Lightweight soft neuroprosthetic hand

Mainly composed of electrical motors and sophisticated mechanical components, existing neuroprosthetic hands 1,2 are typically heavy (>400 g) and expensive (>USD 10,000), and they lack the compliance and tactile feedback of human hands. These limitations hamper neuroprosthetic hands’ innovation and broad utility for amputees 3-5 . Here we report the design, fabrication and applications of a lightweight (292 g) and potentially low-cost (component cost below USD 500) soft neuroprosthetic hand with simultaneous myoelectric control and tactile feedback. The soft neuroprosthetic hand consists of five soft fingers and a palm to give six active degrees of freedom under pneumatic actuation, four electromyography sensors that measure the surface electromyogram signals to control the hand to deliver four common grasp types, and five hydrogel-elastomer capacitive sensors on the fingertips that measure the touch pressure and elicit electrical stimulation on the skin of the residual limb. The soft finger is made of a fiber-reinforced elastomeric structure embedded with rigid segments to mimic the soft-joint/rigid-bone anatomy of the human finger. We use a set of standardized tests 6 to neuroprosthetic devices but also opens an avenue to widespread applications of lightweight, low-cost, and compliant hand replacements for amputees. hand type hand distributed test significant analysis. A two-way analysis of variance (ANOVA) is used to statistically analyze the significant influences of the two factors (neuroprosthetic hand types and task types) on the task in the two (items or S1-S9). According to the two-way ANOVA, there is interaction between the two factors. Bonferroni correction is used to correct for multiple comparisons. paired-samples t -test is used to compare the difference of blocks per and or different

neuroprosthetic devices but also opens an avenue to widespread applications of lightweight, low-cost, and compliant hand replacements for amputees.

Introduction
There are over five million upper-limb amputees worldwide, and the number increases by a substantial margin each year. Losing a hand is generally catastrophic, seriously limiting a person's ability in daily activities 1 . Although artificial prostheses are available, the most widely used prostheses are still cosmetic devices or functional hook-like grippers. While a few anthropomorphic neuroprosthetic hands 1,2 (such as i-Limb Hand, Michelangelo Hand, Bebionic Hand, and Vincent Hand) have been commercialized, they all rely on electrical motors and sophisticated mechanical components. The high weights (>400 g) and high prices (>USD 10,000) of these neuroprosthetic hands severely limit their broad utility for amputees 3,4 . It is also desirable for neuroprosthetic hands to have the compliance and tactile feedback of human hands 1,2,5 . For instance, the Pisa/IIT SoftHand 8-10 composed of electrical motors and tendon-driven mechanisms with compliant joints or skins has been developed as a myoelectric prosthesis with high compliance and 520 g of weight.
Here, we report the design, fabrication and applications of a lightweight (292 g) and potentially low-cost (component cost below USD 500) soft neuroprosthetic hand with simultaneous myoelectric control and tactile feedback for transradial amputees (Fig. 1A and Extended Data Tables 2, 3). The soft neuroprosthetic hand consists of five soft fingers and a palm to give six active degrees of freedom (DoFs), four electromyography sensors that measure the surface electromyogram (EMG) signals of residual forearm muscles to control the hand to deliver four common grasp types, and five hydrogel-elastomer capacitive sensors on the fingertips that measure touch pressure and elicit electrical stimulation on the skin of the residual limb. To evaluate the function of the soft neuroprosthetic hand, two transradial amputees have carried out a set of standardized tests 6 (including the Box and Blocks Test, all seven tasks in the Jebsen-Taylor Hand Function Test, and nine selected tasks in the Southampton Hand Assessment Procedure). One transradial amputee with the soft neuroprosthetic hand has further demonstrated the dexterous and versatile hand functions with primitive tactile sensation and closed-loop control in daily activities such as handling tools, eating, shaking hands, petting animals, and recognizing touch pressure.

