Early Insights into AI and Machine Learning Applications in Hydrogel Microneedles: A Short Review

Abstract

Hydrogel microneedles (HMNs) act as non-invasive devices that can effortlessly merge with the human body for drug delivery and diagnostic purposes. Nonetheless, their improvement is limited by intricate and repetitive issues related to material composition, structural geometry, manufacturing accuracy, and performance enhancement. At present, there are only a limited number of studies accessible since artificial intelligence and machine learning (AI/ML) for HMN are just starting to emerge and are in the initial phase. Data is distributed across separate research efforts, spanning different fields. This review aims to tackle the disjointed and narrowly concentrated aspects of current research on AI/ML applications in HMN technologies by offering a cohesive, comprehensive synthesis of interdisciplinary insights, categorized into five thematic areas: (1) material and microneedle design, (2) diagnostics and therapy, (3) drug delivery, (4) drug development, and (5) health and agricultural sensing. For each domain, we detail typical AI methods, integration approaches, proven advantages, and ongoing difficulties. We suggest a systematic five-stage developmental pathway covering material discovery, structural design, manufacturing, biomedical performance, and advanced AI integration, intended to expedite the transition of HMNs from research ideas to clinically and commercially practical systems. The findings of this review indicate that AI/ML can significantly enhance HMN development by addressing design and fabrication constraints via predictive modeling, adaptive control, and process optimization. By synchronizing these abilities with clinical and commercial translation requirements, AI/ML can act as key facilitators in converting HMNs from research ideas into scalable, practical biomedical solutions.

1. Introduction

HMNs are developing as next-gen tools for transdermal drug delivery, with uses that include insulin delivery, vaccine administration, continuous glucose monitoring, and precise agricultural sensing. They offer a non-invasive substitute to hypodermic needles, enhancing patient compliance and accessibility [1,2]. Integrating the minimal invasiveness of microneedles with the tunability of hydrogels, HMNs serve in drug encapsulation, wound healing, and regenerative medicine [3,4].

Notwithstanding this potential, HMN development encounters ongoing technical challenges: a shortage of biocompatible hydrogels with reliable swelling and degradation characteristics, issues with maintaining consistent tip sharpness and pore structure, deformation risks during production, and complexities in regulating drug-release kinetics for multifaceted therapeutic payloads. Conventional trial-and-error methods are inefficient, resource-demanding, and inadequate for understanding nonlinear connections among material characteristics, design, and in vivo outcomes. Sofian et al. [5] applied ML (Reinforcement learning (RF), Gaussian Processes, convolutional neural networks (CNNs), Bayesian optimization) in the production of bioinks to enhance the printability, strength, and porosity of alginate-carrageenan hydrogels. Wang et al. [6] utilized deep learning and cloud-driven AI in wearable microneedle sensors to enable real-time calibration and detection of multiple fluid biomarkers. AI/ML therefore facilitate extensive analysis, forecasting models, and automated enhancement to tackle these obstacles.

At present, there are limited studies accessible since AI/ML for HMN is still developing and in its initial phase, with data dispersed among separate fields. AI/ML can analyze extensive datasets that exceed traditional techniques [7] and enhance various factors such as hydrogel formulation, shape, and drug release [8]. This advancement necessitates cooperation between data scientists, engineers, and clinicians to guarantee that models are scientifically valid and clinically pertinent [9]. In spite of this potential, the majority of studies concentrate on one specific stage—like material characterization, tip optimization, or defect detection—rather than encompassing multiple stages. This disjointed method hinders iterative feedback across stages, where material characteristics, structural effectiveness, and clinical results could be improved. Without this integration, AI’s capability to speed up timelines, minimize variability, and facilitate personalized, adaptive HMN systems is not achieved.

The goal of this review is to tackle the divided and limited scope of current studies at an initial stage. To address this gap, this review provides a comprehensive synthesis of AI/ML applications covering the entire HMN development continuum. Our discussion revolves around five key thematic areas: (1) materials and manufacturing, (2) diagnostics and treatment, (3) medication delivery, (4) drug innovation, and (5) health/agriculture monitoring. We outline key algorithms and integration methods while also pinpointing outstanding technical and translational challenges. We suggest a systematic development route linking material discovery, structural enhancement, production, and biomedical efficacy toward smooth AI integration, providing a clear pathway for advancing HMNs from research ideas to clinical and market implementation.

Key applications consist of predictive modeling for hydrogel swelling, degradation, and drug release [10], enhancement of hydrogel mechanical properties [11,12], supporting polymer formulation [13], and optimizing microneedle–tissue configurations [14]. In production, evolutionary algorithms and reinforcement learning improve fabrication precision and consistency [15,16]. In addition to fabrication, AI facilitates real-time observation and closed-loop feedback, employing comparable approaches in wound-healing hydrogels and biosensors for dynamic diagnosis and therapy [17,18,19,20].

HMNs provide a minimally invasive method for accurate drug delivery, diagnostics, and biosensing, yet their advancement is obstructed by complex design, material, and production issues. This review seeks to establish a thorough basis for progressing HMN technologies from theoretical models to clinically and commercially feasible solutions by mapping AI applications across five thematic areas and suggesting a structured development pathway.

2. Methodology

The reviewed literature on hydrogel microneedles (HMNs) can be organized into five thematic domains, each representing a distinct application of AI and ML. These domains highlight critical aspects of HMN development, including material design, diagnostics, therapeutics, drug delivery, drug development, and sensor-based monitoring for health and agriculture. Based on this categorization,

Table 1. Thematic grouping of AI/ML applications in HMNs.

	Thematic Grouping	Thematic Group Description
1	Material and Microneedle Design	ML and simulation tools enable predictive material design (e.g., hydrogels, nanocomposites) for specific mechanical, electrical, and drug-release properties; accelerates bio-device innovation.
2	Microneedles for Diagnostics and Therapy	Integration of AI with microneedle devices for real-time biomarker diagnostics, adaptive drug release, and closed-loop feedback systems; enables personalized, minimally invasive therapy via biomarker sensing and intelligent control.
3	Microneedles for Drug Delivery	Focus on hydrogel-forming and polymer-based microneedles that are biocompatible, dissolvable, and suitable for sustained drug delivery; advances in drug dosing enhance comfort and efficacy.
4	Microneedles for Drug Development	AI tools (ANN, CNN, etc.) applied to optimize drug formulations, delivery systems, and bioavailability; revolutionizes pharma R&D with predictive analytics and rapid prototyping.
5	Microneedles Sensors for Health and Agriculture	AI-powered biosensors embedded in wearable tech monitor plant/human hydration, stress etc., offering real-time, adaptive health insights, especially for agriculture, sports and chronic care.

presents a detailed overview of the five thematic domains, summarizing their focus and the contributions of AI/ML tools in each area.

This research employs a systematic thematic synthesis approach to explore the integration of AI/ML into microneedle and hydrogel technologies. The evaluation is grounded in review articles, each chosen for its emphasis on AI/ML uses in material innovation, drug delivery, or wearable diagnostics. The articles were classified into five thematic categories, as detailed in the included reference Table 1: Material and Microneedle Design, Microneedles for Diagnostics and Therapy, Microneedles for Drug Delivery, Microneedles for Drug Development, and Microneedle Sensors for Health and Agriculture.

2.1. Article Selection and Inclusion Criteria

Both review and research articles were included to maintain relevance and scientific rigor. Every article had to explicitly address the involvement of AI or ML regarding hydrogel materials, microneedle production, drug formulation, diagnostics, or intelligent sensor systems. The chosen articles needed to cover not just technical implementations but also obstacles and future pathways for AI/ML integration. Articles that did not have AI/ML relevance or did not address future applications were omitted. The final group of articles was pinpointed via full-text evaluation, guaranteeing an even distribution among fields like biomedicine, materials science, and intelligent sensing.

2.2. Thematic Grouping Developmental Pathways

After conducting a thorough text analysis of the chosen articles, the review employed a thematic grouping approach to organize the literature into five unique categories: (1) Material and Microneedle Design, (2) Microneedles for Diagnostics and Therapy, (3) Microneedles for Drug Delivery, (4) Microneedles for Drug Development, and (5) Microneedle Sensors for Health and Agriculture.

This classification was established according to the primary research focus of each article, the AI methods discussed, the intended biomedical application, and the contextual difficulties encountered. Thematic categorization facilitated a more organized synthesis of the field’s expansiveness and permitted direct comparisons of technological advancements, constraints, and suggested future tasks.

2.3. Analytical Dimensions

Articles within each thematic group were analyzed based on four key parameters:

• Thematic Group Description—A synthesized definition of the core AI/ML application within the domain.
• Author-Year Attribution—Each contributing author was assigned to only one group to maintain exclusivity and balance.
• Data Structure—Parameters for analytical comparison that enable systematic evaluation across groups while preserving domain-specific depth.
• Future Work—Suggestions from the authors for advancing the field, harmonized across thematic groups.

