Daehan Wi

1. Wearable tactile interfaces are increasingly explored as alternative or complementary channels for conveying information when vision is unavailable, overloaded, or impractical (Adame et al., 201...

Background and Objectives Clinical AI/ML models increasingly inform healthcare and decision support, yet credibility appraisal is often ad hoc—especially for non-deterministic, lifecycle-dependent models. A risk-informed, stage-based diagnostic pathway is proposed to plan and audit credibility evidence in in silico clinical trial (ISCT) settings and related evaluation pipelines, without introducing scores or maturity levels. Methods Question of interest (QOI), context of use (COU), and model risk assessment are operationalized by adapting credibility principles from computational modeling guidance (e.g., ASME V&V 40; FDA credibility guidance). Evidence is organized into five stages: data quality assessment, internal validation, external validation, scenario-based evaluation of decision relevance and robustness, and post-deployment monitoring and change control. Explainability, dataset shift/drift susceptibility, and update intent are treated as cross-cutting evidence targets, supported by a minimal documentation set. Results Two retrospective, publication-based walkthroughs were conducted. For an acute ischemic stroke outcome prediction study, gaps were identified in external validation of newly developed models, subgroup calibration evidence, and lifecycle monitoring intent. For a dynamic kidney failure prediction model with prospective external validation, residual implementation-facing gaps persisted despite decision-curve analysis, particularly COU-linked operating thresholds, quantitative calibration/uncertainty at clinically relevant horizons, and monitoring/update planning. Conclusions The pathway offers a practical, risk-tiered approach to plan and audit credibility evidence, shifting attention beyond discrimination toward COU-linked reliability, clinical utility, and lifecycle readiness.

A Nitinol rod-driven discrete continuum robot with two sections and eight units was developed to support clinicians in performing transoral surgery. The robot measures 120 mm in length, with each unit having a diameter of 15 mm and a height of 20 mm. The distal and proximal sections are designed to bend independently, each with two degrees of freedom (DOF) actuated by four Nitinol rods. To validate the independent controllability of the two sections, two-dimensional bending tests and ANSYS simulations were conducted. For the assessment of clinical feasibility, head and neck CT images from ten patients were manually segmented to reconstruct three-dimensional oral cavity models. Ten fictitious reference passages were generated from the lips to the oropharynx, and planar path-planning simulations were performed using these passages. Verification experiments were carried out on three reference passages employing experimentally derived inverse kinematics. The simulation results demonstrated an average reference path-following error within a root mean square (RMS) of 1.9705 mm at maximum insertion length. Experimental path-planning results showed average absolute angular differences of 5.6 degrees in the distal section and 4.1 degrees in the proximal section when compared with the simulations.

Traditional surgical interventions for frontal sinus fractures necessitate a cut on the forehead skin, and extant closed reduction techniques aimed at enhancing accessibility continue to grapple with secure tool fixation, stable bone elevation, and screw breakage risk. To address these challenges and augment surgical efficiency, this study introduces novel surgical devices. Design parameters for models with spiral or L‐shaped tips are established, considering practical medical requirements and constraints, and subsequently validated through finite element method numerical simulations using commercial software, Ansys. Four spiral‐type prototypes are constructed, and three scenarios for each prototype, varying in projection distance from the device handle to the bone‐device contact point, are examined via nonlinear simulation analyses. For the L‐shaped type, three prototypes are developed, and static analyses are conducted for four scenarios per prototype, differing in traction force locations, based on another simulation result concerning moments of inertia calculation with a force boundary condition unlike pressure. Maximum stress results under a specific force are analyzed, and the maximum permissible force is determined under the most unfavorable force application condition. Simulation outcomes indicate that the spiral type offers greater applicability with less force to lift multiple bones, while the L‐shaped type is more suitable under bone hardening conditions.

The authors previously developed an online thin film buckling configuration determination method for a mini basket type mapping catheter prototype, which incorporates eight thin film sensor strips. In the prior study, no external force was applied to the thin film, and only axial displacement was adjusted to induce buckling in the thin film. Extending this prior work, a preliminary methodological validation is conducted for an online configuration determination method of thin film buckling under a point contact force. The overall thin film configuration determination problem is formulated as a constrained optimization problem, involving five variables and five equality constraint functions. Before developing an actual online optimization solver, preliminary numerical calculations, Ansys simulations, and experiments are performed to verify the proposed problem formulation. The comparison between the numerical precalculations, Ansys simulations, and experimental results demonstrated that the proposed problem formulation is consistent with Ansys simulations and experimental outcomes. This indicates that the proposed formulation is capable of calculating accurate solutions using appropriate optimization methodologies.

To address the possibility of assisting the clinicians to perform transoral surgery, a Nitinol rod driven discrete continuum robot with 3 sections and 9 units is developed by authors. The overall length of the discrete continuum robot mechanism is 120 mm and the unit diameter is 15 mm with 20 mm height. The distal and proximal sections are designed to be bendable independently in 2 degree of freedom (DOF) with four Nitinol rod each. To investigate the Nitinol rod displacement change to section bending angle relationship, 2D section bending tests and ANSYS simulation are performed. Moreover, to address the design feasibility of the discrete continuum robot in clinical environment, 11 patients' head and neck CT images are manually segmented to render 11 3D oral cavity and made a fictituous 11 reference passage from lips to oropharynx. Using the 11 fictituous reference passage, planar path planning simulations are performed with the designed 2 sections 8 units discrete continuum robot. To verify the simulations, path following tests are conducted with 3 reference passages. The simulation results show that the average reference path following error is within RMS (root mean square) 1.9705 mm in maximum introducing length and the test results presents that the maximum bending angle difference between the simulation and test is 9.6 degree.

A vibrational resonant cantilever beam device is a type of tactile vision substitution device for the visually impaired that has the potential to achieve a high resolution in a small space. In order to realize a device of this type, it is necessary to accurately model the resonant behavior of the mesoscale (0.1–1.0 mm) metal cantilever beams of which the device is composed. Specifically, the natural frequencies and damping ratios of these beams must be analytically modeled in order to design beam dimensions for the device. In this paper, the resonant frequencies and damping ratios of a set of A2 tool-steel mesoscale cantilever beams are obtained using three different methods: calculation based on a mass–spring–damper model of the beams; a forced response experiment; and a free response experiment. The results are compared, and the size effect on stiffness and elastic modulus is investigated. Based on the modified couple stress theory, the length scale parameter is calculated. The damping ratio difference depending on the dimension of material in the mesoscale is also investigated.

Resonant vibrotactile microbeams use the concept of resonance to excite the vibration of cantilever beams, which correspond to pixels of an image. The primary benefit of this type of tactile display is its potential for high resolution. This paper presents the concept of the proposed system and human skin contact experiments to explore user perception challenges related to beam vibration during skin contact. The human skin contact experiments can be described in five phases: dried skin contact to metal beam tips, wet and soaped skin contact to metal beam tips, skin contact with a constraint, normal force measurement, and skin contact to the tips of silicone rubber beams attached to metal beam tips. Experimental results are analyzed to determine in what cases of skin contact the beams stop vibrating. It is found that the addition of silicone rubber beams allows the primary metal beams to continue vibrating while in contact with skin. Thus, the vibration response of a metal beam with silicone rubber beams is investigated for the better understanding of the effect of silicone rubber beams on the metal beam vibration.