Design
Hand. Each finger of the soft neuroprosthetic hand is based on a fiber-reinforced elastomeric tubular structure, in which two or three rigid segments with specific lengths are embedded to mimic the soft-joint/rigid-bone anatomy of the thumb or other fingers, respectively, of a human hand 35,36 ( Fig. 1B and Supplementary Figs. 1, 2). The thumb-palm connection in the soft neuroprosthetic hand is based on a fiber-reinforced elastomeric hollow pad with a strain-limiting layer ( Supplementary Fig. 3). The soft neuroprosthetic hand is designed to possess six active DoFs (Fig.   1C): each soft finger can be pneumatically actuated to provide one flexion DoF and the thumb-palm connection under pneumatic actuation further gives the thumb another circumduction DoF. In addition, the inherent compliance of the soft fingers can introduce many passive DoFs to the soft hand for dexterous and adaptive grasps even on fragile and soft objects (Fig. 1D), mimicking those passive compliance of human hands 8 .
We choose the pneumatically-actuated soft fingers owing to their advantages of low cost, light weight, and scalable fabrication 21 Table 4).
The five soft fingers are mounted on a three-dimensionally (3D) printed plastic palm (Imagine 8000, SOMOS Inc., Netherlands) in the shape of a human palm. Both the fingers and the palm are covered by a soft elastomeric layer mimicking the skin on the human hand. The palm skeleton is further connected to a customized plastic socket that fits with the residual limb of the transradial amputee. Owing to the pneumatic actuation and modular design of the soft neuroprosthetic hand, the pumps, valves, electronic boards and battery for the hand can be contained in a small bag (total weight of 444 g) on the waist of the amputee (Extended Data Fig. 2A). The pumps and electronic board are connected by soft tubing and electrical wires (hidden under clothes and in the socket) to the soft fingers and sensors on the hand, respectively. This modular design can dramatically reduce the weight of the soft neuroprosthetic hand to 292 g, much lighter than the weights of commerciallyavailable neuroprosthetic hands (420 g -628 g; Extended Data Table 3) and the average human hands (400 g) 1,2 . Furthermore, since the commercially-available neuroprosthetic hands integrate the electric motors, transmission mechanisms and batteries in their palms and sockets, it is challenging to significantly reduce their weights. We further demonstrate that the six active DoFs of the soft neuroprosthetic hand can be independently controlled with one pump and twelve valves (Supplementary Video 2). Notably we can also integrate the pumps, valves, electronic boards and battery in the socket of the soft neuroprosthetic hand if required in any application (Extended Data   Fig. 2B). Such a design increases the weight of the soft neuroprosthetic hand to 604 g, which is still lighter than or comparable to the weights of commercially-available neuroprosthetic hands such as the i-Limb large hand and the Bebionic medium hand (Extended Data Table 3).

EMG sensors.
In the socket of the hand, we implement four customized EMG sensors ( Fig. 1A and Extended Data Fig. 3), which will be mounted on the skin of the residual limb to record the surface EMG signals from the target muscles in the residual limb. Each EMG sensor (weight of 10 g) consists of three electrodes (a reference electrode and a pair of differential electrodes) and readout electronics (including two-level signal amplifier and filter circuits; see Supplementary Fig. 6). The locations of the EMG sensors with respect to the target muscles have been optimized, aiming to achieve a superior performance on decoding the amputees' motion intention. The decoded motion intention will be used to control the soft fingers and palm to deliver the corresponding grasp types.
Touch sensors and electrical stimulators. To sense the touch pressure of the soft neuroprosthetic hand applying on objects, we implement five soft capacitive touch sensors, each on a fingertip of the hand (Fig. 1C). The capacitive touch sensor is composed of an ionic hydrogel-elastomer hybrid structure 37-39 that forms a capacitor for sensing the touch pressure (Supplementary Fig. 7A). An increase in the touch pressure reduces the thickness of the elastomeric layer and, therefore, increases the capacitance of the capacitor (Supplementary Fig. 7B). The measured relative capacitance change is used to control an electrical stimulator (Supplementary Fig. 8), which outputs programmable electrical pulses via a noninvasive stimulation electrode on a specific region of the residual limb (Supplementary Fig. 9) to inform the amputee the touch pressure on the corresponding fingertip. This is the first time that hydrogel-elastomer capacitive touch sensors have been used on the soft neuroprosthetic hand to measure the touch pressure at the fingertips.
Control algorithm. By integrating the EMG sensors, touch sensors and electrical stimulators with the hand, we develop a bidirectional human-machine interface for the soft neuroprosthetic hand (Supplementary Fig. 8 for the description of experimental setup). We choose a pattern recognition approach 40 to classify the surface EMG signals into five classes corresponding to the four common grasp types 41,42 and rest type of human hands (see Supplementary Information for the descriptions of pattern recognition algorithms). The decoded grasp types are then mapped onto the actuation command for the corresponding soft fingers and palm. Based on the relative changes of the measured capacitances of the capacitive touch sensors (caused by the touch pressure), we can program the electrical stimulator to output the corresponding stimulation patterns, which inform the amputee the touch pressure on each fingertip. Based on the informed touch pressure, the amputee can further change the grasp type of the soft neuroprosthetic hand by varying the EMG signals, therefore forming a closed-loop control of the hand.

