Periodic Reporting for period 1 - ROBIOPSY (AI and Robotics for Prostate Biopsy)
Período documentado: 2023-05-01 hasta 2024-04-30
The ROBIOPSY project aims at evolving the robotic device for prostate cancer (PCa) biopsy demonstrated during the ERC-PoC PROST into a product prototype that will be ready for clinical trials. The focus of the project will be two-fold: the engineerization of the PROST prototype, and the accurate analysis of the business case of prostate cancer diagnosis and its potential extension to focal therapy. We will correct the shortcomings of the pre-clinical tests, meet the strict medical certification regulations, and address the Health Economics implications of our solution. ROBIOPSY will achieve better performance than current competitors because it solves the two main causes of PCa diagnostic error: uncertain target identification, and inaccurate needle positioning, as demonstrated in the pre-clinical tests. A novel image fusion method will permit to significantly reduce the target uncertainty, while the robotic device will zero the positioning error. The ROBIOPSY prototype will consist of a single cart holding the robotic needle positioner, and housing all the electronic components. The electronics will be divided into two subsystems: interface and data processing, and safety critical controls. This will reduce the operational risks and will simplify the medical certification. A multi-centric study is ongoing to collect biopsy images to train the Machine Learning algorithms for prostate segmentation and lesion identification. In parallel to the technical development, we will analyze in depth the business case of prostate biopsy in selected European Countries, we will analyze the time/cost reduction due to the new device and will analyze the feasibility and economic benefits of using the ROBIOPSY prototype in the focal treatment of PCa, to be ready for the expected evolution of prostate cancer treatments.
During the first periodic period (PrM1-PrM12), we improved and consolidated the prototype developed during the ERC-PoC project PROST in order to have a solid ground on which defining accurate clinical requirements to be translate into technical specifications. In particular we focused on
• [SW] identification of the prostate boundaries and biopsy targets in the MRI and US images. The PROST-Net Deep Convolutional Neural Network has been further trained with new data coming from hospitals involved in the clinical trials started during the PROST project
• [SW] fusion of MRI and US images: novel NNs have been tested and compared to achieve a better fusion between MRI and US 3D models. This task is critical since lesions are not visible in US scans and so the mapping of lesions detected in MRI into the US space is mandatory to reduce the positioning error. Particular attention has been paid to the computational time since the fusion process must be executed at run-time and so no mayor delays can be accepted
• [SW] refactoring of the software related to the current prototype to remove dependencies of third-party libraries and be more user-friendly
• [HW] partially re-design of the robotic targeting and needle guide devices for inserting the needle (Needle Targeting subsystem and Needle Guide subsystem) and of the US probe motions (US subsystem), to move the probe back and forth and rotate it
• [HW] computation of the direct and inverse kinematics of the prototype. This step is necessary to plan the optimal needle trajectories to reach the lesions detected in the MRI 3D model and “mapped” into the US 3D model
• [HW] design the control architecture to move the robot according to the workflow of the procedure and the commands by the medical personnel
• [HW] design and development of the passive positioner arm to move the biopsy module (Needle Targeting subsystem + Needle Guide subsystem + US subsystem) close to the patient.
• [SW] identification of the prostate boundaries and biopsy targets in the MRI and US images. The PROST-Net Deep Convolutional Neural Network has been further trained with new data coming from hospitals involved in the clinical trials started during the PROST project
• [SW] fusion of MRI and US images: novel NNs have been tested and compared to achieve a better fusion between MRI and US 3D models. This task is critical since lesions are not visible in US scans and so the mapping of lesions detected in MRI into the US space is mandatory to reduce the positioning error. Particular attention has been paid to the computational time since the fusion process must be executed at run-time and so no mayor delays can be accepted
• [SW] refactoring of the software related to the current prototype to remove dependencies of third-party libraries and be more user-friendly
• [HW] partially re-design of the robotic targeting and needle guide devices for inserting the needle (Needle Targeting subsystem and Needle Guide subsystem) and of the US probe motions (US subsystem), to move the probe back and forth and rotate it
• [HW] computation of the direct and inverse kinematics of the prototype. This step is necessary to plan the optimal needle trajectories to reach the lesions detected in the MRI 3D model and “mapped” into the US 3D model
• [HW] design the control architecture to move the robot according to the workflow of the procedure and the commands by the medical personnel
• [HW] design and development of the passive positioner arm to move the biopsy module (Needle Targeting subsystem + Needle Guide subsystem + US subsystem) close to the patient.
Complete, ergonomic device configuration with all of the components involved in the prostate biopsy procedure. The ROBIOPSY prototype will consists of the biopsy device, the graphical data and command interface of the ultrasound imager, and the processing and editing unit for the biopsy procedure. The system includes these elements in a novel ergonomic set-up not available on the market. We are working on the seamless integration of the different SW/HW we already have and on the re-design / improvement of other modules (e.g. GUI, segmentation algorithms with further US/MRI images). A new patent has be filed during the first reporting period and others will be filed soon to consolidate the Needleye Robotics position on the diagnostic and therapy prostate cancer domain.