Thematic summary tables (Table 2, Table 3, Table 4 and Table 5 in Section 3) provide a structured comparative assessment of review studies, detailing AI/ML components, methodologies, and applications. This approach allows identification of patterns or connections that may be overlooked when examining individual articles.

Each table is organized with the following columns:

• Author(s) and Year—Identifies researchers and publication date, placing studies in chronological context.
• AI/ML—Indicates the use of AI, ML, or both.
• HMN—Specifies the type of material or platform, such as microneedles or hydrogels.
• AI/ML Techniques & Algorithms—Details computational models and data-driven approaches (e.g., CNNs, SVM, PCA, RL).
• Key AI/ML Role/Purpose/Application—Defines primary functions of AI/ML, including predictive modeling, formulation optimization, biosignal analysis, or drug release profiling.
• AI/ML Integration in HMN—Explains how AI/ML is embedded in microneedle systems to enhance functionality.
• AI-Enhanced HMN Features—Highlights functional improvements, such as optimized material selection, microneedle design, enhanced drug release, predictive alerts, or adaptive performance.
• Materials Used—Lists polymers, nanomaterials, or conductive substances employed (e.g., PVA, chitosan, graphene, CNTs).
• AI-Enhanced Targeted Design Features—Describes AI-enabled customization of performance metrics, such as hydrogel porosity, degradability, drug release kinetics, or mechanical-electrical synergy.
• Dataset/Data Type/Size—Provides the nature and scale of data used, including experimental, simulated, field-collected, or real-time streaming data.
• Targeted Application—Defines practical use, including drug delivery, biosensing, wound healing, drug discovery, or precision agriculture.
• Target Drug Parameter—Lists compounds or physiological markers studied (e.g., glucose, doxorubicin, environmental toxins).
• Targeted Parameter/Sickness—Specifies targeted diseases or conditions (e.g., diabetes, cancer, crop stress, environmental contamination).
• Results, Limitations, and Future Work—Summarizes performance indicators (accuracy, sensitivity), acknowledged limitations (data availability, computational cost, biocompatibility), and anticipated research directions (closed-loop optimization, clinical testing, AI-health integration).

Together, these columns provide a comprehensive framework for assessing AI-driven innovations in multi-domain systems involving materials and microneedle technologies, enabling readers to compare studies systematically and identify gaps for future research.

2.4. Methodological Strengths and Limitations

The strength of this thematic approach is its capacity to chart the AI/ML landscape within microneedle research while maintaining domain relevance. Nonetheless, constraints involve possible thematic redundancies and the omission of experimental research, which might highlight recent advancements that have not been assessed. Notwithstanding this, the approach provides a thorough and adaptable developmental pathway for forthcoming meta-analyses in AI-augmented biomedical technologies.

3. Developmental Pathway of Five Thematic Groupings

3.1. Thematic Group 1: Material and Microneedle Design

AI/ML tools are transforming material design by enabling predictive modeling of hydrogel properties like swelling, strength, and drug release. This thematic group covers nine studies using algorithms such as CNN, FEM, and Bayesian Optimization. Table 2A,B summarize AI roles, integration strategies, materials used, targeted applications, performance outcomes, and future directions in microneedle design and function.

Table 2. (A) AI/ML Techniques and Design Strategies in Microneedle Material Innovation. (B) Materials, Datasets, Applications, and Outcomes in Microneedle Material Studies.

(A) Thematic Group 1: Material and Microneedle Design
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques& Algorithms	Key AI/ML Role/ Purpose Application	AI/MLIntegration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	He W et al. [21]	ML	HMN	XGBoost, SVM, RF, PCA, K-means, CNN, RNN, GNN, GAT, BO.	Material screening, drug release, pain prediction, stress analysis	Pain minimization via FEA-COMSOL, BO.; CNN for defect detection	Stability, defect detection, optimized length, reduced pain, improved flow & geometry	Design, QC, material selection, geometry, stress, flow, comfort
2	Teodoro et al. [22]	Both	HMN	Predictive modeling, generative design, supervised ML, FEM	Defect prediction, design optimization, signal tuning	Predictive modeling and performance	Geometry, signal response	Geometry, drug release kinetics, Insertion force, signal mapping
3	Negut & Bita [10]	Both	Hydrogel	DOE, PCA ANN, SVM, RF, CNN, GAN, DNN	Modeling, formulation optimization, release kinetics	Virtual screening, ML-guided synthesis, gelation, print QC	Mechanical strength, precise release, swelling, degradation	Porosity, biodegradability, swelling, printability, delivery tuning
4	Finster et al. [23]	Both	Hydrogel	ANN, DNN, XGBoost, PCA, DDPG, DFT, FEA, MD, CFD, SVM, RF, NN, GA	Optimize formulations, predict properties, automate bioprinting	Conductive hydrogel design, real-time sensing, ML-guided tuning	Mechanical strength, precise release, electrical, adhesive properties	Porosity, conductivity, mech–electrical synergy, biocompatibility
5	Ji Wang et al. [24]	Both	HMN	AlphaFold, ML design tool	mRNA sequence design and structure prediction	AlphaFold for LNP-mRNA structure optimization	Enhanced wound healing via mRNA design	Predictive targeting for pro-healing proteins
(B) Thematic Group 1: Material and Microneedle Design
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/Biosensing Parameter	Targeted Parameter/ Sickness	Results	Limitations	Future Work
1	He W et al. [21]	Hydrogel, sugar, Silicon, glass, metals, polymers, sugar, ceramics	Simulated (30 k), empirical (2.4 k)	Drug delivery, biosensing, wound healing, dermatology	Triamcinolone, glucose, Minoxidil, TA, acyclovir, AZA, MTX, VEGF, PBNs, SD-208, EGCG, quercetin, etc.	Diabetes, Alopecia, acne, herpes, lupus, psoriasis, wounds, cancer	CNN = 96% accuracy, GAT = 8.3 × 10−5 MPa MSE, optimized flow & pain scores	Limited clinical trials, High cost (MEMS, TPP), complexity, data volume	IoT integration, personalized models, real-time ML feedback, explainable AI
2	Teodoro et al. [22]	PVA, Ag NPs, methacrylated polymers, Thermoplastics, UV polymers	Simulated + empirical	Biosensing, Diagnostics	Glucose, Lactate	Biosensing, drug delivery, Diabetes, Metabolic disorder	Thiram LOD: 10-710-7 M; 95% Cu release, Enhanced sensitivity and wearable design	Scalability, data gaps, biocompatibility, regulatory hurdles	AI-closed-loop design, smart integration
3	Negut & Bita [10]	Alginate, PNIPAAm, PEG NIPAAm, AAc, PEG, PLGA, PVA, Chitosan	Experimental (small), simulated (large)	Tissue engineering Drug delivery, biosensing	Antibiotic permeability Doxorubicin, glucose, stress	Bacterial infections Cancer, chronic wounds	Accurate modeling (93.5% CNN accuracy)	Data scarcity, model opacity, limited interpretability	DL for personalization, domain-integrated ML, scalable data
4	Finster et al. [23]	Chitosan, PNIPAm, carbon nanotubes, PEDOT:PSS, PVA, polyaniline, alginate	Experimental/Simulated	Drug delivery, wound healing, Bioelectronics Biosensing, Neuromodulation	Doxorubicin, PTEN inhibitors, Glucose, dopamine, electrical signals	Diabetic ulcers, infections, Neurodegeneration	95% gesture recognition, 40% fewer trials, Higher accuracy, faster optimization	Small datasets, high computational cost, Data scarcity, transfer-ability	Multi-objective RL, clinical trials, real-time ML, feedback loop
5	Ji Wang et al. [24]	MXene hydrogel MNs + HA + LNPs	AlphaFold mRNA design	Wound healing therapy	Triplet mRNA: PDGF, FGF-7, VEGF	Diabetic wound healing	92.33% wound closure in 12 days	Scalability, human trial lacking	Anti-inflammatory mRNA, real-time monitoring, clinical trials

Table 2. (A) AI/ML Techniques and Design Strategies in Microneedle Material Innovation. (B) Materials, Datasets, Applications, and Outcomes in Microneedle Material Studies.