Characterization
We use a simple and scalable method to fabricate and assemble the soft neuroprosthetic hand (see Methods section "Fabrication and assembly of the soft neuroprosthetic hand" and Extended Data Fig. 4). We next characterize the performance of the fabricated soft neuroprosthetic hand. To monitor its kinematics, we develop a motion tracking system (Supplementary Fig. 10). For the 1-DoF flexion fingers (i.e., index, middle, ring, and little fingers), we attach markers to their flexible joints and fingertips to measure the bending angles of the joints and the fingertips ( Fig. 2A). We find that an increase of the applied pneumatic pressure increases the bending angles of the fingertip to a maximum of 231° at 120 kPa pneumatic pressure (Fig. 2B). In addition, the measured bending angles of the flexible joints are consistent with the finite-element model's prediction (Fig. 2B, Fig. 1A and Extended Data Table 4). For the 2-DoF thumb, we attach markers to its flexible joints, thumb tip and thumb-palm connection (Fig. 2C). The results indicate that the thumb has a maximum flexion of 69° and a maximum palm circumduction of 28° at 80 kPa pneumatic pressure, which also agrees well with the finite-element model's prediction (Fig. 2D, Fig. 1B, Supplementary Fig. 11 and Extended Data Table 4). The maximum bending angles of the soft fingers are comparable to those of existing rigid counterparts (Extended Data Table 3).

