Chinese-led team’s ultrasound robot uses AI for thyroid scan results ‘close’ to that of doctors

A team of scientists based in China and Canada may have become the first to deploy a fully autonomous ultrasound robot for thyroid scans, with results “close” to that of human doctors.

The robot used artificial intelligence to perform the exams and identify malignant nodules – all without human assistance.

The team, led by researchers at the South China University of Technology (SCUT), used human skeleton recognition, reinforcement learning, and physical force feedback to develop their fully autonomous robotic ultrasound system (FARUS).

“Experimental results on human participants demonstrated that this system can perform high-quality ultrasound, close to manual scans obtained by clinicians”, as well as detect and identify characteristics of nodules, the researchers wrote in a paper published in the peer-reviewed journal Nature Communications earlier this month.

“To the best of our knowledge, this is the first in-human study of fully autonomous robotic ultrasound scanning for thyroid.”

According to Du Guanglong, study author and professor at SCUT, the system is able to “independently scan the thyroid area and identify malignant nodules without human assistance”.

FARUS can provide “characteristic data for diagnosis and treatment, effectively promoting medical efficiency”, Du told Chinese-language news site Science Times.

Performing thyroid ultrasounds is a “physically and cognitively exhausting” task for sonographers and radiologists, and the diagnosis can vary greatly between doctors. Patients can also experience delays in diagnosis and treatment due to long wait times for ultrasounds, according to the paper.

Various autonomous ultrasound robot systems have been proposed to help solve this problem, but adapting these systems to a clinical setting “is still challenging”, the researchers said, as this requires a high level of perception and control while ensuring patient safety.

According to the paper, FARUS was able to provide “a convenient autonomous tool integrating nodule detection, lesion localisation and automatic classification [of whether an anomaly was benign or malignant]”.

“Our work addresses the gap between existing research and clinical application by demonstrating the deployment of this system in a real-world clinical setting,” the researchers said.

The robot “doctor” developed by the team first instructs the patient to turn their head so ultrasound gel can be applied, before it moves a probe on their neck and adjusts its position based on AI reasoning.

Any nodules located are then recorded by the robot, which is also able to halt the scan and adjust positioning and the force at which it presses down if the patient moves during the exam, the researchers said.

And while standard thyroid examinations require a patient to lie down and be still, their system can perform the test while the patient is sitting upright.

“It is remarkable that the entire scanning process, including thyroid searching, force control, image quality optimisation, and suspected nodule detection was completed autonomously,” the researchers wrote. Carleton University in Ontario also took part in the study.

The researchers also compared the results of exams made by FARUS and doctors on the same patients. They found that the robot system identified the same patients who required medical intervention on their nodules, although size and low-contrast nodules appeared to present challenges, as it missed some smaller ones and indicated possible false positives.

Although FARUS demonstrated its ability to scan for nodules and cancer-risk data, the team said that “further clinical studies are essential to assess its safety as a screening tool for probably or definitely malignant nodules.”

The researchers also conducted surveys of the patients, and while some were anxious about undergoing the procedure, most reported that they felt safe and did not feel any pain or discomfort during the exam.

“This non-invasive, rapid, and accurate screening strategy can provide an early warning of thyroid nodule development,” the researchers said.

They said FARUS could be used in hospitals, and also adopted as a screening method in outpatient clinics and remote areas.



Read More

Leave a Reply