(A) Thematic Group 1: Material and Microneedle Design
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques& Algorithms	Key AI/ML Role/ Purpose Application	AI/MLIntegration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	He W et al. [21]	ML	HMN	XGBoost, SVM, RF, PCA, K-means, CNN, RNN, GNN, GAT, BO.	Material screening, drug release, pain prediction, stress analysis	Pain minimization via FEA-COMSOL, BO.; CNN for defect detection	Stability, defect detection, optimized length, reduced pain, improved flow & geometry	Design, QC, material selection, geometry, stress, flow, comfort
2	Teodoro et al. [22]	Both	HMN	Predictive modeling, generative design, supervised ML, FEM	Defect prediction, design optimization, signal tuning	Predictive modeling and performance	Geometry, signal response	Geometry, drug release kinetics, Insertion force, signal mapping
3	Negut & Bita [10]	Both	Hydrogel	DOE, PCA ANN, SVM, RF, CNN, GAN, DNN	Modeling, formulation optimization, release kinetics	Virtual screening, ML-guided synthesis, gelation, print QC	Mechanical strength, precise release, swelling, degradation	Porosity, biodegradability, swelling, printability, delivery tuning
4	Finster et al. [23]	Both	Hydrogel	ANN, DNN, XGBoost, PCA, DDPG, DFT, FEA, MD, CFD, SVM, RF, NN, GA	Optimize formulations, predict properties, automate bioprinting	Conductive hydrogel design, real-time sensing, ML-guided tuning	Mechanical strength, precise release, electrical, adhesive properties	Porosity, conductivity, mech–electrical synergy, biocompatibility
5	Ji Wang et al. [24]	Both	HMN	AlphaFold, ML design tool	mRNA sequence design and structure prediction	AlphaFold for LNP-mRNA structure optimization	Enhanced wound healing via mRNA design	Predictive targeting for pro-healing proteins
(B) Thematic Group 1: Material and Microneedle Design
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/Biosensing Parameter	Targeted Parameter/ Sickness	Results	Limitations	Future Work
1	He W et al. [21]	Hydrogel, sugar, Silicon, glass, metals, polymers, sugar, ceramics	Simulated (30 k), empirical (2.4 k)	Drug delivery, biosensing, wound healing, dermatology	Triamcinolone, glucose, Minoxidil, TA, acyclovir, AZA, MTX, VEGF, PBNs, SD-208, EGCG, quercetin, etc.	Diabetes, Alopecia, acne, herpes, lupus, psoriasis, wounds, cancer	CNN = 96% accuracy, GAT = 8.3 × 10−5 MPa MSE, optimized flow & pain scores	Limited clinical trials, High cost (MEMS, TPP), complexity, data volume	IoT integration, personalized models, real-time ML feedback, explainable AI
2	Teodoro et al. [22]	PVA, Ag NPs, methacrylated polymers, Thermoplastics, UV polymers	Simulated + empirical	Biosensing, Diagnostics	Glucose, Lactate	Biosensing, drug delivery, Diabetes, Metabolic disorder	Thiram LOD: 10-710-7 M; 95% Cu release, Enhanced sensitivity and wearable design	Scalability, data gaps, biocompatibility, regulatory hurdles	AI-closed-loop design, smart integration
3	Negut & Bita [10]	Alginate, PNIPAAm, PEG NIPAAm, AAc, PEG, PLGA, PVA, Chitosan	Experimental (small), simulated (large)	Tissue engineering Drug delivery, biosensing	Antibiotic permeability Doxorubicin, glucose, stress	Bacterial infections Cancer, chronic wounds	Accurate modeling (93.5% CNN accuracy)	Data scarcity, model opacity, limited interpretability	DL for personalization, domain-integrated ML, scalable data
4	Finster et al. [23]	Chitosan, PNIPAm, carbon nanotubes, PEDOT:PSS, PVA, polyaniline, alginate	Experimental/Simulated	Drug delivery, wound healing, Bioelectronics Biosensing, Neuromodulation	Doxorubicin, PTEN inhibitors, Glucose, dopamine, electrical signals	Diabetic ulcers, infections, Neurodegeneration	95% gesture recognition, 40% fewer trials, Higher accuracy, faster optimization	Small datasets, high computational cost, Data scarcity, transfer-ability	Multi-objective RL, clinical trials, real-time ML, feedback loop
5	Ji Wang et al. [24]	MXene hydrogel MNs + HA + LNPs	AlphaFold mRNA design	Wound healing therapy	Triplet mRNA: PDGF, FGF-7, VEGF	Diabetic wound healing	92.33% wound closure in 12 days	Scalability, human trial lacking	Anti-inflammatory mRNA, real-time monitoring, clinical trials

He W et al. [21] utilized ML (NN, decision trees, Bayesian optimization) within a COMSOL–MATLAB environment to enhance MN production, decreasing anticipated pain by 37% while maintaining structural integrity. ML improved the prediction of skin penetration, monitoring pH levels in diabetic wounds, extracting interstitial fluid, and detecting biomarkers. Future efforts focus on developing adaptive MN therapeutic platforms that utilize IoT technology.

Teodoro et al. [22] broadened the emphasis on MN microfabrication, showcasing AI’s involvement throughout development phases. In 3D printing, AI identified and anticipated flaws, enhanced material choice, and optimized parametric design for durability, biocompatibility, and expense. Analytics powered by AI improved the processing of biosensor data for glucose, urea, and pH, whereas integration with IoT facilitated cloud monitoring without wires. Future directions involve developing biodegradable MNs and expanding ML datasets to tackle environmental and patient variability.

Negut & Bita [10] reviewed AI/ML strategies for hydrogel design and optimization, employing NN, SVM, RF, and CNNs to predict swelling, strength, and biocompatibility. RF models classified cardiomyocyte yields (75% accuracy) and predicted antibiotic skin permeability, while CNNs improved 3D bioprinting precision (93.5%). RL and GANs accelerated formulation, enabling self-healing, stimuli-responsive hydrogels for drug delivery, wound healing, and tissue engineering.

Finster et al. [23] highlighted AI/ML–simulation integration (ANN, DNN, CNN, XGBoost, SVR, PCA, DDPG) for hydrogel optimization in wound healing, tissue engineering, and biosensing. BO reduced bioink iterations by 40%, ANN/DNN improved mechanical and drug release properties, and CNNs achieved > 95% biosensing accuracy. RL balanced porosity and conductivity, with future work targeting multi-objective RL and expanded in vivo validation.

Ji Wang et al. [24] exhibited the integration of AI and biomaterials, creating MXene–hyaluronic acid hydrogel MNs infused with AI-enhanced triplet mRNA (PDGF, FGF-7, VEGF) for healing diabetic wounds. Protein design using AlphaFold (version 2.3.2) improved cell migration, collagen deposition, and angiogenesis. Preclinical findings indicated a 92.33% healing of wounds within 12 days. Ongoing obstacles encompass mass production and clinical validation, with forthcoming intentions for anti-inflammatory mRNA, combined wound monitoring, and regulatory progress.

Together, these studies combine computational modeling, sophisticated fabrication, and AI-enhanced optimization to speed up targeted material synthesis. This thematic group illustrates the importance of interdisciplinary cooperation between materials scientists, engineers, and data scientists in discovering new therapeutic options. To tackle issues like data limitations and model transparency, they suggest strong, data-informed approaches. Combined, their efforts indicate a transformative change in the creation of sophisticated materials and nanocomposites for medical use, providing both theoretical understanding and actionable methods for swift biomedical prototyping.

3.2. Thematic Group 2: Microneedles for Diagnostics and Therapy

Integration of AI with MN systems allows for immediate diagnostics and smart drug administration. This thematic group emphasizes the transformation of MNs into closed-loop, personalized healthcare tools through AI-driven logic, signal filtering, and adaptive control. Table 3A,B provide a summary of AI techniques, control methods, sensor incorporation, materials, targets, diseases, results, and potential future avenues in the studies reviewed.

Table 3. (A) AI/ML Techniques and Smart Integration in Diagnostic and Therapeutic Microneedles. (B) Materials, Datasets, Applications, and Performance Outcomes in Smart Diagnostic Microneedles.

