Crossmodal Pattern Discrimination in Humans and Robots: A Visuo-Tactile Case Study

AbstractThe quality of crossmodal perception hinges on two factors: The accuracy of the independent unimodal perception and the ability to integrate information from different sensory systems. In humans, the ability for cognitively demanding crossmodal perception diminishes from young to old age.To research to which degree impediments of these two abilities contribute to the age-related decline and to evaluate how this might apply to artificial systems, we replicate a medical study on visuo-tactile crossmodal pattern discrimination utilizing state-of-the-art tactile sensing technology and artificial neural networks. We explore the perception of each modality in isolation as well as the crossmodal integration.We show that in an artificial system the integration of complex high-level unimodal features outperforms the comparison of independent unimodal classifications at low stimulus intensities where errors frequently occur. In comparison to humans, the artificial system outperforms older participants in the unimodal as well as the crossmodal condition. However, compared to younger participants, the artificial system performs worse at low stimulus intensities. Younger participants seem to employ more efficient crossmodal integration mechanisms than modelled in the proposed artificial neural networks.Our work creates a bridge between neurological research and embodied artificial neurocognitive systems and demonstrates how collaborative research might help to derive hypotheses from the allied field. Our results indicate that empirically-derived neurocognitive models can inform the design of future neurocomputational architectures. For crossmodal processing, sensory integration on lower hierarchical levels, as suggested for efficient processing in the human brain, seems to improve the performance of artificial neural networks.

Download Full-text

Crossmodal Pattern Discrimination in Humans and Robots: A Visuo-Tactile Case Study

Frontiers in Robotics and AI ◽

10.3389/frobt.2020.540565 ◽

2020 ◽

Vol 7 ◽

Author(s):

Focko L. Higgen ◽

Philipp Ruppel ◽

Michael Görner ◽

Matthias Kerzel ◽

Norman Hendrich ◽

...

Keyword(s):

Neural Networks ◽

Artificial Neural Networks ◽

State Of The Art ◽

Pattern Discrimination ◽

Tactile Sensing ◽

Processing Stream ◽

Sensing Technology ◽

Crossmodal Processing ◽

Crossmodal Perception ◽

Artificial Neural

The quality of crossmodal perception hinges on two factors: The accuracy of the independent unimodal perception and the ability to integrate information from different sensory systems. In humans, the ability for cognitively demanding crossmodal perception diminishes from young to old age. Here, we propose a new approach to research to which degree the different factors contribute to crossmodal processing and the age-related decline by replicating a medical study on visuo-tactile crossmodal pattern discrimination utilizing state-of-the-art tactile sensing technology and artificial neural networks (ANN). We implemented two ANN models to specifically focus on the relevance of early integration of sensory information during the crossmodal processing stream as a mechanism proposed for efficient processing in the human brain. Applying an adaptive staircase procedure, we approached comparable unimodal classification performance for both modalities in the human participants as well as the ANN. This allowed us to compare crossmodal performance between and within the systems, independent of the underlying unimodal processes. Our data show that unimodal classification accuracies of the tactile sensing technology are comparable to humans. For crossmodal discrimination of the ANN the integration of high-level unimodal features on earlier stages of the crossmodal processing stream shows higher accuracies compared to the late integration of independent unimodal classifications. In comparison to humans, the ANN show higher accuracies than older participants in the unimodal as well as the crossmodal condition, but lower accuracies than younger participants in the crossmodal task. Taken together, we can show that state-of-the-art tactile sensing technology is able to perform a complex tactile recognition task at levels comparable to humans. For crossmodal processing, human inspired early sensory integration seems to improve the performance of artificial neural networks. Still, younger participants seem to employ more efficient crossmodal integration mechanisms than modeled in the proposed ANN. Our work demonstrates how collaborative research in neuroscience and embodied artificial neurocognitive models can help to derive models to inform the design of future neurocomputational architectures.

Download Full-text

Мethods of Machine Learning in Ophthalmology: Review

Ophthalmology in Russia ◽

10.18008/1816-5095-2020-1-20-31 ◽

2020 ◽

Vol 17 (1) ◽

pp. 20-31

Author(s):

D. D. Garri ◽

S. V. Saakyan ◽

I. P. Khoroshilova-Maslova ◽

A. Yu. Tsygankov ◽

O. I. Nikitin ◽

...

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Neural Networks ◽

Artificial Neural Networks ◽

Laboratory Data ◽

Age Related Macular Degeneration ◽

Digital Data ◽

Age Related ◽

Artificial Neural ◽

Sensitivity Specificity

Machine learning is applied in every field of human activity using digital data. In recent years, many papers have been published concerning artificial intelligence use in classification, regression and segmentation purposes in medicine and in ophthalmology, in particular. Artificial intelligence is a subsection of computer science and its principles, and concepts are often incomprehensible or used and interpreted by doctors incorrectly. Diagnostics of ophthalmology patients is associated with a significant amount of medical data that can be used for further software processing. By using of machine learning methods, it’s possible to find out, identify and count almost any pathological signs of diseases by analyzing medical images, clinical and laboratory data. Machine learning includes models and algorithms that mimic the architecture of biological neural networks. The greatest interest in the field is represented by artificial neural networks, in particular, networks based on deep learning due to the ability of the latter to work effectively with complex and multidimensional databases, coupled with the increasing availability of databases and performance of graphics processors. Artificial neural networks have the potential to be used in automated screening, determining the stage of diseases, predicting the therapeutic effect of treatment and the diseases outcome in the analysis of clinical data in patients with diabetic retinopathy, age-related macular degeneration, glaucoma, cataracts, ocular tumors and concomitant pathology. The main characteristics were the size of the training and validation datasets, accuracy, sensitivity, specificity, AUROC (Area Under Receiver Operating Characteristic Curve). A number of studies investigate the comparative characteristics of algorithms. Many of the articles presented in the review have shown the results in accuracy, sensitivity, specificity, AUROC, error values that exceed the corresponding indicators of an average ophthalmologist. Their introduction into routine clinical practice will increase the diagnostic, therapeutic and professional capabilities of a clinicians, which is especially important in the field of ophthalmic oncology, where there is a patient survival matter.

Download Full-text