Extended Data
To evaluate the load capacity of the soft neuroprosthetic hand, we use the hand (at 100 kPa pneumatic pressure to the 1-DoF fingers and 80 kPa pneumatic pressure to the 2-DoF thumb) to grasp a 55 mm-diameter cylinder while measuring the grasping force with an electronic dynamometer ( Supplementary Fig. 12). The results indicate that the soft neuroprosthetic hand gives maximum grasping forces of 18 N and 17 N along the vertical and horizontal directions, respectively ( Fig. 2E, F). Therefore, the soft neuroprosthetic hand can perform most grasping tasks in daily activities, which generally require grasping forces below 10 N for human hands 43 . The grasping forces of the soft fingers can be potentially increased in future studies by using stiffer elastomers for the fingers and applying higher pneumatic pressures.
Furthermore, we demonstrate that our soft fingers have repeatable pressure-flexion relations with small hysteresis over 10,000 cycles of actuations (Fig. 2G), and they are resilient to be bent by arbitrary angles, struck with a steel hammer, and run over back and forth by one wheel of a 1500 kg vehicle (Supplementary Fig. 13 and Supplementary Video 3).
Based on the usage frequencies of grasp types in daily activities 41,42 , we choose to recognize the most commonly used four grasp types (i.e., Power, Precision disk, Tripod and Lateral pinch) of amputees through the myoelectric control interface (Extended Data Fig. 5). Using the four-channel EMG sensors (one sensor at each channel), the soft neuroprosthetic hand can decode the intended four grasp types and rest type ( Fig. 2H and Supplementary Fig. 14).
We next calibrate the capacitive touch sensor by uniformly compressing the sensor and then measuring its capacitance. Under zero pressure, we denote the measured capacitance of the sensor as . With the increase of the applied pressure, we can obtain the current capacitance C of the touch sensor and calculate the change of the capacitance ∆ . The experimental results demonstrate that, when the uniformly applied pressure on the sensor increases from 0 kPa to 55.8 kPa, the ∆ / varies from 0 to 0.85, giving a sensitivity of 0.016 kPa -1 (Supplementary Fig. 7B).  Table 5) further demonstrate that, compared to the rigid neuroprosthetic hand, the soft neuroprosthetic hand has significantly superior performances in 7 items (i.e., BBT, J3, J4, J7, S1, S3 and S5), significantly inferior performances in 3 items (i.e., J5, S6 and S7), and statistically similar performances (p > 0.05) in 7 items (i.e., J1, J2, J6, S2, S4, S8 and S9). Based on the same set of standardized tests and statistical analyses (Extended Data Fig. 8 and Extended Data Table   6), another subject with transradial amputation demonstrates that the performances of the soft neuroprosthetic hand are significantly superior in 5 items (i.e., J3, J5, S4, S5 and S8), significantly inferior in 4 items (i.e., J1, S6, S7 and S9), and statistically similar in 8 items (i.e., BBT, J2, J4, J6, J7, S1, S2 and S3) compared to those of the rigid counterpart ( Supplementary Fig. 16) [45][46][47] . (Note the reported results have only provided mean values of their performances, whose statistical difference from the performances of the soft neuroprosthetic hand cannot be evaluated.)

Applications on amputees
Next, we perform comparative experiments for the same subject wearing the soft neuroprosthetic hand and the rigid i-Limb hand to grasp fragile objects (e.g., strawberry, bread, paper cup). The results indicate that the rigid neuroprosthetic hand damages the strawberry and bread, and tends to crush the paper cup (Extended Data Fig. 9A and Supplementary Video 7).
Owing to the inherent compliance, the soft neuroprosthetic hand can guarantee safe interactions with these fragile and soft objects (Extended Data Fig. 9B and Supplementary Video 7).
Furthermore, we demonstrate that one subject with the soft neuroprosthetic hand can intuitively perform the four common grasp types to grasp different objects (Supplementary Video 8) and handle commonly-used items in daily activities, such as food (e.g., potato chips, cakes, strawberries and apples), commodities (e.g., clothes, bags, laptops, water glasses, bottles, tissues and dishes), and tools (e.g., hammers and pliers). The subject also achieves safe interactions with other persons (e.g., shaking hands), animals (e.g., petting a cat) and environments (e.g., touching a flower) ( Fig.   3B and Supplementary Video 9). We further demonstrate that the subject can successfully carry out delicate tasks to handle objects with complex shapes and different sizes and then insert them in the corresponding slots precisely (Fig. 3C, and Supplementary Video 10). In a load test, the subject wearing the soft neuroprosthetic hand can lift a payload of 2.3 kg ( Fig. 3D and Supplementary

Video 11).
Tactile feedback and closed-loop control. Furthermore, we demonstrate that the subject with the soft neuroprosthetic hand can restore primitive touch sensation based on the capacitive touch sensors and electrical stimulators. When the effective pressure on the touch sensor of a fingertip reaches a threshold (i.e., threshold ∆ / ), the electrical stimulator will be triggered to generate an electrical pulse (amplitude of 4.0 mA, pulse width of 200 μs, and pulse frequency of 20 Hz) to stimulate a specific region on the residual limb corresponding to the fingertip ( Fig. 4A and Supplementary Fig. 9). We set the threshold effective pressure to be 2.3 kPa (i.e., the threshold ∆ / = 0.1), so that the touch sensors are sufficiently sensitive to touch pressures commonly experienced in daily activities 44 yet unaffected by environmental noises and crosstalk among sensors (Supplementary Fig. 17). In a blindfolded and acoustically-shielded interaction experiment, we demonstrate that, in the blindfolded and acoustically-shielded experiment, the subject can restore the graded tactile feedback to discriminate three cylinders with different diameters (i.e., 60 mm, 70 mm and 80 mm) with an accuracy of 96.25% (Extended Data Fig. 10 and Supplementary Video 14).