(A) Thematic Group 2: Microneedles for Diagnostics and Therapy
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques &Algorithms	Key AI/ML Role/Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Ashraf et al. [25]	Both	MN	ML, DL, supervised/unsupervised, signal denoising	Noise reduction, biomarker prediction, sensor calibration, signal filtering, diagnostics	Integrated into wearable MNs for real-time diagnostics	Predictive sensing, personalized biosensing, improved accuracy, adaptive calibration	Signal calibration, multi-biomarker detection
2	S. Wang et al. [26]	Both	HMN	AI, ANN, Logic Encoding, IoT	Logic encoding, real-time monitoring, smart tracking, adaptive therapy	Programmable MNs, AI-driven control, therapeutic pathways	Adaptive release, biosensing, multi-therapy MNs, remote control	Multi-disease targeting, drug timing control
3	Merzougui et al. [27]	AI (proposed)	HMN	Not specified (general AI/ML for data analysis)	Enhance diagnostic accuracy, real-time POC analysis	AI for biomarker signal interpretation	Signal filtering, noise reduction, automated quantification	Data processing efficiency
4	Liu et al. [28]	Both	HMN	XGBoost, LSTM	Closed-loop inflammation monitoring	Real-time biosensing with integrated feedback	Theranostic system for RA	Prediction of cytokine levels & treatment response
5	Xiao et al. [29]	Both	HMN	MobileNet, PCA	Smartphone-assisted pH detection	AI-visualization via smartphone fluorescence	Real-time wound pH monitoring	Dynamic healing status identification
(B) Thematic Group 2: Microneedles for Diagnostics and Therapy
	Author (Year)	Materials Used	Data Type/Size	Targeted Application	Biosensing Parameter	Targeted Parameter/ Sickness	Results	Limitations	Future Work
1	Ashraf et al. [25]	Metals, semiconductors, polymers, colorimetric/electrochemical MN	Simulated/empirical (small data), High-dimensional time-series biosensor data	Biosensing, therapy, Non-invasive diagnostics, AI-guided monitoring	Glucose, lactate, cortisol, nucleic acids, pH, electrolytes	Glucose, Diabetes, cancer, inflammation, autoimmune disease	AI-enabled real-time monitoring with high-fidelity, adaptive biosensing	High accuracy, limited scalability; ISF variability, signal drift, costly integration	AI-driven closed-loop systems, novel materials, Standardized datasets, clinical trials, cost savings, regulatory compliance
2	S. Wang et al. [26]	GelMA, PLGA, PVP, metals MeHA, PEGDA, PDA, GO, AuNR, BP, MOFs	Experimental (in vivo/in vitro)	Drug delivery, biosensing Wound healing, cancer, diabetes, obesity, alopecia	Insulin, chemotherapeutics, ROS, TNF-α, VEGF, NO	Cancer, Tumor, wounds, diabetes	12-day wound healing, tumor suppression Strong therapeutic outcomes	Preclinical stage; no AI; mechanical design; complex data handling	AI-driven MN chips, clinical trials, Wireless AI control, full wearable systems
3	Merzougui et al. [27]	Polymers, hydrogels, metals	Simulated/empirical (small datasets)	Biosensing (cancer biomarkers)	Tyrosinase, CA15-3, GPC1 exosomes	Breast cancer, melanoma, colorectal cancer	Detection limits: 0.52 μM–1.1 copies/μL	Low sample volume, no clinical AI validation	AI-driven portable readers, multiplex detection
4	Liu et al. [28]	PPy-based HMNs(PEGDA + dopants)	Impedance, temp, clinical score (RA rats)	RA drug delivery + inflammation monitoring	Methotrexate (MTX), Rhodamine B	Rheumatoid Arthritis	Paw swelling ↓40%, cytokines ↓, ML AUC = 0.787	ML accuracy (76.74%), rat-only study	Closed-loop system, multiplexing, human trials
5	Xiao et al. [29]	MOF-HMNs, fluorescent reagents	Smartphone image dataset (fluorescence pH data)	Wound pH monitoring	No drug (self-sterilization + pH monitoring)	Chronic wounds, sepsis prevention	Reliable pH detection via ML, self-sterilization confirmed	No clinical data, device scalability	Clinical trials, biomarker expansion, IoT integration

Table 3. (A) AI/ML Techniques and Smart Integration in Diagnostic and Therapeutic Microneedles. (B) Materials, Datasets, Applications, and Performance Outcomes in Smart Diagnostic Microneedles.

(A) Thematic Group 2: Microneedles for Diagnostics and Therapy
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques &Algorithms	Key AI/ML Role/Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Ashraf et al. [25]	Both	MN	ML, DL, supervised/unsupervised, signal denoising	Noise reduction, biomarker prediction, sensor calibration, signal filtering, diagnostics	Integrated into wearable MNs for real-time diagnostics	Predictive sensing, personalized biosensing, improved accuracy, adaptive calibration	Signal calibration, multi-biomarker detection
2	S. Wang et al. [26]	Both	HMN	AI, ANN, Logic Encoding, IoT	Logic encoding, real-time monitoring, smart tracking, adaptive therapy	Programmable MNs, AI-driven control, therapeutic pathways	Adaptive release, biosensing, multi-therapy MNs, remote control	Multi-disease targeting, drug timing control
3	Merzougui et al. [27]	AI (proposed)	HMN	Not specified (general AI/ML for data analysis)	Enhance diagnostic accuracy, real-time POC analysis	AI for biomarker signal interpretation	Signal filtering, noise reduction, automated quantification	Data processing efficiency
4	Liu et al. [28]	Both	HMN	XGBoost, LSTM	Closed-loop inflammation monitoring	Real-time biosensing with integrated feedback	Theranostic system for RA	Prediction of cytokine levels & treatment response
5	Xiao et al. [29]	Both	HMN	MobileNet, PCA	Smartphone-assisted pH detection	AI-visualization via smartphone fluorescence	Real-time wound pH monitoring	Dynamic healing status identification
(B) Thematic Group 2: Microneedles for Diagnostics and Therapy
	Author (Year)	Materials Used	Data Type/Size	Targeted Application	Biosensing Parameter	Targeted Parameter/ Sickness	Results	Limitations	Future Work
1	Ashraf et al. [25]	Metals, semiconductors, polymers, colorimetric/electrochemical MN	Simulated/empirical (small data), High-dimensional time-series biosensor data	Biosensing, therapy, Non-invasive diagnostics, AI-guided monitoring	Glucose, lactate, cortisol, nucleic acids, pH, electrolytes	Glucose, Diabetes, cancer, inflammation, autoimmune disease	AI-enabled real-time monitoring with high-fidelity, adaptive biosensing	High accuracy, limited scalability; ISF variability, signal drift, costly integration	AI-driven closed-loop systems, novel materials, Standardized datasets, clinical trials, cost savings, regulatory compliance
2	S. Wang et al. [26]	GelMA, PLGA, PVP, metals MeHA, PEGDA, PDA, GO, AuNR, BP, MOFs	Experimental (in vivo/in vitro)	Drug delivery, biosensing Wound healing, cancer, diabetes, obesity, alopecia	Insulin, chemotherapeutics, ROS, TNF-α, VEGF, NO	Cancer, Tumor, wounds, diabetes	12-day wound healing, tumor suppression Strong therapeutic outcomes	Preclinical stage; no AI; mechanical design; complex data handling	AI-driven MN chips, clinical trials, Wireless AI control, full wearable systems
3	Merzougui et al. [27]	Polymers, hydrogels, metals	Simulated/empirical (small datasets)	Biosensing (cancer biomarkers)	Tyrosinase, CA15-3, GPC1 exosomes	Breast cancer, melanoma, colorectal cancer	Detection limits: 0.52 μM–1.1 copies/μL	Low sample volume, no clinical AI validation	AI-driven portable readers, multiplex detection
4	Liu et al. [28]	PPy-based HMNs(PEGDA + dopants)	Impedance, temp, clinical score (RA rats)	RA drug delivery + inflammation monitoring	Methotrexate (MTX), Rhodamine B	Rheumatoid Arthritis	Paw swelling ↓40%, cytokines ↓, ML AUC = 0.787	ML accuracy (76.74%), rat-only study	Closed-loop system, multiplexing, human trials
5	Xiao et al. [29]	MOF-HMNs, fluorescent reagents	Smartphone image dataset (fluorescence pH data)	Wound pH monitoring	No drug (self-sterilization + pH monitoring)	Chronic wounds, sepsis prevention	Reliable pH detection via ML, self-sterilization confirmed	No clinical data, device scalability	Clinical trials, biomarker expansion, IoT integration

Ashraf et al. [25] applied deep learning (CNNs, RNNs) to optical and electrochemical MN sensor data, reducing noise, calibrating outputs, and detecting biomarker patterns. Clustering aided disease subtyping, while RL achieved 98.4% glucose monitoring accuracy in closed-loop diabetes care. Applications included stress and cortisol monitoring, though challenges remain with ISF variability and the scarcity of large, diverse datasets for robust model generalization.

Expanding on the concept of adaptive systems, S. Wang et al. [26] conceptually examined the integration of AI/ML with MNs, imagining programmable, smart MNs for immediate monitoring and drug release responsive to biomarkers. Suggested designs featured neural network–driven dosing and adaptive photodynamic therapy, focusing on improving treatment accuracy. Despite the absence of experimental AI execution, the review highlighted the possibilities for data-driven, autonomous MN platforms.

Merzougui et al. [27] addressed MN array–based dermal ISF biopsy for cancer diagnostics, focusing on material and sensor innovations. While AI/ML was not implemented, the authors noted its potential for noise filtering, real-time analysis, and automated biomarker quantification. Electrochemical and fluorescent MN sensors detected tyrosinase at 0.52 μM for melanoma screening, with AI expected to improve multiplex detection accuracy.

Liu et al. [28] developed a wearable hydrogel MN device for rheumatoid arthritis, integrating impedance sensing, methotrexate delivery, and electrical stimulation. Logistic regression predicted inflammation with 76.74% accuracy in rat models, enabling therapy adjustments. Treatment reduced paw thickness by 40% and preserved bone density. Biodegradable within four weeks, the system shows promise for closed-loop chronic disease management despite limited predictive accuracy and lack of human validation.

