Robot Assistant for Sonographic Thyroid Examination and Reconstruction (RASTER) - a robotic system for 3D thyroid volume estimations
Abstract
While the exact prevalence of thyroid nodules in the wider population is not known, studies have shown between a 2% and 65% prevalence rate in the human population. The vast majority of these tumors are benign. However, the rate of thyroid cancer has at least doubled since 1980 while a corresponding increase in mortality has not occurred, leading researchers to conclude the rise is a measure of how many cancerous nodules are being found. Physicians have turned to ultrasonography as a relatively low-cost method of imaging thyroid nodules in a non-invasive way. These exams, when conducted by physicians and trained ultrasound technicians, are prone to human error which can cause inconsistent results between exams. This thesis proposes a novel robotic system for assisting physicians in conducting thyroid ultrasound exams. The proposed system uses a 6 DoF robotic manipulator mounted with a wireless ultrasound probe to capture ultrasound images at known locations. A non-deep learning, real-time image segmentation framework, initialized by the physician, tracks the region of interest throughout the travel assisted by a force, probe orientation, and position control framework. A second non-deep learning, non-real time image segmentation framework combined with the positional data from the robot generates an approximation of the thyroid as a 3D volume. Testing completed on a training phantom demonstrated the ability of the system to track the patient profile accurately and in real-time. From the experimental data, a volume was generated and visualized properly.