Extended Data Fig. 9: Evaluation of the resolution dependent tracking accuracy using the ultrasound wristband.
Artificial ultrasound datasets with reduced resolution are generated by applying Gaussian blur to the original images (a), simulating imaging at lower center frequencies (b. 8 MHz, c. 5 MHz, and d. 3 MHz). The AI model is trained and tested on each dataset to assess performance under degraded resolution. a-d, Left: Representative examples of artificially blurred ultrasound images. Right: Corresponding tracking accuracy for each condition. (All data in this figure are presented as mean values +/− standard deviation, n = 5 independent measurements.).