Xiao et al. [29] introduced a MOF–hydrogel MN patch for wound pH monitoring and localized antimicrobial action, using smartphone fluorescence imaging. ML classified pH levels with >90% accuracy. In vitro and ex vivo tests confirmed stability, bacterial inhibition, and clinically relevant sensitivity. Limitations include no clinical validation and limited dataset diversity, with future work targeting broader biomarker detection and improved scalability.

Together, this thematic group foresees future MN systems effortlessly combined with wearable devices and cloud-driven data platforms. The research highlights addressing mechanical and regulatory obstacles via AI-driven, iterative design enhancements. Collectively, they delineate a strategy for smart MN platforms that can provide diagnostics and therapeutics instantly. These adaptive, clinically applicable, and predictive systems connect the divide between laboratory prototypes and commercial healthcare solutions, promoting personalized, minimally invasive digital health.

3.3. Thematic Group 3: Microneedles for Drug Delivery

This thematic group covers hydrogel and polymer-based microneedles for sustained drug delivery, emphasizing biocompatibility, dissolution, and formulation control. Reviewed studies apply AI/ML methods like DL and FEM to optimize geometry, kinetics, and dosing. Table 4A,B summarize AI strategies, materials, drug targets, diseases treated, outcomes, limitations, and proposed improvements for future delivery systems.

Table 4. (A) AI/ML Techniques and Design Optimization in Microneedle-Based Drug Delivery. (B) Materials, Drug Targets, Outcomes, and Future Directions in Microneedle-Based Drug Delivery.

(A) Thematic Group 3: Microneedles for Drug Delivery
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques & Algorithms	Key AI/ML Role/Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Albayati et al. [30]	Both	HMN	DNN, ANN, BioSIM, COMSOL, KNN, SVM	Personalized delivery, predictive skin permeability, diffusion & release modeling	Optimized MN geometry, ML-integrated skin data, AI-driven real-time delivery	Permeability prediction, real-time dosing, personalized delivery, spatial MN design, kinetic control	Drug release precision, Microneedle geometry, diffusion control
2	Aghajani et al. [31]	Proposed	HMN	No specific technique applied	Future optimization of delivery systems	Briefly mentioned (AI for future optimization)	None (theoretical potential)	Sustained release, drug penetration, Riluzole CNS targeting
3	Suriyaamporn et al. [32]	Both	HMN	RF, PCA, SVM, ANN	Optimization of formulation & release prediction	AI-guided optimization of PEGylated formulations	Improved drug permeation and cancer cell apoptosis	Feature extraction for enhanced therapeutic targeting
4	Bagde et al. [33]	Both	Hydrogel	CNN	Quality control and drug diffusion modeling	CNN-based image inspection of MN prints	Controlled ibuprofen release	Biphasic delivery modeling with CNN
5	Yuan et al. [34]	Both	HMN	XGBoost, Ridge, Lasso, SVR	Drug flux prediction	XGBoost outperforming traditional Fick models	Data-driven simulation of MN performance	Multi-drug transport modeling
6	Zong et al. [35]	Both	HMN	Real-time sensor monitoring, adaptive ML algorithms	Optimize gas production, drug penetration, safety control	Dynamic feedback for microbial metabolism and gas/drug control	Real-time monitoring, 94% dermal retention, controlled gas propulsion	Spatiotemporal drug release, minimized systemic exposure
(B) Thematic Group 3: Microneedles for Drug Delivery
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/Biosensing Parameter	Targeted Parameter/Sickness	Results	Limitations	Future Work
1	Albayati et al. [30]	Chitosan, ethosomes, liposomes, PLA, hydrogels, metallic/biodegradable MNs (referenced)	Clinical trials, simulated & empirical wearable biosensor data	Personalized transdermal drug delivery and biosensing	Acyclovir, SARS-CoV-2 vaccine, insulin, lidocaine, fentanyl, biologics; pH, temp, hydration	HSV, COVID-19, Alzheimer’s, Diabetes, pain, hormone therapy, neurological diseases	Fewer side effects; improved dose prediction, adaptive delivery, enhanced bioavailability	Data bias, dataset standardization, skin variability, wearable integration complexity	AI-nano integration, real-time AI patches, IoT feedback, smart skin models, AI-guided fabrication
2	Aghajani et al. [31]	Polymers, lipids (nano-particles); phospho-lipids (liposomes), micelles	Preclinical (rodents), small trials, review-based analysis	Drug delivery	Riluzole	ALS (Amyotrophic Lateral Sclerosis)	Improved CNS riluzole delivery in animal models, better compliance	Regulatory barriers, scalability, cost, BBB penetration, patient variability, no AI	AI-driven optimization, MN refinement, AI/ML for personalization & scale-up
3	Suriyaamporn et al. [32]	Crosslinked hydrogel MNs (PEGylated liposomes, HPβCD, limonene)	75 formulations (experimental dataset)	Transdermal chemotherapy	5-FU (fluorouracil)	Non-melanoma skin cancer (BCC, SCC)	91.5% apoptosis, 41.78% permeation, >90% recovery	Limited to in vitro/ex vivo, scalability issues	Clinical trials, other drugs, real-time tracking, long-term stability
4	Bagde et al. [33]	PEGDAMA-based resin MNs	CNN-analyzed print images + in vivo drug data	Transdermal drug delivery	Ibuprofen (IBU)	Pain/inflammation	Sustained release (AUC = 62,812 ng/mL*h), biphasic delivery	Single drug, small-scale testing	Biologic delivery, clinical trials, IoT integration
5	Yuan et al. [34]	Hydrogel & plastic MNs	191 points (6 drugs) for ML prediction	Transdermal drug delivery simulation	Caffeine, Lidocaine, Rhodamine B, GHK, BSA, Copper	Generic delivery modeling	XGBoost: R2 = 0.98; better than Fick’s law	Small, unbalanced dataset	Expand data, model validation, clinical integration
6	Zong et al. [35]	Gas-permeable hydrogels, thermosensitive polymers	Experimental, porcine & murine models	Transdermal drug delivery	Calcipotriol, H2, NO, H2S	Psoriasis, tumors, chronic wounds	94% dermal drug retention, 1000 μm depth, effective symptom reduction	Microbial safety, gas control precision, clinical scalability	Use safer microbes, smart hydrogels, regulatory advancement

Table 4. (A) AI/ML Techniques and Design Optimization in Microneedle-Based Drug Delivery. (B) Materials, Drug Targets, Outcomes, and Future Directions in Microneedle-Based Drug Delivery.

