Robot-assisted biopsy on MR-detected lesions

In women, breast cancer is the most common cancer and it is the leading cause of cancer death in many countries. In the diagnostic work up of breast cancer, the biopsy is a crucial step to determine malignancy of a lesion and its success rate is largely dependent on accurate lesion localization and needle placement. Especially when lesions are MRI-visible only, there are limitations in the accuracy which can negatively impact the outcome of the biopsy.

Robotic assistance can take place inside and outside the MR-bore, and has the potential to improve the accuracy of the biopsy procedure. In the MRI and Ultrasound Robot-assisted biopsy (MURAB) project, a robotic setup is presented to assist the radiologist with an ultrasound-guided biopsy on an MR-detected lesion. The robotic setup consists of a seven-degrees-of-freedom robotic arm equipped with an end-effector which is positioned under a patient bed. The patient lies on this bed with the examined breast through a hole such that it is freely accessible by the robot. The robot achieves an accurate notion of the current lesion position by combining the preoperatively acquired MRI images with stereo vision and intraoperatively acquired ultrasound images. Additionally, a patient-specific biopmechanical model is built utilizing elastography which is used to predict deformations caused by needle insertion.

In this thesis, several aspects of the setup are worked out in detail. The end-effector design is presented which contains an actuated needle guide, an ultrasound probe, a set of stereocameras, a projector and lighting. A compliant controller is introduced which which limits the energy in the null space of the robot, optimizes the joint positions with respect to their limits and avoids the joint limits. Furtermore, it is shown how ultrasound feedback is taken advantage of during ultrasound acquisitions and the biopsy procedure.

In several phantom experiments, it is shown that the robot more accurately registers the preoperatively acquired MRI images with the patient and more accurately targets the lesion than is currently the case in clinical practice. During robotic ultrasound acquisitions, ultrasound feedback in the form of confidence maps is utilized to correct a preoperatively planned trajectory to obtain higher quality ultrasound images, and to fully autonomously scan a beforehand unknown surface. Additionally, confidence maps are used to determine when to start tracking the lesion position when approaching the patient, and needle detection is used to correct for errors in the current needle position. All in all, this thesis nicely shows the potential benefits of the introduction of robotics to the diagnosis of breast cancer.