Ultrasound image features of the wrist are linearly related to finger positions

Abstract

Ultrasound imaging is a widespread technique to gather live images of the interiors of the human body. It is safe and provides high spatial and temporal resolution. In this paper we show that features extracted from the ultrasound section of the human wrist can be used to fully reconstruct the hand movements, including flexion of all fingers and the rotation of the thumb. Surprisingly, it turns out that there is a clear linear relationship between image features and finger positions. The related matrix can be estimated on a rather small subset of samples, and the reconstruction is quite robust across single- and multi-finger movements. This technique can be used to control advanced mechatronic hands, and it finds its paradigmatic application in the case of hand amputees.