Fluids

In this study, we proposed and evaluated a new dielectrophoresis (DEP)-based device for the high-throughput, non-invasive separation of cells with similar physical properties. Although DEP is an established technology for separating biological particles based on their dielectric properties, the accurate separation of populations of similar cell types—such as normal cells and cancer cells of the same species—remains a significant technical challenge. In this study, separation experiments using a prototype device were conducted with normal (MCF10A) and tumor (MDA-MB-231) human mammary cells. The experiments systematically analyzed the effects of three key parameters—A.C. frequency (f), flow rate (Q), and solution conductivity (${\sigma }_f$)—on the cell separation ratio ($\Phi$). The experimental results identified optimal operating conditions at f = 57.6 kHz, Q = 5.0 mL/h, and ${\sigma }_f$ = 40 mS/m, achieving a maximum separation ratio of $\Phi$ = 72.1%. These findings demonstrate that the potential of this device to perform high-speed processing without the need for fluorescent labeling or invasive pretreatment. This study provides a fundamental framework for optimizing DEP parameters and outlines a pathway toward more efficient clinical applications and the isolation of rare cell populations.

Patient-specific computational fluid dynamics (CFD) is increasingly used to investigate intracranial aneurysm hemodynamics. However, differences in modeling practices can affect simulated flow fields and derived metrics, complicating interpretation and comparison. We conducted a review to characterize methodological variability and reporting across the CFD workflow. We searched PubMed, Web of Science, and Scopus from 1 January 2020 to 6 November 2025. Patient-specific saccular intracranial aneurysm CFD studies were included; idealized geometries, treatment-device simulations, and methodological comparisons were excluded. Domains included geometry reconstruction, meshing, boundary conditions, numerical implementation, temporal settings, mesh independence testing, and external validation. Among 112 studies, geometry source (92.9%) and inlet boundary conditions (91.1%) were commonly reported, whereas solver type (25.0%), minimum element count (33.9%), maximum element size (22.3%), first inflation-layer height (6.3%), and explicit mesh-independence testing (8.0%) were reported less frequently. Validation was not described in 80.4% of studies, and 8.0% reported explicit methods. Modeling patterns were broadly similar when studies were stratified by scientific purpose, while reporting gaps persisted. These studies demonstrate substantial methodological variability, incomplete reporting, and infrequent verification and validation, limiting independent appraisal and cross-study comparison of CFD-derived hemodynamic findings.

Force measurement is a fundamental diagnostic technique in experimental fluid mechanics. However, conventional instrumentation, such as mechanical balances and pressure taps, is often intrusive, geometry-dependent, and mechanically constrained. This study proposes a non-contact, image-driven force-estimation framework wherein convolutional neural networks (CNNs) infer the instantaneous lift and drag coefficients of a circular cylinder directly from wake streakline flow-visualization images. A comprehensive hybrid numerical–experimental dataset was established across laminar and turbulent regimes. Navier–Stokes simulations (DNS at $Re=150$; LES at $Re=500$ and 3900) were post-processed into synchronized streakline–force pairs, complemented by water-tunnel dye-visualization experiments at $Re=500$, 8000, and $\mathrm{12,000}$. Five CNN architectures, including lightweight custom networks and vision backbones (AlexNet, ResNet18), were trained and validated. For numerical datasets, all models show high prediction accuracy in lift and drag coefficients. In simulation-to-experiment transfer at $Re=500$, the models capture the dominant vortex-shedding Strouhal number within $12.0\%$ ($St=0.233$ vs. $0.208$), while CNNModel_v5 predicts the mean drag coefficient within $0.35\%$. On experimental datasets, ResNet18 achieved the best predictive performance among the models tested under random splitting. These results indicate that wake flow-visualization images contain rich physical representations for non-intrusive fluid force estimation.