Deciphering the mechanoreceptor network of human hands

See allHide authors and affiliations

Science Translational Medicine  26 Jun 2019:
Vol. 11, Issue 498, eaax9569
DOI: 10.1126/scitranslmed.aax9569


Machine learning based on gloves with tactile sensor arrays helps scientists understand the complex patterns of hand grasping.

Have you ever wondered whether the dexterous, bionic prosthetic limbs from science fiction movies could become real? It may soon be possible. Engineers and neuroscientists are working on developing prosthetic hands that could convert contact forces to electronic signals and transmit them to neural interfaces for the amputee, enabling the patient to “feel” an object while handling it. Such bidirectional prosthesis may eventually enable highly sophisticated hand activities, such as playing musical instruments. However, many technological gaps must be addressed to achieve this goal. One major obstacle is the lack of a platform to measure how healthy persons synchronize hand motions and mechanoreceptor signaling. Once available, such a platform could produce the crucial datasets to help design a prosthesis that better resembles the patient’s lost hand.

Sundaram and co-workers addressed the above technological gap by using machine learning. Using a glove embedded with 548 tactile sensors covering the full hand, the authors recorded the motions and force patterns of a human hand when grasping 26 different objects of diverse shapes and weights. This work generated a large-scale tactile dataset with over 135,000 frames, which were subsequently analyzed by machine-learning algorithms based on deep convolutional neural networks. Particularly important to prosthetics applications, the above analyses helped the authors decipher how healthy persons synchronize hand motions and mechanoreceptor signaling. This was achieved by evaluating the correlations between different regions of the hand, including the palm and different finger segments, in changing positions and increasing contact forces under different arrays of tactile signals associated with different objects.

The glove embedded with tactile sensors could serve as a testbed to advance the understanding of the basic principles of dexterous manipulation, which would be crucial to prosthetic hand designs. In addition, this platform may benefit other technology based on human-machine interactions, such as robotic nursing and remote surgery.

Highlighted Article

View Abstract

Navigate This Article