AI energized hydrogel design, optimization and application in biomedicine

Abstract

Traditional hydrogel design and optimization methods usually rely on repeated experiments, which is time-consuming and expensive, resulting in a slow-moving of advanced hydrogel development. With the rapid development of artificial intelligence (AI) technology and increasing material data, AI-energized design and optimization of hydrogels for biomedical applications has emerged as a revolutionary breakthrough in materials science. This review begins by outlining the history of AI and the potential advantages of using AI in the design and optimization of hydrogels, such as prediction and optimization of properties, multi-attribute optimization, high-throughput screening, automated material discovery, optimizing experimental design, and etc. Then, we focus on the various applications of hydrogels supported by AI technology in biomedicine, including drug delivery, bio-inks for advanced manufacturing, tissue repair, and biosensors, so as to provide a clear and comprehensive understanding of researchers in this field. Finally, we discuss the future directions and prospects, and provide a new perspective for the research and development of novel hydrogel materials for biomedical applications.

Keywords: Artificial intelligence; Biomedicine application; Design; Hydrogel; Optimization.

Graphical abstract

Fig. 1

(A) The development history of hydrogels. (B) Hydrogels for various tissue engineering applications, such as repair the tissues of bone, cartilage, oral, meniscus, muscle, skin, cardiac, cornea, neural, vascular, hepatic, gastric, and so on. (C) Problems and challenges in the design, optimization and biomedical applications of hydrogels.

Fig. 2

(A) The main types of ML, including supervised learning, unsupervised learning, and reinforcement learning. (B) Several common ML models offer the tools and methodologies for AI systems.

Fig. 3

(A) Illustration of bio-inks development according to mathematical strategies to predict tissue engineering issues. (B) Schematic diagram for developing 3D printable naturally derived bio-inks. (C) Scheme illustrating the preparation of the cell-laden 3D biomimetic structure composed by low viscosity hydrogel (1% collagen) as cell vehicles and high viscosity hydrogel bio-inks as frameworks. Adapted with permission [73]. Copyright © 2020, IOP Publishing Ltd.

Fig. 4

(A) Diagram of optimizing hydrogel formulations by using ML. Rheological information of hydrogels must be first collected through experiments or obtained from literature. Then, the collected information is fed into ML-based algorithms. Robust algorithms can predict the properties of different components and help optimize the formulation of hydrogels. (B) Illustration of a method for rapidly and autonomously characterizing the rheological properties of hydrogels by high-throughput through automated sensing and physically guided supervised ML. Adapted with permission [76]. Copyright © 2022, Elsevier Ltd. (C) The impact of innovative autonomous sensors and data-driven HTC strategies on achievable screening throughput compared to conventional characterization methods. Adapted with permission [76]. Copyright © 2022, Elsevier Ltd.

Fig. 5

(A) Overview of the AI-energized materials discovery. Adapted from CC-BY open access publications [88]. Copyright © 2016, Elsevier Ltd. (B) An illustration of material discovery, using titanium alloy as an example. (i) Properties prediction processes, such as filtering of martensite start (Ms) temperature and combining maps. (ii) Constraining the prediction in βLow-assisted alloy development for low-modulus and low-β stabilizer β-titanium alloys. Adapted with permission [89]. Copyright © 2020, Elsevier Ltd.

Fig. 6

(A) AI strategies, such as RF, ANNs, and SVM, have been applied at multiple steps, including (1) forecasting hydrogel formation according to previous ingredients, (2) improving 3D printing performance, (3) adjusting injectable properties, (4) optimizing supporting functions, (5) optimizing and forecasting drug release curves, and (6) upgrading clinical effects, to enhance the preparation of hydrogel drug delivery systems. Adapted with permission [108]. Copyright © 2022, Elsevier Ltd. (B) Synthesis, analysis and optimization of specific injectable hydrogels for delivering proteins by the high-throughput strategies. Adapted with permission [119]. Copyright © 2019 American Chemical Society.

Fig. 7

(A) Selected administration routes of long-acting injectables formulations approved by FDA. (B) Typical trial-and-error loop conventionally used in the development of classical long-acting injectables formulation. (C) Training and analyzing ML models to enhance the developing of new long-acting injectables systems, termed “Data-driven long-acting injectables formulation development”. Adapted with permission [122]. This is an open access article distributed under the terms of the Creative Commons CC BY license.

Fig. 8

(A) The 3D bioprinting process includes pre-bioprinting, bioprinting, and post-bioprinting. (B) Constructs printed from bio-inks and applications and schematic of pneumatic-, piston-, and screw-driven printing. Adapted with permission [124]. Copyright © 2023 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. (C) AI can be involved in various stages of 3D bio-printing in different ways, including AI-energized fabrication and data-driven fabrication. The illustration indicates the bio-printing procedure for different levels AI involvement in fabrication, such as 3D printing without AI, open-loop AI printing, closed-loop AI printing and predictive AI printing. Adapted with permission [132]. Copyright © 2020, Springer Nature Limited.

Fig. 9

(A) Illustration of developing bioinspired 3D-printed hydrogels by DNA-induced biomineralization. (B) ML modeling applying Gaussian process regression to predict the printability score of the prepared hydrogel bio-inks. (C) Various scores based on variable nozzle size, temperature, pneumatic pressure, and FSA concentration. Adapted with permission [134]. This is an open access article distributed under the terms of the Creative Commons CC BY license.

Fig. 10

(A) Description of the wound recognition. Personalized 3D printed hydrogel wound dressings can match the shape and size of the wound by recognizing wound characteristics. (B) Flow chart of intelligent wound monitoring with multifunctional hydrogel dressings, including (i) wound recognition, (ii) real-time status supervising and (iii) customized wound management. Adapted with permission [143]. Copyright © 2022, Elsevier Ltd.

Fig. 11

Illustration of the AI-HTPCSS for fast screening of the optimized extrusion bio-printing conditions of a given bio-printer and bio-ink combination. (A) Overview of the AI-HTPCSS. (B) The morphologies of extrusion patterns under different bio-printing parameters, such as droplets, lines of droplets, or lines. (C) Graphical representing the line uniformities of extruded patterns under different bio-printing parameters. (D) Optimized bio-printing conditions are sued to prepare multi-layer 3D mesh-like hydrogel scaffolds with different structures. (E) The application of the optimal printed hydrogel dressings for accelerating the healing of diabetic wounds. Adapted with permission [148]. Copyright © 2022, Wiley-VCH GmbH.

Fig. 12

(A) Overview of high-throughput multi-channel feeder. (B) Schematic diagram of material synthesis reactions. (C) The device structure of the hydrogel-based capacitive sensor and its response to applied pressure before cutting and after self-healing are analyzed. Adapted with permission [157]. Copyright © 2021, Wiley-VCH GmbH.

Fig. 13

The current development trends and potential applications of AI-energized design and optimization of hydrogels in biomedicine.