Shiming Zhang

Hydrogels are widely used in biomedical and bioelectronic applications owing to their tissue-like properties, including biocompatibility, softness and three-dimensional architecture. In recent years, semiconducting behaviour was demonstrated in hydrogels through the network design of π-conjugated polymers, extending their potential to advanced electronic applications such as transistors. In this Review, we provide an overview of the design, fabrication, characterization and benchmarking standards of π-conjugated hydrogel semiconductors, a rapidly evolving field requiring interdisciplinary knowledge across organic electronics, electrochemistry and soft materials. We demonstrate how to understand and regulate ion and electron transport, as well as their interactions, both thermodynamically and kinetically, in these π-conjugated supramolecular systems. Finally, we envision the potential of these materials to advance spatiotemporal biological research, wearable healthcare, implantable medicine and beyond.

Objectives: This systematic review aims to evaluate the clinical applicability of automated 3D facial asymmetry assessment methods based on 3D facial scans. Search Methods: A comprehensive search of electronic databases (PubMed, Web of Science, EMBASE, Medline, and Scopus) and manual literature searches were conducted in April 2025. Selection Criteria: Studies that were published in English and evaluated the validity or reliability of automated asymmetry assessment methods in 3D facial scans for medical or biological settings were included. Data Collection and Analysis: Two reviewers independently screened the articles for eligibility. Risk of bias was assessed using QUADAS-2, and the certainty of evidence was graded using the Grading of Recommendations Assessment, Development, and Evaluation framework. Results: Fourteen studies met the inclusion criteria and were analyzed for methodology, validity, and reliability. Methodologies for assessing facial asymmetry were categorized into four approaches: landmark-based, depth-stratified, original-mirror alignment, and template-based approaches. While landmark and depth-stratified methods rely on sparse data, original-mirror and template-based methods enable comprehensive surface analysis. Six studies evaluating validity against alternative methods, synthetic ground truth, or human ratings consistently demonstrated moderate-to-strong correlation coefficients and classification accuracy. Reliability was examined across nine studies using repeated measurements and multi-observer designs, generally showing minimal measurement variation. Notably, methodological analysis revealed that original-mirror alignment, typically implemented using unconstrained iterative closest point (ICP)-based best-fit registration, is susceptible to registration errors (the “Pinocchio effect”) in cases of severe asymmetry, whereas template-based methods mitigate this through correspondence transfer and weighted registration strategies that stabilize anatomical alignment. Limitations: The review is limited by a high risk of bias in primary studies and significant methodological heterogeneity. Conclusions: Despite “very low” certainty evidence, template-based approaches that transfer anatomical correspondence and apply weighted registration appear preferable due to their robustness against localized deformities. Conversely, original-mirror alignment, typically implemented via unconstrained ICP, remains a practical alternative for mild asymmetry. Future research should prioritize end-to-end deep learning automation, dynamic analysis, and the development of accessible, open-source tools to bridge the gap between technical innovation and routine clinical practice. Registration: PROSPERO (CRD420251025105).

Continuous glucose monitoring (CGM) is a commercially available biochemical sensor for diabetes management. A growing trend in this field is the transition from long needles to microneedles to reduce pain and avoid bleeding. A promising strategy for microneedle‐based CGM involves using hydrogel as a filler to bridge the interstitial fluid and externally deployed glucose sensors on the skin surface. However, the glucose diffusion dynamics, critical to the sensing process, remain poorly understood. In this study, we investigate the dynamics of glucose diffusion between subcutaneous tissue and hydrogel‐filled microneedles. Specifically, we examine how hydrogel thickness, microneedle size, and glucose concentration influence diffusion. This study provides an experimental reference to the wearable biosensor community researching microneedle‐based CGM.

