Fig. 5: Fully autonomous ultrasound imaging.

a The control strategy of FARUS. The autonomous scanning is achieved based on the fusion of force and visual information. b, c Evolution of force, probe position, and probe orientation for a patient with nodules during the experiment, which included the transverse scanning phase and longitudinal scanning phase. During the transverse scanning phase, the orientation of the probe at the end of the robot remained unchanged. Guided by the thyroid gland segmentation network, FARUS maintained the thyroid lobe in the center while moving the probe from the upper pole to the lower pole of the thyroid. When the nodule segmentation network detected a suspected nodule, the location of the suspected nodule was recorded. In the longitudinal phase, FARUS re-scanned the suspected nodules from a different angle to determine whether they were lesions (please refer to Supplementary Movie 3 for additional details). d–g Performance of the scanned image quality of FARUS on 19 patients. The performance metrics for evaluating the US image quality included confidence map, centering error, orientation error and image entropy. The shadowed area represents mean ± SD over the different experiments, while the curves inside the shadowed areas are the average results. Source data are provided as a Source Data file.