(A) Thematic Group 3: Microneedles for Drug Delivery
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques & Algorithms	Key AI/ML Role/Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Albayati et al. [30]	Both	HMN	DNN, ANN, BioSIM, COMSOL, KNN, SVM	Personalized delivery, predictive skin permeability, diffusion & release modeling	Optimized MN geometry, ML-integrated skin data, AI-driven real-time delivery	Permeability prediction, real-time dosing, personalized delivery, spatial MN design, kinetic control	Drug release precision, Microneedle geometry, diffusion control
2	Aghajani et al. [31]	Proposed	HMN	No specific technique applied	Future optimization of delivery systems	Briefly mentioned (AI for future optimization)	None (theoretical potential)	Sustained release, drug penetration, Riluzole CNS targeting
3	Suriyaamporn et al. [32]	Both	HMN	RF, PCA, SVM, ANN	Optimization of formulation & release prediction	AI-guided optimization of PEGylated formulations	Improved drug permeation and cancer cell apoptosis	Feature extraction for enhanced therapeutic targeting
4	Bagde et al. [33]	Both	Hydrogel	CNN	Quality control and drug diffusion modeling	CNN-based image inspection of MN prints	Controlled ibuprofen release	Biphasic delivery modeling with CNN
5	Yuan et al. [34]	Both	HMN	XGBoost, Ridge, Lasso, SVR	Drug flux prediction	XGBoost outperforming traditional Fick models	Data-driven simulation of MN performance	Multi-drug transport modeling
6	Zong et al. [35]	Both	HMN	Real-time sensor monitoring, adaptive ML algorithms	Optimize gas production, drug penetration, safety control	Dynamic feedback for microbial metabolism and gas/drug control	Real-time monitoring, 94% dermal retention, controlled gas propulsion	Spatiotemporal drug release, minimized systemic exposure
(B) Thematic Group 3: Microneedles for Drug Delivery
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/Biosensing Parameter	Targeted Parameter/Sickness	Results	Limitations	Future Work
1	Albayati et al. [30]	Chitosan, ethosomes, liposomes, PLA, hydrogels, metallic/biodegradable MNs (referenced)	Clinical trials, simulated & empirical wearable biosensor data	Personalized transdermal drug delivery and biosensing	Acyclovir, SARS-CoV-2 vaccine, insulin, lidocaine, fentanyl, biologics; pH, temp, hydration	HSV, COVID-19, Alzheimer’s, Diabetes, pain, hormone therapy, neurological diseases	Fewer side effects; improved dose prediction, adaptive delivery, enhanced bioavailability	Data bias, dataset standardization, skin variability, wearable integration complexity	AI-nano integration, real-time AI patches, IoT feedback, smart skin models, AI-guided fabrication
2	Aghajani et al. [31]	Polymers, lipids (nano-particles); phospho-lipids (liposomes), micelles	Preclinical (rodents), small trials, review-based analysis	Drug delivery	Riluzole	ALS (Amyotrophic Lateral Sclerosis)	Improved CNS riluzole delivery in animal models, better compliance	Regulatory barriers, scalability, cost, BBB penetration, patient variability, no AI	AI-driven optimization, MN refinement, AI/ML for personalization & scale-up
3	Suriyaamporn et al. [32]	Crosslinked hydrogel MNs (PEGylated liposomes, HPβCD, limonene)	75 formulations (experimental dataset)	Transdermal chemotherapy	5-FU (fluorouracil)	Non-melanoma skin cancer (BCC, SCC)	91.5% apoptosis, 41.78% permeation, >90% recovery	Limited to in vitro/ex vivo, scalability issues	Clinical trials, other drugs, real-time tracking, long-term stability
4	Bagde et al. [33]	PEGDAMA-based resin MNs	CNN-analyzed print images + in vivo drug data	Transdermal drug delivery	Ibuprofen (IBU)	Pain/inflammation	Sustained release (AUC = 62,812 ng/mL*h), biphasic delivery	Single drug, small-scale testing	Biologic delivery, clinical trials, IoT integration
5	Yuan et al. [34]	Hydrogel & plastic MNs	191 points (6 drugs) for ML prediction	Transdermal drug delivery simulation	Caffeine, Lidocaine, Rhodamine B, GHK, BSA, Copper	Generic delivery modeling	XGBoost: R2 = 0.98; better than Fick’s law	Small, unbalanced dataset	Expand data, model validation, clinical integration
6	Zong et al. [35]	Gas-permeable hydrogels, thermosensitive polymers	Experimental, porcine & murine models	Transdermal drug delivery	Calcipotriol, H2, NO, H2S	Psoriasis, tumors, chronic wounds	94% dermal drug retention, 1000 μm depth, effective symptom reduction	Microbial safety, gas control precision, clinical scalability	Use safer microbes, smart hydrogels, regulatory advancement

Albayati et al. [30] utilized DNN and SVM to forecast drug skin permeability and enhance TDDS parameters, increasing flux by 50% compared to conventional methods. COMSOL/BioSIM (version 5.3) simulations demonstrated temperature-dependent absorption, whereas ANN-optimized hydrogels facilitated prolonged Alzheimer’s medication release. Patches with integrated sensors modified dosing based on physiological feedback, shortening rivastigmine trial time by 30%. Future efforts aim at merging AI with nanotechnology for targeted oncology.

Expanding to neurodegenerative therapies, Aghajani et al. [31] reviewed HMN delivery of riluzole for ALS, tackling low bioavailability and limited blood–brain barrier (BBB) penetration. Although AI/ML was not applied, the authors proposed its use for optimizing formulations, predicting release kinetics, and tailoring MN geometry to drug properties—enabling scalable, reproducible, patient-specific systems. A related example of AI integration is shown in [32] (Graphical abstract), where AI modeling guided the optimization of crosslinked hydrogel microneedles for localized 5-FU delivery.

Predictive modeling using a Gaussian Process Regressor optimized crosslinking parameters, enabling incorporation of 5-FU–HPβCD liposomes to improve permeation, stability, and localized delivery. In vitro/ex vivo studies demonstrated significant cancer cell apoptosis, underscoring the translational potential of AI-driven formulation strategies in microneedle-mediated drug delivery. Adapted from Suriyaamporn et al. [32].

Bagde et al. [33] developed an AI-optimized 3D-printed PEGDAMA MN patch for biphasic ibuprofen delivery. CNNs refined MN geometry for fidelity and reproducibility via DLP printing. The patch achieved burst and sustained release over 72 h, with strong IVIVR and first-then zero-order kinetics. Validation in vitro, ex vivo, and in vivo supports broader potential, though scalability, clinical testing, and expanded APIs remain future needs.

Yuan et al. [34] used ML to predict drug permeation through MN-treated skin, analyzing 191 experiments with six drugs. Among Fick’s law, MLR, RF, and XGBoost, the latter performed best (R² = 0.98). Permeation time and drug loading emerged as key predictors. While highlighting ML’s promise for MN design optimization, dataset homogeneity limits clinical and industrial generalizability.

Zong et al. [35] presented a HMN platform powered by microbial metabolism for self-sufficient drug delivery. Encapsulated microbes produced gas to drive calcipotriol into the dermis, achieving 94% retention. A thermoresponsive barrier halted leakage, allowing for regulated release without outside energy. Although AI/ML was not utilized, the modular biohybrid method demonstrates potential for personalized medicine, cancer treatment, and wound healing, yet microbial safety and gas regulation continue to be significant challenges.

Together, this collection of research shows how microneedles made from hydrogels—when paired with smart design concepts—can produce systems that are both efficient and minimally invasive. Numerous writers anticipate developments arising from AI-based biosensors for immediate therapeutic observation. Collectively, these investigations enhance transdermal drug delivery through the incorporation of intelligent polymers with sophisticated manufacturing, providing dosage-efficient, personalized, and clinically pertinent solutions while fostering sustainable, budget-friendly treatment approaches.

3.4. Thematic Group 4: Drug Development for Microneedles

This thematic group examines AI applications in drug development, including screening, formulation, toxicity prediction, and bioavailability. Studies highlight tools like deep learning, neural networks, and systems modeling. While not MN-specific, they inform MN-compatible delivery strategies. Table 5A,B summarize AI techniques, design roles, materials, datasets, drug targets, medical conditions, outcomes, limitations, and future directions.

Table 5. (A) AI/ML Techniques and Applications in Drug Development for Microneedle-Enabled Therapeutics. (B) Materials, Datasets, Therapeutic Targets, and Outcomes in AI-Supported Drug Development.

(A) Thematic Group 4: Drug Development for Microneedles
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques & Algorithms	Key AI/ML Role/ Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Biswas et al. [15]	Both	HMN	Voting Regressor, Ensemble ML	Prediction of transdermal drug delivery	Model ensemble for flux prediction	Improved RMSE in simulated release	Framework validation across various drugs
2	Reddy et al. [36]	Both (conceptual)	HMN	Predictive modeling, deep learning (conceptual)	Design optimization, personalized dosing, feedback systems	Conceptual integration with biosensors and smart release	Proposed: Real-time feedback, adaptive dosing	Responsive release, closed-loop regulation (future)
(B) Thematic Group 4: Drug Development for Microneedles
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/ Biosensing Parameter	Targeted Parameter/Sickness	Results	Limitations	Future Work
1	Biswas et al. [15]	Hydrogel MNs and solid MNs	Dataset from Yuan et al., ML predictions	Drug permeation prediction	Unspecified	General drug delivery	Voting regressor RMSE: 3.24 (percentage), 654.94 (amount)	No experimental validation	Advanced models, expanded datasets, clinical validation
2	Reddy et al. [36]	Chitosan, HA, gelatin, PEG, PVA, smart hydrogels	Review, multiple cited works	Smart drug delivery	Insulin, doxorubicin, 5-FU, curcumin, antibiotics	Diabetes, cancer, infection, pain, inflammation	Improved compliance, targeted release, reduced toxicity	Toxicity, clinical translation, regulation, cost	AI integration, feedback DDS, better targeting and biocompatibility

Table 5. (A) AI/ML Techniques and Applications in Drug Development for Microneedle-Enabled Therapeutics. (B) Materials, Datasets, Therapeutic Targets, and Outcomes in AI-Supported Drug Development.

