Deep Learning-Enhanced Metal-Organic Framework E-Skin for Health Monitoring

Academic Background

Electronic skin (e-skin) is a technology capable of sensing physiological and environmental stimuli, mimicking human skin functions. In recent years, the potential applications of e-skin in fields such as robotics, sports science, and healthcare monitoring have gradually emerged. However, current e-skin technology faces several challenges: first is the realization of multifunctionality, i.e., how to simultaneously detect multiple physiological signals (such as biomolecules, motion signals, etc.) in a single device; second is the issue of signal differentiation, particularly how to accurately distinguish and identify different signals when detecting multiple stimuli simultaneously.

Traditional multifunctional e-skin typically requires the integration of various sensing materials, which not only increases the complexity of manufacturing but may also lead to unstable device performance. Additionally, existing e-skin faces limitations in terms of signal-to-noise ratio, sensitivity, and stability. Therefore, the development of a high-performance, multifunctional, and easily manufacturable e-skin has become a research focus.

Metal-Organic Frameworks (MOFs) are considered a promising sensing material due to their unique electrochemical properties, open-pore structures, and high customizability. However, the preparation of MOF films often faces challenges related to stability and continuity. To overcome these limitations, researchers have explored new fabrication methods, such as Atomic Layer Deposition (ALD), to improve the quality and performance of MOF films.

Paper Source

This paper was authored by Xinyi Ke, Yifan Duan, Yifei Duan, et al., from multiple research institutions including Fudan University and Donghua University. It was published in the journal Device on April 18, 2025 (DOI: 10.1016/j.device.2024.100650). The paper’s topic is “Deep Learning-Enhanced Metal-Organic Framework E-Skin for Health Monitoring.”

Research Overview

The study proposes a multifunctional e-skin based on self-locking chitosan films and conductive MOF films, capable of simultaneously detecting lactic acid, glucose, and motion signals. By leveraging the Transformer neural network in deep learning, the e-skin can accurately recognize and differentiate various signals, particularly facial micro-expressions. This technology represents a significant advancement in the field of e-skin, enhancing its comprehensiveness and precision in health monitoring.

Research Process

1. Material Preparation and Characterization:

   • First, the research team extracted chitosan fibers from crab shells to prepare self-locking chitosan films.
   • Then, zinc oxide (ZnO) nanomembranes were deposited on the chitosan films via Atomic Layer Deposition (ALD) to serve as an induction layer for MOF film growth.
   • Next, a conductive MOF (Cu-HHTP) film was grown on the chitosan films using ALD-assisted assembly, ensuring its uniformity and adhesion.

2. E-Skin Fabrication:

   • The Cu-HHTP film was combined with the chitosan film to form the basic structure of the e-skin.
   • Gold electrodes were deposited on the e-skin to create a sensor array for signal output.

3. Performance Testing:

   • The research team conducted detailed tests on the mechanical properties, electrochemical performance, and sensing capabilities of the e-skin.
   • Tests included sensing of biomolecules (lactic acid and glucose), detection of electrocardiogram (ECG) signals, and recognition of motion signals.

4. Data Analysis and Deep Learning Application:

   • The Transformer neural network was used to process and analyze the data output by the e-skin, enabling the recognition of micro-expressions and differentiation of various stimulus signals.

Key Findings

1. Biomolecular Sensing Performance:

   • The e-skin exhibited high sensitivity and low detection limits for lactic acid and glucose. The sensitivity for lactic acid detection was 11,480 µA mM⁻¹ cm⁻², with a linear range of 0.001–2 mM and a detection limit of 0.335 µM.
   • Glucose detection also demonstrated a good linear range and stability, with high anti-interference capability in the presence of interfering substances.

2. Motion Sensing Performance:

   • The e-skin was able to detect subtle human motions, such as bending, stretching, and pressure changes.
   • In a test of over 800 loading-unloading cycles, the e-skin exhibited excellent long-term stability.

3. Micro-Expression Recognition:

   • With the help of the Transformer neural network, the e-skin could accurately recognize different micro-expressions, such as crying, smiling, and pouting.
   • The application of deep learning technology improved the precision of signal analysis, enabling the recognition of complex motion patterns.

Conclusion and Outlook

This study prepared a multifunctional MOF-based e-skin using ALD-assisted assembly methods, demonstrating its excellent performance in biomolecular sensing, motion detection, and micro-expression recognition. With the help of the Transformer neural network, the e-skin could precisely differentiate various physiological signals, further enhancing its application value in health monitoring.

Future research may focus on the following aspects: first, improving the scalability of e-skin for mass production; second, introducing self-healing and adaptive materials to further enhance the durability and adaptability of the device. With the further development of machine learning models, e-skin is expected to evolve into an intelligent system capable of autonomously analyzing complex physiological data, advancing healthcare and human-computer interaction technologies.

This study not only provides new ideas for the design of multifunctional e-skin but also lays the foundation for the future development of health monitoring technology. By combining MOF materials with deep learning, significant progress has been made in the sensitivity, multifunctionality, and signal processing capabilities of e-skin, offering broad application prospects.