Abstract
Autonomous robotic ultrasound has recently gained considerable interest, especially for collaborative applications. Existing methods for acquisition trajectory planning are solely based on geometrical considerations, such as the pose of the transducer with respect to the patient surface. PURPOSE: This work aims at establishing acoustic window planning to enable autonomous ultrasound acquisitions of anatomies with restricted acoustic windows, such as the liver or the heart. METHODS: We propose a fully automatic approach for the planning of acquisition trajectories, which only requires information about the target region as well as existing tomographic imaging data, such as X-ray computed tomography. The framework integrates both geometrical and physics-based constraints to estimate the best ultrasound acquisition trajectories with respect to the available acoustic windows. We evaluate the developed method using virtual planning scenarios based on real patient data as well as for real robotic ultrasound acquisitions on a tissue-mimicking phantom. RESULTS: The proposed method yields superior image quality in comparison with a naive planning approach, while maintaining the necessary coverage of the target. CONCLUSION: We demonstrate that by taking image formation properties into account acquisition planning methods can outperform naive plannings. Furthermore, we show the need for such planning techniques, since naive approaches are not sufficient as they do not take the expected image quality into account.