(A) Thematic Group 4: Drug Development for Microneedles
	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques & Algorithms	Key AI/ML Role/ Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Biswas et al. [15]	Both	HMN	Voting Regressor, Ensemble ML	Prediction of transdermal drug delivery	Model ensemble for flux prediction	Improved RMSE in simulated release	Framework validation across various drugs
2	Reddy et al. [36]	Both (conceptual)	HMN	Predictive modeling, deep learning (conceptual)	Design optimization, personalized dosing, feedback systems	Conceptual integration with biosensors and smart release	Proposed: Real-time feedback, adaptive dosing	Responsive release, closed-loop regulation (future)
(B) Thematic Group 4: Drug Development for Microneedles
	Author (Year)	Materials Used	Dataset/Data Type/Size	Targeted Application	Target Drug/ Biosensing Parameter	Targeted Parameter/Sickness	Results	Limitations	Future Work
1	Biswas et al. [15]	Hydrogel MNs and solid MNs	Dataset from Yuan et al., ML predictions	Drug permeation prediction	Unspecified	General drug delivery	Voting regressor RMSE: 3.24 (percentage), 654.94 (amount)	No experimental validation	Advanced models, expanded datasets, clinical validation
2	Reddy et al. [36]	Chitosan, HA, gelatin, PEG, PVA, smart hydrogels	Review, multiple cited works	Smart drug delivery	Insulin, doxorubicin, 5-FU, curcumin, antibiotics	Diabetes, cancer, infection, pain, inflammation	Improved compliance, targeted release, reduced toxicity	Toxicity, clinical translation, regulation, cost	AI integration, feedback DDS, better targeting and biocompatibility

Biswas et al. [15] evaluated ML models—including stacking regressor, voting regressor, and ANN—for forecasting drug permeation from hydrogel and solid MNs based on Yuan et al. [34] dataset. The voting regressor reached the highest accuracy (RMSE: 3.24% and 654.94 µg), exceeding previous models. A web application based on Flask was developed for practical purposes. Constraints involve the diversity of the dataset and the absence of experimental validation, with future efforts aimed at deep learning and real-time quality assurance.

Complementing this modeling focus, Reddy et al. [36] reviewed smart drug delivery systems using hydrogels, microneedles, and other stimuli-responsive carriers for site-specific release triggered by pH, temperature, or enzymes. While lacking experimental data, they proposed integrating AI/ML for predictive modeling, biosensor feedback, and closed-loop delivery. Applications span diabetes, cancer, and infection, but challenges remain in toxicity, cost, and regulatory approval.

Collectively, this thematic group highlights the significance of cooperation between AI experts and pharmaceutical researchers in developing tailored, adaptive medication treatments. Their research demonstrates AI’s capability to revolutionize all phases of pharmaceutical development—from molecular design to clinical implementation—while providing a developmental pathway for quicker, more dependable formulation. These investigations aid in establishing new benchmarks for efficiency, accuracy, and adherence to regulations, indicating a shift toward smart, data-informed drug discovery methods that are both creative and clinically significant.

3.5. Thematic Group 5: Microneedles Sensors for Health and Agriculture

This thematic group explores AI-integrated microneedle sensors for real-time monitoring in healthcare and agriculture. Studies highlight applications in disease management, crop stress detection, and VOC tracking. AI enhances signal fidelity, feedback control, and scalability.

Table 6. (A) AI/ML Techniques and Smart Microneedle Sensor Integration for Health and Agriculture. (B) Materials, Biomarker Targets, Applications, and Outcomes in AI-Integrated Microneedle Sensors.

(A) Thematic Group 5: Microneedle Sensors for Health and Agriculture
#	Author (Year)	AI/ML Both	Hydrogel Microneedle HMN	AI/ML Techniques & Algorithms	Key AI/ML Role/Purpose Application	AI/ML Integration Innovation in HMN	AI-Enhanced HMN Features	AI-Enhanced Targeted Design Features
1	Kharb et al. [37]	Both	MN	RNN, ML algorithms DL, CNN, edge AI, RL, clustering	Predictive analytics, noise reduction, personalized health insights, Bio-signal processing,	AI enhances data analysis, predictive modeling, and real-time feedback for wearables	Improved disease prediction, real-time monitoring, user compliance, dynamic calibration	Sensor calibration, multimodal data fusion, adaptive feedback
2	Ermis et al. [38]	Both	HMN	ML (predictive modeling, calibration algorithms) SVM, NN, DL	Real-time biomarker analysis, personalized insights, biosignal feedback	ML optimizes microneedle sampling efficiency, Real-time biosignal analysis	Accuracy, real-time response	Dynamic calibration for skin variability, Signal robustness, hydration mapping
3	A. Sen et al. [39]	Both	HMN	Predictive analytics, trend detection, ML alerts, (Lacks in-depth AI/ML details)	Glucose prediction, treatment personalization	Indirect (via wearable analytics)	Remote tracking, early alerts	Glucose sensing, insulin release
4	Kachouei et al. [40]	Both	MN	AI, ML, DL, Pattern Recognition Predictive analytics, CRISPR-Cas12a	Pathogen detection, crop optimization Decision support, real-time alerts, prediction	IoT-AI fusion, smartphone-based real-time sensing, wireless data transmission	Real-time decisions & detection, predictive alerts, smart integration, scalable systems	Customizable sensing, VOC detection, hormone fluctuation, nitrate detection,
5	Fucheng Zhang et al. [41]	Both	MN	ANN, deep learning, edge computing Data fusion algorithms, IoT, ML-enabled optimization	Predictive analytics, data fusion, stability compensation, precision agriculture, real-time plant monitoring	No AI focus; ML integrated into multimodal, wearable, implantable plant sensors	Stability, multimodal data fusion; microclimate sensing, disease/stress prediction	Environmental adaptability, accuracy Plant hormone detection, stress prediction
6	Ausri et al. [42]	Both	HMN	SVM, Random Forest	Ketone detection modeling	AI-enhanced biosensor for metabolic monitoring	AI-tuned sensor thresholds for diabetic patients	Ketone monitoring & early DKA prediction
(B) Thematic Group 5: Microneedle Sensors for Health and Agriculture
	Author (Year)	Materials Used	Dataset/ Data Type/ Size	Targeted Application	Target Drug/Biosensing Parameter	Targeted Parameter/Sickness	Results	Limitations	Future Work
1	Kharb et al. [37]	Graphene, biodegradable polymers, flexible bioelectronics, Flexible sensors, nanocomposites	Empirical (patient datasets), Large-scale IoT	Biosensing, drug delivery, Biosensing + diagnostics	Lactate, VOCs Glucose, cortisol, sweat ions	Metabolic disorders, chronic diseases, Diabetes, depression, neurostress	High early detection accuracy, less invasive; personalized, real-time predictive diagnostics	data privacy, environmental variability, power, clinical reliability	AI-IoT closed-loop systems, AI twins, federated learning, edge-powered wearables
2	Ermis et al. [38]	PDMS, graphene, PVA, chitosan, silk fibroin	Empirical (sweat, blood, tears)	Biosensing (glucose, cortisol, electrolytes)	Glucose, cortisol, Na+, K+ Electrolytes,	Diabetes, stress, dehydration	95% correlation with blood tests, High accuracy, low latency	Power constraints, data security, Motion artifacts, privacy	Biodegradable materials, AI-driven closed-loop systems, blockchain, personalization, hormone tracking
3	A. Sen et al. [39]	Chitosan, PLGA, graphene, PANI, CNT, ZnO, metal oxides	In vitro/in vivo studies Secondary data discussed	Drug delivery & biosensing	Glucose, Insulin, metformin, GLP-1, sulfonylureas	Diabetes mellitus	Reduced injections, Better glycemic control, wearable convenience	Scalability, biocompatibility, Scale-up issues, toxicity concerns, regulatory compliance	Smart polymers, nanocomposites, digital health, AI-health integration, digital medicine
4	Kachouei et al. [40]	Graphene, CNTs, polymers, Au NPs PDA, MN, nanofibers	Empirical (field data), streaming data, low-volume	Food safety, plant health, disease detection, sustainability, stress monitoring	Toxins, Pathogens, VOCs, Hormones, Nitrate	Heavy metals, pathogens, VOCs, Foodborne illness, plant stress	Early disease detection, reduced pesticide use, Highly accurate, responsive	High cost, energy inefficiency Scalability, cost, infra-structure	Solar sensors, blockchain, 5G, green tech, AI-automated models
5	Fucheng Zhang et al. [41]	Graphene, MXene, LIG, PI, AgNO3 PVA, conductive hydrogel, CNTs, sensors	Empirical/field data (large-scale) Empirical + Real-field data	Crop monitoring, soil analysis Crop growth, drought, stress sensing	NH4+, H2O2, VOCs, ethylene Salicylic acid, ethylene, nitrate, VOCs	Crop diseases, nutrient deficiency, drought, plant stress	Ethylene LOD: 0.084 ppm; VOC accuracy: 97% High stretchability, energy harvesting, accuracy	Cost, durability, environmental impact Sensor longevity, calibration stability	Self-powered sensors, biodegradable materials Integration of AI-powered feedback loops, data mining, closed-loop optimization
6	Ausri et al. [42]	DA-HA hydrogel MNs, PEDOT:PSS, NAD+, HBD	Porcine & rat ketone levels + ML model	Continuous ketone sensing	No drug (biosensor only)	Type 1 Diabetes (DKA)	MAD: 0.184 mM, R2 = 0.95, MARD: 7.68%	2 hr tests only, animal models	Longer trials, CGM integration, multiplex sensing

A,B summarize AI techniques, integration methods, materials, targets, conditions monitored, outcomes, limitations, and future directions across wearable and implantable sensor platforms.