Conclusion
We report a lightweight and potentially low-cost soft neuroprosthetic hand for transradial amputees to restore versatile hand functions and primitive tactile sensation. We choose a modular design to enable efficient iteration of the design, fabrication and control, as well as rapid replacement of the components in case of wear or damage. Compared to commercially-available neuroprosthetic hands, our soft neuroprosthetic hand has advantages including the intrinsic compliance, light-weight (292 g), potentially low-cost (component cost below USD 500), and embedded soft touch sensors, while maintaining the similar active DoFs, the number of joints, and the maximum bending angles of the joints (Extended Data Table 3). As the first demonstration of the soft neuroprosthetic hand on transradial amputees, we employ the most commonly used EMGdecoding algorithm and the electrotactile feedback. To further improve the performance of the soft neuroprosthetic hand, advanced EMG-decoding algorithms 48-50 and sensory feedback approaches [11][12][13][14][15][16][17][18] can be implemented in the future. Overall, this work has the potential to provide the nextgeneration personalized neuroprosthetic hands that are intrinsically soft, lightweight and potentially low-cost for upper-limb amputees, and to broaden the future applications of soft robotic systems.

Synthesis of the ionic hydrogel
The ionic hydrogel 37-39 is a polyacrylamide (PAAm) hydrogel containing lithium chloride (LiCl). The PAAm-LiCl hydrogel is synthesized by using the acrylamide (AAm; J&K) as the monomer, N,N'-methylenebisacrylamide (MBAA; Molbase) as the crosslinker, LiCl monohydrate (LiClꞏH2O; Sinopharm Chemical Reagent) as the ionic conductive medium, and 2-Ketoglutaric Acid (Adamas) as the photoinitiator. The monomer solution is prepared by mixing AAm, LiClꞏH2O and deionized water with a mass ratio of 9.98%: 16.16%: 73.86%. MBAA solution is dissolved into deionized water with a mass ratio of 1.2%. Then, the monomer solution, MBAA solution, and 2-Ketoglutaric Acid are mixed with the mass ratio of 96.67%: 1.13%: 2.20% to form the hydrogel precursor ink for further fabrication. Note that the LiCl is added into the hydrogel, not only serving as a conductive medium but also as a hygroscopic salt to maintain water in the hydrogel in ambient environments [37][38][39] .

Fabrication and assembly of the soft neuroprosthetic hand
We present a simple and scalable method to fabricate and assemble the soft neuroprosthetic hand (Extended Data Fig. 4). For the soft fingers, we use the Dragon Skin 10 (Smooth-On Inc., USA) silicone rubber for the inner elastomeric tubular structure, the Ecoflex 0030 (Smooth-On Inc., USA) silicone rubber for the outer elastomeric skin, and the polyethylene thread for the fiber reinforcement. We attach the carbon fiber-reinforced plastics lamination with the heat-shrink tubes as the embedded rigid segments of the soft fingers. The palm skeleton is 3D printed with the commercial photosensitive resin (Imagine 8000, SOMOS Inc., Netherlands) and the covered elastomeric skin is made of Ecoflex 0030 silicone rubber. To tune the color of the elastomeric skin close to the amputee's skin color, we can add Slic Pig pigment PMS 488C (Flesh color, Smooth on Inc., USA) into the Ecoflex 0030 silicone rubber (see Supplementary Fig. 18 for an illustration). The socket is fabricated with the vacuum-forming thermoplastic acrylic resin and the commercial gold-plated copper blocks are embedded in the socket as the EMG electrodes. We fabricate the capacitive touch sensors by curing the hydrogel precursor (containing acrylamide, crosslinker, photoinitiator, and ionic conductive medium) into a VHB elastomeric matrix. Notably, all the mechanical components can be constructed with low-cost commercially available materials (Extended Data Table 2). The fabrication and assembly steps of the soft neuroprosthetic hand are detailed in the Supplementary Information.

