Title of the Talk
Credible Digital Twin Architectures for Regulated Therapy and Imaging Devices: Verification, Validation, and Bounded Personalization
Abstract
Digital twins, computational representations of medical devices, clinical environments, and patient physiology, are increasingly recognized as a means to augment traditional verification and validation (V&V) activities. Regulatory guidance such as ASME V&V 40-2018 and the U.S. FDA’s credibility framework for computational modeling and simulation explicitly define expectations for context of use, model risk, and decision consequence, yet practical system architectures that operationalize these principles across therapy delivery and medical imaging workflows remain limited. This talk presents a credibility-first digital twin architecture designed to support regulated medical device development across therapy delivery systems and imaging pipelines, integrating device behavior models, physics-based environment models, and physiology-aware patient representations. The architecture emphasizes traceability from ISO 14971 risk controls to simulation evidence, explicit verification and validation planning, and uncertainty quantification aligned with decision impact. Core platform components include a versioned twin registry, scenario orchestration engine, evidence capture layer for auditability, and controlled interfaces compatible with continuous integration workflows such as automated regression simulations and parameter sweeps. Validation strategies are discussed across multiple evidence sources commonly used in regulated development, including bench measurements, phantom studies, animal data, and retrospective clinical datasets, with stratification over device settings, anatomical variability, motion, and workflow conditions. The approach also enables controlled personalization, where patient-specific parameters are estimated only within predefined safety envelopes and verified algorithmic bounds, rather than open-ended adaptation. By grounding digital twin use in established regulatory standards and reproducible engineering workflows, this architecture demonstrates how simulation can safely expand verification coverage, support risk management, and accelerate development without compromising patient safety or regulatory credibility.
Brief Profile
Shrikant Chikhalkar is a senior engineering leader with over two decades of experience designing and delivering high-performance software for regulated healthcare, neuromodulation, and medical imaging systems. His work spans safety-critical medical devices, PET/MR reconstruction platforms, and cloud-native, AI-enabled software architectures developed under IEC 62304 and FDA 510(k) regulatory frameworks. Currently a Senior Staff Software Engineer, Shrikant leads backend and platform initiatives supporting clinician and patient applications, including consent management systems, dynamic configuration platforms, secure telemetry pipelines, and cross-platform MAUI architectures spanning iOS, Android, and Windows. He has played a key role in enabling global neuromodulation therapy launches across 27 European countries while ensuring GDPR, HIPAA, and regional data-sovereignty compliance. His work includes regulated SDLC execution, microservice modernization, DevOps automation, and the application of generative AI to improve engineering productivity and regulatory documentation workflows. Previously, Shrikant spent a decade at GE Healthcare as a Staff Software Architect, where he led PET/MR reconstruction and acquisition software for clinical imaging systems. He was responsible for productizing advanced MR attenuation correction (MRAC) algorithms, motion-corrected PET/MR reconstruction pipelines, fault-tolerant imaging systems, and GPU-accelerated reconstruction frameworks that achieved up to 200x performance gains. His contributions supported FDA 510(k) submissions and were validated through phantom and clinical image-quality studies. Earlier in his career, Shrikant held technical leadership and engineering roles at IGATE, Persistent Systems, Accenture, Manchitra, and a computational research lab, working on legacy modernization, distributed graph mining, seismic simulation, GIS rendering engines, and large-scale HPC systems using MPI, OpenMP, and GPU computing. Shrikant holds a Bachelors degree in Computer Science from the University of Pune and has completed graduate-level coursework in artificial intelligence, deep learning, cloud computing, and algorithm analysis at the University of Wisconsin-Milwaukee. He has earned certifications in Generative AI with Large Language Models, SAFe, and project management, and has submitted intellectual property related to AI-assisted code analysis. He has also presented at international conferences, including ISMRM, and has served as an architecture lead and mentor across global engineering teams.