Jurdana, Janko (University of Zagreb, Faculty of Mechanical Engineering and Nava), Markulin, Matija (University of Zagreb, Faculty of Mechanical Engineering and Nava), Šiktar, Luka (University of Zagreb, Faculty of Mechanical Engineering and Nava), Matijević, Luka (University of Zagreb, Faculty of Mechanical Engineering and Nava), Šekoranja, Bojan (University of Zagreb, Faculty of Mechanical Engineering and Nava), Šuligoj, Filip (University of Zagreb, Faculty of Mechanical Engineering and Nava), Švaco, Marko (University of Zagreb, Faculty of Mechanical Engineering and Nava)

Robotically Navigated Prostate Biopsy

Scheduled for presentation during the Regular Session "Surgical Robots" (WeAT4), Wednesday, July 16, 2025, 11:40−12:00, Room 108

Joint 10th IFAC Symposium on Mechatronic Systems and 14th Symposium on Robotics, July 15-18, 2025, Paris, France

This information is tentative and subject to change. Compiled on July 16, 2025

Abstract

This paper presents the scientific project PRONOBIS, which proposes the use of medical robots for prostate biopsy procedures. Robot-guided prostate biopsy imposes a number of challenges. Firstly, patient localization with respect to preoperative images and current environment is done. Once the patient is localized, the robot guides the biopsy needle towards specified targets. Using ultrasound data, an analysis of procedure accuracy is performed with respect to ground truth data. Patient localization is challenging due to factors like patient movement, organ movement/deformation, equipment imperfections, and gross errors. To address these issues, PRONOBIS adjusts robotic system parameters in real time using advanced machine learning algorithms and continuous patient registration, enabling precise prostate gland sweeps. Before the sweep, a surgeon inserts the probe. The robot then rotates the probe around its axis to acquire ultrasound images for 3D model generation. The ultrasound image analysis process commences with prostate segmentation and the construction of a detailed 3D prostate model. We introduce a novel medical image annotation platform, MedAP, which employs the Segment Anything Model (SAM). MedAP is utilized to generate a comprehensive segmentation training dataset for automated segmentation algorithm MicroSegNet. MedAP uses SAM’s flexible prompting capabilities to enhance prostate segmentation from micro-ultrasound scans. Although micro-ultrasound images are inherently grainy and may pose challenges to SAM’s performance, MedAP addresses this limitation by incorporating a manual segmentation mode. This additional feature allows for precise adjustments and refinements, ensuring accurate segmentation results even in challenging cases.

Registration is performed on the 3D model with respect to preoperative images. The registered model helps surgeons identify suspect regions and target them. The system can also determine standardized biopsy points. Robotics offers another major benefit: precise navigation to targets, guiding the biopsy needle. This allows less experienced clinicians to perform prostate biopsies without requiring advanced dexterity or coordination.