Morphological Neural Computation Restores Discrimination of Naturalistic Textures in Trans-radial Amputees

Abstract

Humans rely on their sense of touch to interact with the environment. Thus, restoring lost tactile sensory capabilities in amputees would advance their quality of life. In particular, texture discrimination is an important component for the interaction with the environment, but its restoration in amputees has been so far limited to simplified gratings. Here we show that naturalistic textures can be discriminated by trans-radial amputees using intraneural peripheral stimulation and tactile sensors located close to the outer layer of the artificial skin. These sensors exploit the morphological neural computation (MNC) approach, i.e., the embodiment of neural computational functions into the physical structure of the device, encoding normal and shear stress to guarantee a faithful neural temporal representation of stimulus spatial structure. Two trans-radial amputees successfully discriminated naturalistic textures via the MNC-based tactile feedback. The results also allowed to shed light on the relevance of spike temporal encoding in the mechanisms used to discriminate naturalistic textures. Our findings pave the way to the development of more natural bionic limbs.

Figure 1

Neuromorphic artificial touch system encoding grating stimuli with type-1 mechanoreceptor model and performance of coarseness discrimination achieved by transradial amputees. (a) Experimental setup to evaluate the restoration of the perception of textural features in transradial amputees. (b) Transduction based on morphological neural computation (MNC) transforms the geometrical features of the surfaces into a sequence of spikes that are then delivered to the median or ulnar nerves by means of intraneural electrical stimulation. This specific illustration shows example patterns obtained with grating tactile stimuli with 3.0 mm SP1 and 0.5 mm SP2. Artificial fingertip touched the first half-grating for 2 s at 10 mm/s (20 mm sliding), followed by another tactile exploration of a second half-grating. Raster plots of the generated spikes are overlapped to the explored stimuli. (c) Overall performance achieved by the subjects in a three-alternative forced-choice experimental protocol (3AFC) of coarseness discrimination with the set of gratings described in Supplementary Table 1a.

Figure 2

Psychophysical performance in 3AFC discrimination of grating coarseness, scaling with relative variation of spatial period, and comparison between type-1 and type-2 receptor models. (a) Confusion matrix of 3AFC psychophysical task. (b) Confidence interval of discrimination for the ΔSP of each grating. (c) 3AFC psychometric curve as a function of ΔSP for the reference set of stimuli (black) and for finer set (red). Vertical dashed lines indicate 3AFC perceptual thresholds as determined by the logistic fit. (d) Same as (c), but sorting the stimuli based on relative variation of spatial period ΔSP/SP. (e,f) Structure of the biomimetic fingertip with sensor location mimicking location of type-1 (e) and type-2 (f) mechanoreceptors. (g) Same as (d), comparing the psychometric function resulting from the encoding of the spatial period via a type-1 receptor model (green) to that obtained via a type-2 receptor model (cyan). Note the shift in 3AFC perceptual threshold due to the change of the receptor model. (h) Spike trains generated by the same texture in the two directions during sliding of type-1 (green, top) and type-2 (cyan, bottom) mechanoreceptor.

Figure 3

Frictional and roughness characteristics of the experimented naturalistic textures. (a) Picture of naturalistic textures (b). 3D scatter plot of normalized roughness, friction coefficient, and max indentation force for each presentation of each of the naturalistic textures. (c) Confusion matrix of kNN decoding of textures based on the aforementioned features. (d) Average performance and confidence for single-texture and overall decoding.

Figure 4

Identification performance with naturalistic textures. (a) Confusion matrix of subject ALM performance in the identification of 4 naturalistic textures. (b) Confidence interval of identification of the same textures as a. (c,d) Same as a-b, for subject LOP.

Figure 5

Analysis of candidate decoding mechanisms for discrimination of naturalistic tactile stimuli. (a,b) Same as 4a-b, extended to 6 naturalistic textures. (c) Scatter plot of neural features related to texture properties for all presentations of all textures. (d,e) Performance of a decoder using kNN clustering of the spike trains based on neural features of (c), represented as confusion matrix (d) or displaying the confidence interval for all textures (e).