Skin-integrated transistors have been explored for wearable sensing, computing and programmable synthetic bioelectronics. However, their high modulus, low sweat permeability, and limited scalability result in multidimensional mismatches with the real skin. Here, we introduce skin-affinitive organic electrochemical transistors (SA-OECTs) based on all-hydrogel, designed to simultaneously overcome these limitations. Our all-hydrogel SA-OECTs incorporates a semiconducting hydrogel channel and thus capable of conducting active sensing and computing functions. This unified all-hydrogel strategy yields the SA-OECTs with skin range softness (10-500 kPa), thickness (500 nm-500 μm), biocompatibility, breathability - a combination of properties unattainable with conventional transistors - while achieving tunable transistor characteristics via ion-electron conversion, achieving an on/off ratio exceeding 10 ⁴ . Additionally, these all-hydrogel SA-OECTs exhibit a substantial enhancement in operational speed over conventional poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) based OECTs of equivalent dimensions. Leveraging these attributes, we constructed a SA biocomputing platform using an all-hydrogel SA-OECTs array, paving the way for biohybrid AI that bridges artificial and biological intelligence.

Synthetic bioelectronics is rapidly advancing, propelled by breakthroughs in synthetic biology and bioelectronics. This convergence is key to next‐generation wearable and implantable devices, enabling seamless integration with living systems. Here, we introduce an enzymatic hydrogel electrode (GelZymes) developed via a synthetic bioelectronic strategy to overcome the mechanical and interfacial limitations of conventional enzyme electrodes. GelZymes deliver two core advances: i) a monolithic and scalable 3D architecture that unifies the enzyme membrane and electrode, simplifying fabrication and eliminating interfacial instability; and ii) tissue‐like viscoelasticity—combining stretchability and adhesiveness—rarely achievable with rigid enzyme membranes. GelZymes are synthesized through three steps: engineering a stretchable, mixed‐conducting 3D hydrogel; implementing an enzyme‐compatible, cascading crosslinking scheme to immobilize enzymes within the network; and balancing the trade‐off between electronic/ionic conductivity and the density of redox‐active enzyme sites to maximize bio‐electrochemical performance. We further show that GelZymes enable a shift from invasive, tissue‐interfaced biosensing to noninvasive, tissue‐integrated biosensing, offering a practical pathway to bridge current biosensor technologies with living systems.

Effective blood glucose management is an increasing demand worldwide. Traditional solutions separate glucose detection and insulin delivery, which is less efficient compared to emerging closed‐loop wearable systems controlled by continuous glucose monitors (CGMs). However, CGM‐controlled systems raise new safety risks, as false CGMs readings can cause insulin overdose, which results in hypoglycemia and fatal consequences. This work proposes a concept of a dual closed‐loop insulin delivery system (DuoLoop) to mitigate the risk issue of CGM‐controlled systems. The first closed‐loop is automated insulin delivery controlled by CGM. The second closed‐loop is the controlled release of glucose‐responsive insulin (GRI), whose release rate depends on actual glucose levels. A customized algorithm is trained and embedded into the wearable CGMs for edge computing. The DuoLoop system shows improved safety in preliminary in vivo test (longer normoglycemia durations, 98.82% vs 92.10%), encouraging its deployment toward precision diabetes care.

The rapid expansion of AI models has intensified concerns over energy consumption. Analog in-memory computing with resistive memory offers a promising, energy-efficient alternative, yet its practical deployment is hindered by programming challenges and device non-idealities. Here, we propose a software-hardware co-design that trains randomly weighted resistive-memory neural networks via edge-pruning topology optimization. Software-wise, we tailor the network topology to extract high-performing sub-networks without precise weight tuning, enhancing robustness to device variations and reducing programming overhead. Hardware-wise, we harness the intrinsic stochasticity of resistive-memory electroforming to generate large-scale, low-cost random weights. Implemented on a 40 nm resistive memory chip, our co-design yields accuracy improvements of 17.3% and 19.9% on Fashion-MNIST and Spoken Digit, respectively, and a 9.8% precision-recall AUC improvement on DRIVE, while reducing energy consumption by 78.3%, 67.9%, and 99.7%. We further demonstrate broad applicability across analog memory technologies and scalability to ResNet-50 on ImageNet-100.