Analytical model for the design of the soft finger
The method for efficiently predicting the bending angles of the flexible joints in the soft fingers plays an important role in mimicking the structure and function of human fingers 24 . We first develop an analytical model to analyze the pneumatic response of a flexible joint. We use a non-linear elasticity approach to analytically model the response of the flexible joint under pneumatic actuation. Specifically, we model the flexible joint as a hollow cuboid of isotropic incompressible neo-Hookean solid with shear modulus of μ and a stiff inextensible layer beneath the cuboid. We take the thickness, height, and width of the hollow cuboid at the undeformed state as t, H, and W, respectively (Supplementary Fig. 4A). Due to the lateral constraint from the stiff fibers surrounding the cuboid, we assume the cross-sections of the cuboid remain planar upon pressurization, which has been validated by the finite-element simulation as well. Furthermore, since the arrangement of fibers are symmetric without twisting, we express the axial stretch of the cuboid cross-section in a linear form of [ ] is the axial stretch at the top surface with the elongated length of l at the top surface (Supplementary Fig. 4A). The deformation gradient at the side surfaces reduces to We model the elastomeric cuboid as an incompressible neo-Hookean solid with strain energy of . Therefore, the axial nominal stress in the cuboid is Assuming the deformation of the hollow cuboid is pure bending, we can obtain the geometrical relation between the bending angle α and the dimensions of the flexible joint (i.e., l, L, and H) l L H a -= Since the bending actuation is driven by the moment created by the internal pressure imposing on the cap of the flexible joint Mp, the relation between the pneumatic pressure p and the bending angle Solving Eq. (4) and Eq. (5) yields the relation between the pneumatic pressure and the bending angle as To verify the above analytical model, we further simulate the pneumatic response of the flexible joint, which consists of a cuboid elastomeric chamber, a network of stiff fibers surrounding the elastomeric chamber, and a stiff layer beneath the chamber. We model the elastomeric chamber as a neo-Hookean solid with solid elements (C3D10H), the strain energy density of which is where is the shear modulus,  is the bulk modulus, det F . We set =80 kPa and  / = 1000 to impose the nearly incompressibility of the material. The stiff layer is modeled as a skin of shell elements with Young's modulus of 210 GPa. The fiber is modeled as a beam element (B32H) with Young's modulus of 1 GPa. The fiber direction is set perpendicular to the axial direction of the finger to constrain the lateral expansion of the flexible joints during inflation 17 . Static simulations are performed by applying pressure on all internal faces of the elastomer chamber with zero displacements at the base portion of the finger as the boundary conditions. We first compare the axial stretch at the top surface of the flexible joint in experiments and simulations, verifying the main deformation mode of the flexible joints is the bending motion (Supplementary Fig. 4B). We further compare the increase of the bending angle α as a function of the applied inflation pressure p for the flexible joints with various lengths in the finite-element simulation and the analytical model, showing good agreement (Supplementary Fig. 4C). Despite the nonlinear large deformation of the elastomeric cuboid, we also show that the bending angle of the joint is almost linearly proportional to its initial length L under a constant inflation pressure p (Supplementary Fig. 4D).
We further study the pneumatic response of a soft finger consisting of an inner hollow cuboid elastomeric tubular structure surrounded by a network of fibers, rigid segments, and a stiff layer beneath the elastomeric tube (Supplementary Fig. 5A). Mimicking the dimensions of the distal phalanx, middle phalanx, and proximal phalanx in human fingers 17,18 , we set the total length of the finger as 85 mm and the length ratio of the top, middle, and bottom parts as 2: 3: 5 in the model (Supplementary Fig. 5A). We show that the ratio of the bending angles of the three joints are nearly constant due to the same pneumatic pressure applied to the three joints (Supplementary Fig. 5B). Moreover, the ratio of the bending angles of the three joints are independent of the total length of all flexible joints as long as the ratio of the lengths of the three joints are fixed (e.g., : : 2: 4: 1, Supplementary Fig. 5C). In existing prosthetic hands 1,2 , the bending motion of a finger typically follows two characteristics: 1) the bending angle of α is approximately half of that of α ; and 2) the bending angle of α is approximately the same as α . Based on our analysis, we show that the central idea to mimic the motion trajectory of human fingers is to design the length ratio of the three flexible joints. Specifically, the length of the distal interphalangeal joint needs to be half of the length of the metacarpophalangeal joint , while the length of the proximal interphalangeal joint is preferred to be the same as the length of the metacarpophalangeal joint . Although we analyze the soft fingers with three flexible joints, the conclusion also applies to the thumb with two flexible joints, which will also be verified in the simulation with the finite-element model.

