Monday, December 2 | 2:20 p.m.-2:30 p.m. | M6-SSMK03-6 | Room E450A
In this presentation, audience members will see the results of an ultrasound-based AI risk grading system for sarcopenia in older adults.
Xinyi Tang from West China Hospital in Chengdu, Sichuan, China will present findings showing that the system can successfully classify sarcopenia risk in a cross-regional, multi-ethnic population.
In the study, sonographers obtained ultrasound images for patients presenting with sarcopenia, the degeneration of muscles due to age or immobility. The team connected four ultrasound images of each sample and input them into the Resnet50 network. From there, it connected factors such as muscle thickness, body mass index (BMI), and age to create a multilayer neural network.
Final analysis included 1,145 people. For measuring muscle decline, the system achieved an area under the curve (AUC) of 0.879 in men and 0.813 in women, respectively. The system also achieved higher sensitivity and specificity values for men than it did for women.
In addition, the team classified patients based on risk. Out of 380 people identified by the system as low-risk, 1.3% were diagnosed with sarcopenia. Among the 466 people classified as middle-risk, 80.7% had at least one sarcopenia-related symptom, and 17.6% were diagnosed with sarcopenia. Finally, out of the 299 people classified as high-risk, more than half were diagnosed with sarcopenia, and about one-third were diagnosed with severe sarcopenia.
What else did the researchers observe that makes a case for this system’s utility? Attend this session to find out more.
