Periodic Reporting for period 1 - EndoTheranostics (Multi-sensor Eversion Robot Towards Intelligent Endoscopic Diagnosis and Therapy)
Okres sprawozdawczy: 2024-07-01 do 2025-12-31
Colorectal cancer prevention relies heavily on colonoscopy, yet the procedure can be uncomfortable and technically demanding in patients with complex anatomy. Current tools often struggle to provide stable, high-quality visualisation and to support precise therapeutic actions while the bowel is moving and deforming. EndoTheranostics aims to overcome these limitations by developing a new generation of soft, minimally invasive robotic endoscopy technology that supports both accurate assessment and targeted treatment within a single procedure.
At the core of the project is a soft “everting” robot: a device that advances by gently turning a thin sleeve inside-out (similar to rolling a sock outward). This mechanism is designed to reduce friction against the intestinal wall and to provide a compliant, patient-friendly way to traverse the colon. Building on this locomotion concept, the project is developing a platform that can deliver miniaturised sensing and treatment modules to the area of interest.
At the core of the project is a soft “everting” robot: a device that advances by gently turning a thin sleeve inside-out (similar to rolling a sock outward). This mechanism is designed to reduce friction against the intestinal wall and to provide a compliant, patient-friendly way to traverse the colon. Building on this locomotion concept, the project is developing a platform that can deliver miniaturised sensing and treatment modules to the area of interest.
During the first reporting period, the consortium established the main technical foundations needed for later system integration. Work advanced on selecting and experimentally validating suitable sleeve materials and on developing steering concepts to improve manoeuvrability. In parallel, the project progressed on compact 3D imaging at the device tip, including a calibrated stereo camera capsule with tailored illumination and initial real-time depth reconstruction.
Early prototypes and design concepts were also developed to create a localised working space inside the colon, an inflatable chamber concept intended to stabilise the immediate environment around a lesion so that sensing and treatment can be performed more reliably. Alongside this, the project advanced miniaturised manipulation concepts compatible with tight endoluminal size constraints, including approaches that can temporarily increase stiffness when needed for precision.
A complementary strand focused on safety and workflow support through data and simulation. The project implemented the foundations for a “guardian” approach: collecting multimodal data during representative tasks, developing automated annotation and analysis pipelines, and building teleoperation and simulation tooling. These elements are intended to support objective performance assessment and to enable future shared-autonomy features that assist, rather than replace, the clinician, with safety as a primary constraint. A control architecture has been developed for safe transitions between different levels of autonomy, including control strategies that implement shared surgeon-robot autonomy during task execution based on model predictive control and control barrier functions.
The project also advanced the experimental infrastructure for validation by developing realistic simulators and measurement strategies that reduce reliance on animal or patient testing. Two main simulators were initiated: a colon model based on published anatomical measurements and a complementary abdominal-region model to account for clinically relevant mobility constraints.
Early prototypes and design concepts were also developed to create a localised working space inside the colon, an inflatable chamber concept intended to stabilise the immediate environment around a lesion so that sensing and treatment can be performed more reliably. Alongside this, the project advanced miniaturised manipulation concepts compatible with tight endoluminal size constraints, including approaches that can temporarily increase stiffness when needed for precision.
A complementary strand focused on safety and workflow support through data and simulation. The project implemented the foundations for a “guardian” approach: collecting multimodal data during representative tasks, developing automated annotation and analysis pipelines, and building teleoperation and simulation tooling. These elements are intended to support objective performance assessment and to enable future shared-autonomy features that assist, rather than replace, the clinician, with safety as a primary constraint. A control architecture has been developed for safe transitions between different levels of autonomy, including control strategies that implement shared surgeon-robot autonomy during task execution based on model predictive control and control barrier functions.
The project also advanced the experimental infrastructure for validation by developing realistic simulators and measurement strategies that reduce reliance on animal or patient testing. Two main simulators were initiated: a colon model based on published anatomical measurements and a complementary abdominal-region model to account for clinically relevant mobility constraints.
EndoTheranostics combines several advances into a single clinically motivated pathway. It introduces a soft, low-friction locomotion approach for colon traversal and couples it with compact 3D perception designed for reliable intra-procedural assessment. It also adds a dedicated concept for stabilising the local environment at the target site, enabling more controlled sensing and treatment actions in a moving, deformable organ.
In addition, the project is building a safety-oriented development framework that links data collection, automated analysis, teleoperation, and simulation to support systematic benchmarking and the progressive introduction of clinician-assistive autonomy. Together, these elements establish a structured route from component innovation to integrated validation, with the long-term aim of improving examination completeness and tolerance, strengthening decision-making through better sensing, and increasing the reliability of targeted endoscopic therapy.
The eversion robot findings in WP1 advance the state of the art not only in the context of colonoscopy but also across eversion robotics more broadly. The proposed improvements to the eversion robot body signifi cantly enhance navigational capability and apply to a wide range of environment-guided eversion robot applications, including other endoluminal procedures, pipe inspection, and search-and-rescue operations. This work has the potential to establish a new generation of highly maneuverable eversion robots.
The results of WP5 advance the state of the art of concentric tube robots by developing techniques that integrate a physics-based model and learned corrections implemented by mixture density networks for accurate kinematic modelling.
In addition, the project is building a safety-oriented development framework that links data collection, automated analysis, teleoperation, and simulation to support systematic benchmarking and the progressive introduction of clinician-assistive autonomy. Together, these elements establish a structured route from component innovation to integrated validation, with the long-term aim of improving examination completeness and tolerance, strengthening decision-making through better sensing, and increasing the reliability of targeted endoscopic therapy.
The eversion robot findings in WP1 advance the state of the art not only in the context of colonoscopy but also across eversion robotics more broadly. The proposed improvements to the eversion robot body signifi cantly enhance navigational capability and apply to a wide range of environment-guided eversion robot applications, including other endoluminal procedures, pipe inspection, and search-and-rescue operations. This work has the potential to establish a new generation of highly maneuverable eversion robots.
The results of WP5 advance the state of the art of concentric tube robots by developing techniques that integrate a physics-based model and learned corrections implemented by mixture density networks for accurate kinematic modelling.