Finite-element model for the design of the soft finger
To further verify the experimental results of the hand with five soft fingers, we develop a finiteelement model of the soft fingers in ABAQUS to quantitatively predict the bending angles of the flexible joints and the motion trajectories of the fingers upon inflation. All material parameters are experimentally measured from mechanical characterizations 51 . The elastomeric tube is modeled as a neo-Hookean solid (Eq. 7) with the shear modulus =85 kPa and the ratio of bulk modulus to shear modulus  / = 1000 to impose the near incompressibility of the material. The silicone skins on the outer surface are modeled as a linear elastic material with elastic modulus 0. To reduce the computational cost, we use the skin module to model the heat-shrinkable tube, the carbon fiber-reinforced plastics, and the fiberglass grid. The Ecoflex 00-30 layer are modeled as skins with uniform thickness as the outer surfaces of the fiber-reinforced inner elastomeric tube. An extra skin of a stiffer rubber material is added on the surfaces of the thumb-palm connection to model the constraint effect of the silicone layer. The elastomer of the inner tube is modeled as solid elements C3D4H. The fibers are modeled as beam elements B31. Tie constraints are set between the fibers and the elastomeric chambers of the finger actuators, the thumb actuator, and the thumbpalm connection actuator. The proximal ends of the thumb-palm connection are fixed to the rigid palm skeleton. The gravity load and a pressure load are applied to the inner surface of the tube of the soft fingers. The finite-element model is in quantitative agreement with the experimental results (Extended Data Fig. 1, Extended Data Table 4 and Supplementary Video 1).

Participant recruitments
All experiments were conducted in accordance with the declaration of Helsinki and approved by the Ethics Committee of Human and Animal Experiments of Shanghai Jiao Tong University. The transradial amputees participated in this study were recommended by Shanghai Liankang Prosthetics and Orthotics Manufacturing Co. Ltd, Shanghai, China. The amputees did not have any prior neuromuscular disorders, and were informed about the experimental procedure and signed the informed consent forms (ICFs) prior to the participation.

Training process for EMG decoding
During the training process, the subject who performs the predefined grasp types (Fig. 2H) will be notified by the vibration from a motor placed on the socket. Each grasp type is maintained for 5 seconds and a rest period is required between two grasp types to prevent possible fatigue of the subject. In the experiments, short-time vibrations are provided at the beginning and end of each grasp type to notify the subject to transit between different hand grasps, and a long-time vibration is provided at the end of the training session. Each channel (recorded by each EMG sensor) of the EMG signals from the subject performing a predefined grasp type is segmented into data fragments with 200-ms windows. Thereafter, a time-domain feature set from the data fragments of each channel for the predefined grasp type is extracted, including the mean absolute value, waveform length, zero-crossings, and slope sign changes. The features from different channels for the same predefined grasp type are cascaded to give a feature vector. Subsequently, the feature vectors obtained from the training data are fed into the linear discriminant analysis (LDA) 40  Although the LDA-based surface EMG decoding algorithms can accurately classify the 5 class in laboratory settings (Supplementary Fig. 14), the performance will be reduced for amputees in realistic activities 52 . Therefore, in many demonstrations of the functions of the soft neuroprosthetic hand, we choose a single grasp classifier based on testing items of the standardized tests. We have indicated in the Video Captions either 5 class classifier or a single grasp classifier has been chosen.

