Abstract: Robots for biomedical applications, including surgical assist systems and experimental automation platforms, demand higher precision, reliability, and robustness under uncertainty than general-purpose robots. To enable their autonomy, the integration of Physical AI is being explored. However, extensive real-world experimentation for AI training is often impractical. Digital twin technologies, combined with precision robotic systems, are being investigated to enable efficient and safe learning.
Abstract: This presentation highlights ongoing research at KTH Royal Institute of Technology focused on the use of sensors and measurement systems to enhance the accuracy and performance of machines and robotic systems. The work includes the development of virtual environments that allow for testing digital versions of sensors and systems prior to physical implementation. This approach accelerates innovation and improves system design, particularly in advanced manufacturing and biomedical applications.
Abstract: Robotic surgery was introduced at Karolinska already 2001 by pioneer prof Peter Wiklund, then performing prostatectomies using the early 3-armed da Vinci system; now the single-port system. Multiple clinical applications in upper gastric, lower abdominal and pelvic surgery were later developed using leading commercial systems, and an orthopaedic robotic system is currently being introduced. Spinal scoliosis surgerfy is employed using navigated screw fixation. There is currently a sincere interest to develop thoracic and orthopaedic robotic surgery.
Abstract: Long bone fracture surgery presents technical challenges, including the difficulty of achieving precise rotational alignment with 2D fluoroscopic imaging, the strenuous manual effort required for traction, high manpower demands, and significant radiation exposure for medical staff. To overcome these issues, we developed a Stewart platform–based robotic system with six degrees of freedom and a detachable driver pack for sterilization, integrated with a dedicated navigation system. The robot delivers precise movements and sufficient traction, while the navigation platform provides real-time bone tracking and facilitates target positioning. Under surgeon supervision, the system enables automatic fracture reduction. In cadaver studies with six tibiae, robot-assisted automatic reduction showed superior accuracy and reduced operative time compared to both manual robotic and traditional methods. For regulatory approval, we conducted tests for electromechanical safety, electromagnetic compatibility, biocompatibility, and system performance. After obtaining regulatory approval, insurance reimbursement is necessary to enable use in real patients and generate revenue, so we are preparing for this next step.
Abstract: Correcting complex deformities remains one of the greatest technical challenges in orthopaedic surgery. Precise alignment, stable fixation, and reliable bone healing are critical for long-term function, yet existing techniques leave room for variability. Current practice relies heavily on surgeon experience, intraoperative imaging and empirical knowledge, but these methods can be limited in accuracy, time efficiency, and adaptability to complex anatomical variations. This lecture will explore both current challenges and possibilities offered by emerging technologies. 3D mirrored drill guides represent one step forward, enabling patient-specific solutions that translate preoperative planning into more accurate execution. Looking further ahead, robot-assisted surgery may allow for unprecedented precision, while virtual reality can be used for preoperative rehearsal, intraoperative guidance, and surgical education. Fixation devices with integrated sensor technology have the potential to deliver continuous data on bone healing and function, combining biological and mechanical feedback into personalized treatment strategies. In summary, emerging technologies have the potential to move deformity correction toward a future that is more predictable, data-driven, and patient-centered.
Abstract: Autonomous manipulation of soft tissues is a key challenge in surgical robotics due to tissue deformability and safety requirements. We present a vision-based modeling framework that constructs a digital twin of the surgical scene and enables real-time tracking of tissue deformation. Combined with reinforcement learning-based countertraction control, our approach supports safer and more efficient laparoscopic procedures.
Abstract: Humans and AI-powered technologies increasingly act together as hybrid agents, jointly producing situationally and sequentially relevant action. Drawing on Feminist Human-Robot Interaction, Human Computer Interaction, and Conversational User Interaction this talk surfaces how qualitative methods can be used to uncover the practices that this human-AI collaboration entails, supporting the technical design of systems that are responsive to the complexities of the physical, social and institutional settings in which they are deployed.
Abstract: In this talk, we present a full Eulerian method for solving fluid-structure interaction problems, which is particularly well-suited for handling medical image data in voxel formats. Using the method, direct numerical simulations are performed to investigate blood flows containing both red blood cells (RBCs) and platelets in small vessels. It is well known that RBCs exhibit an axial migration in capillary vessels, leading to the formation of a so-called cell-free layer near the vessel wall. Platelets tend to migrate toward this cell-free layer, and these processes play a crucial role in platelet adhesion to the vessel wall, which marks the initial stage of thrombosis. In this talk, we present a modeling framework for these phenomena and discuss platelet dynamics in the presence of RBCs.
Abstract: The enhanced risk of flow-induced thrombosis remain a challenge when designing new medical devices such as blood pumps, membrane lungs and cannula used in extracorporeal life and organ support. To avoid flow-induced complications, the level of stress would need to be kept close to physiological levels. However, this is inherently difficult due to the operating conditions and stress dynamics associated with the flow structures that develops within these devices. This talk will focus on these medical devices, how the link between thrombogenicity and flow is currently modelled and assessed, and a future perspective on improving our understanding of the processes leading to thrombus formation and modeling thereof.
Abstract: Endothelial cells (ECs) cover the inner surface of blood vessels and are exposed to mechanical factors such as shear stress caused by blood flow and stretching tension induced by blood pressure. ECs sense changes in mechanical forces (mechanosensing), convert them into intracellular biochemical signals, and transmit them to the interior of the cell (mechanotransduction), triggering cellular responses accompanied by changes in cell morphology, function, and gene expression. To investigate the role that changes in cellular function mediated by mechanosensing play in the development and progression of cerebral aneurysms, a vascular disease involving blood flow factors, we performed computational fluid dynamics analysis on clinical cases to develop a diagnostic tool that can be used to predict rupture of cerebral aneurysms and recurrence after coil embolization.
Abstract: Predicting the behaviour of fluids in which microscopic structure matters remains an open challenge: fine-scale physics governs macroscopic observables, yet a fully resolved description is computationally prohibitive. I will outline a framework that links particle-scale and continuum descriptions through neural operators, providing a data-driven “bridge”. I will present the general strategy which is based on embedding discrete simulations into a reduced latent space, learning scale-transfer maps, and coupling the resulting surrogate with conventional CFD. Then I will specialise it to blood clot formation, a test-bed where multiscale interactions are unavoidable.
Abstract: Patient-specific simulations are increasingly expanding toward clinical applications. This talk highlights recent trends, with a particular focus on two aspects of cerebral circulation. The first concerns patient-specific blood flow simulations that integrate large-scale, multi-modal data with AI techniques. Such approaches are beginning to yield novel insights previously unattainable, bridging information from medical images to hemodynamic simulations. At the same time, despite significant advances in imaging technologies, limitations remain in elucidating underlying mechanisms. To address these gaps, mathematical modeling plays a crucial role in linking unresolved factors—such as collateral flow—that current imaging methods alone cannot capture. Together, these approaches underscore the growing potential of patient-specific simulations to deepen our understanding of cerebral circulation and ultimately inform clinical decision-making.