Sensor-free contact force guidance for ultrasound imaging

Objectives: The primary aim of this study is to develop a sensor-free contact force guidance to enable automatic control of the contact force exerted by an ultrasound probe during robotic ultrasound scanning. By estimating the force differential between the left and right sides of the ultrasound transducer without relying on force sensors, the system avoids complexities related to sensor installation and data alignment. A deep learning approach was employed to estimate the contact force using sequences of ultrasound images, offering a novel solution for sensor-free force measurement in robotic ultrasound applications. Materials & Methods: Experiments were conducted under two conditions: with and without force sensors attached to the transducer surface. Consistent positioning and force conditions were maintained using precise control of the robotic arm, allowing accurate matching of force data from sensors with ultrasound image sequences obtained from sensor-free experiments. Ultrasound images were captured using a Vantage ultrasound system with a linear array transducer (transmit frequency 7.8 MHz) and three different tissue-mimicking phantoms. A total of 109 sets of ultrasound image sequences and corresponding force values were collected. The force prediction network, combining spatial and temporal feature extraction architectures, was trained using this dataset. The model's performance was evaluated by calculating the mean squared error (MSE) between predicted and measured force differences. Results: The proposed method was tested on tissue-mimicking phantoms and integrated into a real-time feedback control loop to dynamically adjust the probe's orientation. Results demonstrate that the model predicts the force difference and effectively maintains optimal imaging conditions through real-time adjustments. This study highlights the potential of sensorless techniques to simplify robotic ultrasound systems while enhancing their accuracy and reliability.