Training process for tactile feedback
In the training process for tactile feedback, the stimulation current is set as bi-phasic, rectangular current pulses. Based on our previous results 53 , we set the current amplitudes, pulse widths and frequencies for all five channels as 4 mA, 200 μs and 20 Hz, respectively. During the training process, we compress each soft finger three times in a random order to check if the subject can discriminate the compressed finger based on the electrical stimulation. The training process lasts about 5 min.

Data analysis and statistics.
All data are analyzed using the available built-in functions of MATLAB (R2016b, The Mathworks Inc., USA) and SPSS (version 22, IBM Inc., USA). All data are reported as mean values with standard deviations when indicated. In the statistical analyses of standardized tests, the factors (independent variables) are the neuroprosthetic hand type (the soft neuroprosthetic hand and the rigid neuroprosthetic hand) and task type (items J2-J7 or S1-S9). The dependent variable is the task time. All the data are demonstrated normally distributed through the Kolgomorov-Smirnov test (p > 0.05) before significant analysis. A two-way analysis of variance (ANOVA) is used to statistically analyze the significant influences of the two factors (neuroprosthetic hand types and task types) on the task time in the two sessions (items J2-J7 or S1-S9). According to the two-way ANOVA, there is interaction between the two factors. Bonferroni correction is used to correct for multiple comparisons. Two-tailed, paired-samples t-test is used to compare the difference of blocks per minute (in the Box and Blocks Test) or words per minute (item J1) with different neuroprosthetic hand types.

Data availability
The datasets generated or analyzed during the current study are available from the corresponding author on reasonable request.    The myoelectric control interface is designed for intuitive control of the soft neuroprosthetic hand. The myoelectric control is achieved by a customized onboard measurement and control system consisting of four-channel EMG sensors, control unit (including the signal processing unit for EMG decoding and the micro-controller for pneumatic actuation), pumps, valves, solid-state relays, and the power (battery and voltage regulators). The four-channel EMG sensors (embedded in the socket and mounted on the skin of residual forearm muscles) record the muscle activities of amputees, which is processed by the readout electronics with the amplification and Butterworth filtering . The signal processing unit receives the amplified and filtered signals from EMG sensors and classifies the signals into several discrete classes related to the grasp types of amputees' intention. Through a universal asynchronous receiver/transmitter (UART) port, the classification results are sent to a micro-controller (Nano, Arduino Inc., Italy). The micro-controller employs the classified grasp types to control the pumps and valves through two solid-state relays, resulting in the intuitive control of the soft neuroprosthetic hand. The pins (D2-D7) and pins (D8-D13) connect the output pins of the micro-controller relating to the corresponding pins of pump and valve relays.  Material: gold-plated copper ~100

Extended Data Table 3 | Comparison of the soft neuroprosthetic hand with representative commercially-available neuroprosthetic hands. (A) Comparison of the weight. (B)
Comparison of the actuation mechanism, active DoFs, the number of joints, and bending angles of the joints.
Note: NA and NP are abbreviations for not applicable and not provided, respectively.

Extended Data Table 4 | Relative prediction errors of the finite-element model for the 1-DoF finger and 2-DoF thumb.
Note: The relative error e is calculated by 100%, where and are bending angles of the finite-element model and the average experimental results (n = 3), respectively. Sim and Exp are abbreviations for simulation and experiment, respectively. Table 5 | Comparisons of the performances of the soft neuroprosthetic hand and a conventional rigid neuroprosthetic hand 7 on the same subject evaluated with the standardized tests. (Each item in the standardized tests is performed three times and the mean values represent the average for n = 3 measurements.) Significant results (p <0.05) are boldface. Table 6 | Comparisons of the performances of the soft neuroprosthetic hand and a conventional rigid neuroprosthetic hand 7 on another subject evaluated with the standardized tests. (Each item in the standardized tests is performed three times and the mean values represent the average for n = 3 measurements.) Significant results (p < 0.05) are boldface.