVM compartmentalization, even if theoretical, would be useful for optimal planning of endovascular and multimodal AVM therapies. We aimed to develop a biomathematical AVM model for theoretical investigations of intranidal regions of increased mean intravascular pressure (Pmean) and flow representing hemodynamic compartments, upon simulated AVM superselective angiography (SSA). We constructed an AVM model as a theoretical electrical circuit containing four arterial feeders (AF1-AF4) and a three-dimensional nidus of 97 interconnected plexiform and fistulous components. We simulated SSA by increases in Pmean in each AF (with and without occlusion of all other AFs), and then used network analysis to establish resulting increases in Pmean and flow within each nidus vessel. We analyzed shifts in hemodynamic compartments consequent to increasing AF injection pressures. SSA simulated by increases of 10 mm Hg in AF1, AF2, AF3, or AF4 resulted in dissipation of Pmean over 38, 66, 76, or 20% of the nidus, respectively, rising slightly with simultaneous occlusion of other AFs. We qualitatively analyzed shifting intranidal compartments consequent to varying injection pressures by mapping the hemodynamic changes onto the nidus network. Differences in extent of nidus filling upon SSA injections provide theoretical evidence that hemodynamic and angioarchitectural features help establish AVM nidus compartmentalization. This model based on a theoretical AVM will serve as a useful computational tool for further investigations of AVM embolotherapy strategies.

Large-scale ensemble simulations of biomathematical brain arteriovenous malformation models using graphics processing unit computation.

BACKGROUND: Theoretical modeling allows investigations of cerebral arteriovenous malformation (AVM) hemodynamics, but current models are too simple and not clinically representative. We developed a more realistic AVM model based on graphics processing unit (GPU) computing, to replicate highly variable and complex nidus angioarchitectures with vessel counts in the thousands-orders of magnitude greater than current models. METHODS: We constructed a theoretical electrical circuit AVM model with a nidus described by a stochastic block model (SBM) of 57 nodes and an average of 1000 plexiform and fistulous vessels. We sampled and individually simulated 10,000 distinct nidus morphologies from this SBM, constituting an ensemble simulation. We assigned appropriate biophysical values to all model vessels, and known values of mean intravascular pressure (Pmean) to extranidal vessels. We then used network analysis to calculate Pmean and volumetric flow rate within each nidus vessel, and mapped these values onto a graphic representation of the nidus network. We derived an expression for nidus rupture risk and conducted a model parameter sensitivity analysis. RESULTS: Simulations revealed a total intranidal volumetric blood flow ranging from 268 mL/min to 535 mL/min, with an average of 463 mL/min. The maximum percentage rupture risk among all vessels in the nidus ranged from 0% to 60%, with an average of 29%. CONCLUSION: This easy to implement biomathematical AVM model, allowed by parallel data processing using advanced GPU computing, will serve as a useful tool for theoretical investigations of AVM therapies and their hemodynamic sequelae.

Are high lumbar punctures safe? A magnetic resonance imaging morphometric study of the conus medullaris.

A high lumbar puncture (LP) at L2-L3 or above is often necessary to consider on technical grounds, but complications of conus medullaris (CM) damage during high LP are potentially concerning. We hypothesized that a high LP might be safer than previously thought by accounting for movements of the CM upon patient positional changes. We retrospectively reviewed standard normal supine lumbar spine magnetic resonance imaging of 58 patients and used electronic calipers on axial images at the T12-L1, L1-L2, and L2-L3 disc levels to measure the transverse diameter of the CM relative to the size of the dorsal thecal sac space (DTSS) through which a spinal needle could be inserted. On 142 axial images, the means for CM diameters were 8.2, 6.0, and 2.9 mm at the three levels, respectively. We then used known literature mean CM displacement values in the legs flexed and unflexed lateral decubitus position (LDP) to factor in CM shifts to the dependent side. We found that at all three levels, the likely positional shift of the CM would be too small and insufficient to displace the entire CM out of the DTSS. However, if needle placement could be confined to the midsagittal plane, an LP in the unflexed LDP would theoretically be entirely safe at both L1-L2 and L2-L3, and almost so at L2-L3 in the legs flexed LDP. Thus, high LPs at L1-L2 and L2-L3 are in theory likely safer than considered previously, more so in the legs unflexed than in the flexed LDP. Clin. Anat. 32:618-629, 2019. © 2019 Wiley Periodicals, Inc.