The Schedule

Welcome address by his Excellency the Ambassador of Japan to Sweden

Masaki Noke

Importance of Japan-Sweden academic exchange and the role of JSPS

Prof. Kuroda Kazuyuki, JSPS Stockholm Office  

Next-generation surgical support systems which transcend size, distance, and time

Prof. Mamoru Mitsuishi, UTokyo

Abstract: The author's group has developed microsurgical systems for neurosurgery, ophthalmology, and pediatric surgery. These are systems that transcend the size barrier. Analysis in micro size has also been performed in osteotomies in joint replacement surgery to control the increase in bone resection temperature and reduce force. Furthermore, remote surgical support systems transcend the limitations of distance, but there is the problem of time delay, which is crucial to solve. Additionally, the introduction of AI, including machine learning, is to utilize knowledge beyond time, and a system utilizing this knowledge in surgery will be constructed.

Dense FixMatch: a Simple Semi-supervised Learning Method for Pixel-wise Prediction Tasks 

Prof. Atsuto Maki, KTH

Abstract: We discuss equivariant semi-supervised deep learning on partially labeled datasets for semantic image segmentation. Specifically, we introduce Dense FixMatch as an extension of the popular FixMatch method which combines pseudo-labeling and consistency regularization via strong data augmentation. By including an extra matching operation for geometric transformations, it allows us to use the full strength of the original pipeline. We evaluated it on two commonly used computer vision datasets with different percentages of labeled data. Dense FixMatch outperforms supervised baselines, with the most significant improvements observed when fewer labeled samples are available. We suggest that it is a promising direction when limited labeled data is available, including possible applications to medical image diagnosis..

Robotic Imitation Learning for Biomedical Applications 

Dr. Kento Kawaharazuka, UTokyo

Abstract: Imitation learning is being extensively utilized as a highly valuable tool for various motion learning of robots. This presentation will discuss how the tasks in fundamentals of laparoscopic surgery are solved using imitation learning. In our journey towards tasks requiring delicate manipulations like those in biomedical applications, we have conducted various trials in imitation learning, including simple manipulation by mobile manipulators, visual servoing by low-rigidity robots, and whole-body task realization by life-sized humanoids. I will reflect on the technological advancements in imitation learning up to the present day.

Robotic treatment of long bone fractures

Prof. Hans E. Berg, Department of Trauma and Orthopaedic surgery, Karolinska University Hospital- KTH

Abstract: Robot-assisted orthopaedic surgery of the spine, pelvis and hip joint has developed rapidly during the last decades. Automated screw placement and joint implant positioning using robots has shown great promise when compared to freehand surgery. Fractures in the long bones have received less attention. This might be related to the challenge of repositioning multiple bone fragments in addition to fixating the necessary nails, plates, and screws. The presentation will focus on existing and emerging technologies in robotic fracture surgery of the long bones of the lower limb and specifically the repositioning tools and the various fixation methods.

Enhancing Robotic Surgery AI Training with Photorealistic Synthetic Data from VR Simulation

Assist. Prof. Saul Heredia, UTokyo

Abstract: In the realm of robotic surgery, training artificial intelligence (AI) models necessitates vast amounts of annotated data, imposing significant manual labor and time constraints. This talk explores the utilization of photorealistic synthetic data generated via virtual reality (VR) simulation to alleviate this burden. By leveraging VR technology, we can produce highly realistic surgical scenarios, providing diverse and annotated datasets for AI training. This innovative approach aims to enhance the efficiency of AI model training for robotic surgery applications, facilitating robot autonomy while minimizing the labor-intensive task of manual data annotation.

Enabling technologies for small-scale laboratory automation

Associate Professor Christian Smith, KTH

Abstract: An emerging trend in industrial automation is to move robots into task domains and work places that still require human skills, and let a robot perform simpler tasks alongside skilled human workers. Examples of this include many tasks in smaller research laboratories that are simple and repetitive, but still require significant effort and time from highly qualified staff. Some tasks require minor attending at regular intervals, perhaps once every few hours, making them expensive to run outside of office hours. This leads to very inefficient use of both staff resources and facilities. By applying small scale automation to these tasks, resource usage could be significantly improved, as repetitive tasks could run autonomously while the staff can focus on more qualified tasks. However, a common trait for many of these tasks is that there may be frequent variation of the details of the task operations, requiring frequent alterations of workflows and/or introduction of novel tasks, leading to a large overhead for programming the robot. In this talk, we will discuss how to automate different lab tasks with minimal human effort.

(Semi) Autonomous Robots for Medicine, Space, Science, and Nuclear

Dr. Murilo M. Marinho, University of Manchester- UTokyo

Abstract: Robotic assistance has the potential to enable breakthroughs in healthcare, scientific exploration on Earth and beyond, and nuclear decommissioning. Although at first glance these applications might be seemingly disjoint, all these fields can benefit from cleverly designed model-based robotic control strategies that can be customized to provide accurate motion with guaranteed safety. This will briefly describe these robot control strategies, their applications, case studies, and their ongoing combination with data-driven strategies and empower them to perform completely automated – and complex – tasks in challenging scenarios.

Reliability and performance of machines and robotics systems

Prof. Andreas Archenti, KTH

Abstract: The reliability and performance of machines and robotic systems are crucial, particularly in safety-critical and harsh environments such as autonomous robotic surgery and industrial robotics with contact applications. Limited understanding of system physical characteristics, including thermo-kinematic, quasi-static, and dynamic stiffness, as well as their spatial and temporal variations, poses a barrier to autonomous application. The presentation will showcase results from past and ongoing projects focusing on identifying sources of variation and will exemplify methods and tools for calibration.

Ultrasound for biomedical applications 

Prof. Matilda Larsson, KTH

Abstract: Ultrasound is one of the most widely used imaging modalities in healthcare today because it is a safe, non-invasive, real-time and cost-effective imaging modality. Advances in ultrasound technology have drastically changed what can be obtained from an ultrasound examination in terms of tissue constitutive properties. This presentation will describe our research on quantitative ultrasound methods that can be used, for example, to assess whether an atherosclerotic plaque is vulnerable to cause a stroke or to improve the diagnosis of maternal pelvic floor muscle birth injuries. The methods are being developed in programmable ultrasound systems and tested in advanced, patient-specific models.

Holographic augmentation 

Jonny Gustafsson, KTH 

Abstract: The group of Industrial Metrology and Optics (now Precision Engineering) has been involved in three-dimensional imaging techniques since the 1960s when research in holography for measurement purposes was started. The main research results were made in the field of holographic and other optical techniques for measurement of displacement, deformation, vibration, and shape. One of the few remaining holographic laboratories that can produce high-quality holograms are within the group. This research has led to the development of a unique 3D display that does not require either glasses or other accessories to experience virtual objects in full 3D of high optical quality which floats freely in space. The scope of this display is multidisciplinary and covers everything from interactive product design to the modeling of 3D environments and simulation of surgical procedures in medical education. My talk will describe some possible applications of this display with an emphasis on the medical field.

Previous exchange experiences within the KTH and University of Tokyo cooperation

Monica Gonzalez (KTH), Theodoros Laspas (KTH), Hung-Ching Lin (UTokyo), Elia Martinelli (KTH)

Open discussion: Collaboration, expectations and future work 

Moderator: Theodoros Laspas, KTH