Kharb et al. [37] investigated the integration of AI/ML with non-invasive MN patches, smart tattoos, and implantable biosensors for the real-time analysis of biomarkers in ISF, sweat, and breath. CNNs and RNNs eliminated noise and identified substances such as glucose, lactate, and cortisol. Uses encompassed detecting COVID-19 from data collected by wearable devices. Obstacles encompass calibration, variety in datasets, privacy concerns, and ethical issues, with advances dependent on strong models and interdisciplinary teamwork.

Ermis et al. [38] reviewed wearable materials, including hydrogels and MNs, for health and sports, noting AI/ML’s potential for personalized hydration and stress management. ML-calibrated MN glucose sensors using light polarization achieved up to 95% correlation with invasive tests. While AI/ML was not the main focus, the study highlights predictive analytics as an emerging tool for athletic performance optimization.

Similarly, A. Sen et al. [39] synthesized advancements in polymer nanocomposite (PNC) devices for enhancing glucose biosensors and enabling controlled insulin release. While AI/ML was not explored in depth, the review positioned these technologies as emerging enablers of real-time glucose monitoring and predictive analytics in diabetes care.

Kachouei et al. [40] analyzed IoT-integrated sensors utilizing AI/ML for sustainable farming and food security. The applications featured CRISPR-Cas12a combined with AI-enhanced colorimetry for detecting pathogens (LoD: 3.4 × 10² CFU/mL) and machine learning-enabled drone imaging for evaluating crop health. Systems also analyzed plant VOCs for early detection of stress. Challenges encompass energy requirements of IoT, absence of data standards, and calibration in varying environments.

Fucheng Zhang et al. [41] utilized AI/ML on agricultural sensor data for forecasting yields and predicting diseases. ANNs measured IAA (LoD: 10.8 pg/mL) and forecasted crop stress based on ethylene concentrations (0.5–20 ppm). Hyperspectral imaging utilizing deep learning identified disease VOCs with over 97% accuracy, whereas ML-driven 3D-printed strain sensors monitored plant growth. Ongoing challenges involve field signal drift and sensor expenses, with upcoming efforts focusing on self-calibration and environmentally friendly designs.

Ausri et al. [42] developed a dopamine–hyaluronic acid HMN patch for continuous ketone monitoring in DKA management. Using β-hydroxybutyrate dehydrogenase and NAD⁺ for selective β-HB detection, a GBDT model corrected ISF–blood delays (R² = 0.95; MARD = 7.68%). The device showed high biocompatibility and stability in animal models. Limitations include short testing duration and no human trials, with future work targeting multiplex sensing and CGM integration.

Together, this thematic group demonstrates how AI/ML-driven, non-invasive wearable sensors are enhancing healthcare and agriculture with personalized, dependable monitoring. By tackling issues like sensor precision, data unification, and adaptability to environments, these research efforts establish the groundwork for future smart wearables. Their advancements illustrate the collaboration between cutting-edge biosensing materials and smart data analytics, facilitating ongoing health and agricultural monitoring with prompt actions. In addition to enhancing personal health, farming, and sports performance, these systems offer promise for proactive public health, creating a strong foundation for AI-enhanced wearable technologies in both clinical and agricultural sectors.

4. Development Pathways for AI/ML Applications in Hydrogel Microneedles

HMNs demonstrate potential for transdermal drug administration because of their biocompatibility, adaptability, and low invasiveness. Achieving their complete clinical potential necessitates enhancement in various areas—material design, structural engineering, production, and biomedical functionality.

Recent developments demonstrate this capability: Xu et al. [43] used Bayesian optimization to enhance hydrogel compositions; Seifermann et al. [44] trained machine learning models on extensive photodegradation datasets for automated choices; Lin et al. [45] and Vora et al. [46] emphasized neural network models that advanced drug release forecasts and nanoparticle toxicity evaluations. These instances illustrate how AI can aid in both material development and treatment enhancement.

To guide this effort, we introduce AIM-DO (AI-driven Microneedle Design Optimization), a conceptual model outlining five interconnected pillars for AI/ML integration in HMN development. These include: (1) Material Discovery and Optimization, (2) Structural and Mechanical Design, (3) Manufacturing Process Optimization, (4) Biomedical Applications and Performance Enhancement, and (5) Advanced AI Integration.

Each pillar addresses significant obstacles in microneedle research—spanning from predictive material modeling to adaptive smart delivery systems—utilizing AI to enhance discovery, fabrication, and clinical translation. The AIM-DO concept is initially depicted in Figure 1 and subsequently represented on translational timelines in Figure 2, with a summary of detailed applications provided in

Table 7. AIM-DO Development for Hydrogel Microneedles.

Development Stage	AI/ML Application	Description	Benefits
1. Material Discovery & Optimization	Data-Driven Material Selection	AI models analyze datasets of hydrogel compositions to predict optimal formulations.	Faster material discovery, improved biocompatibility.
	Predictive Modeling	ML predicts hydrogel behavior under various conditions (hydration, pH, temperature).	Reduces trial-and-error experimentation.
	High-Throughput Screening	Deep learning and reinforcement learning identify promising hydrogel candidates.	Accelerates discovery of novel materials.
	Multi-Objective Optimization	Genetic algorithms balance mechanical strength, biodegradability, and drug release.	Tailored microneedles for specific applications.
2. Structural & Mechanical Design	Generative Design	AI-driven topology optimization for novel microneedle geometries.	Enhances penetration efficiency and drug delivery.
	FEA	ML predicts microneedle stress distribution and mechanical failure points.	Improves durability and reliability.
	Bio-Inspired Design	AI mimics biological structures for better adhesion and penetration.	Increased effectiveness and patient comfort.
	Swelling Dynamics Prediction	AI models hydrogel swelling behavior for optimal drug diffusion.	Ensures controlled fluid absorption and drug release.
3. Manufacturing Process Optimization	Process Control & Automation	AI optimizes 3D printing, micro-molding, and photopolymerization.	Reduces defects, improves fabrication efficiency.
	Defect Detection	ML-based image recognition detects microscopic defects in microneedles.	Enhances quality control, reduces waste.
	3D Printing Optimization	AI refines printing techniques and layer-by-layer deposition.	Ensures structural integrity and printability.
	Self-Correcting Fabrication Systems	Reinforcement learning enables real-time process adjustments.	Reduces variability, increases yield.
4. Biomedical Applications & Performance Enhancement	Drug Delivery Optimization	AI predicts drug release kinetics based on hydrogel properties.	Enables sustained and controlled drug release.
	Personalized Medicine	ML analyzes patient data to tailor microneedle formulations.	Enhances treatment efficacy, reduces side effects.
	Biosensing Integration	AI optimizes sensor placement and biomarker detection algorithms.	Improves real-time health monitoring accuracy.
	AI-Assisted Pharmacokinetics Modeling	Deep learning models simulate drug absorption and metabolism.	Speeds up drug development and regulatory approval.
5. Advanced AI Integration	AI-Powered Smart Microneedles	Embedded AI sensors adjust drug delivery in real-time.	Enables adaptive and responsive therapy.
	Blockchain for Data Security	Secure storage and sharing of microneedle treatment data.	Ensures regulatory compliance and patient privacy.
	Hybrid AI-Physics Models	AI-driven simulations combined with physics-based modeling.	Enhances predictive accuracy for microneedle performance.

. Combined, these components offer a guide for future AI-augmented HMN systems.

AIM-DO spans material optimization, mechanical design, manufacturing, biomedical applications, and advanced AI integration—each targeting specific innovation domains to accelerate microneedle advancement.

This table outlines AI/ML applications, functional descriptions, and anticipated benefits across five development pillars: material optimization, structural design, manufacturing, biomedical applications, and intelligent system integration.

4.1. Pillar 1: Material Discovery & Optimization

This pillar utilizes ML to forecast hydrogel swelling, degradation, and drug-loading by analyzing polymer characteristics, crosslinking, and biocompatibility profiles—reducing trial-and-error. Robotic screening with high throughput confirms predictions, and multi-objective optimization harmonizes elements like strength, biodegradability, and release rates, allowing for designs customized for objectives like swift burst release or extended delivery.

4.2. Pillar 2: Structural & Mechanical Design

Generative design algorithms develop microneedle structures for maximum penetration and minimal discomfort. FEA guarantees mechanical stability, while designs inspired by biology enhance efficiency and biocompatibility. Models of swelling forecast hydrogel enlargement to regulate drug release. These instruments allow for adjustable shapes—conical or pyramidal ends—and pore designs tailored to the characteristics of drugs and the requirements of patients.

4.3. Pillar 3: Manufacturing Process Optimization

AI-based controls modify temperature and humidity levels during hydrogel synthesis to minimize variations. Computer vision identifies imperfections such as microcracks instantaneously. In the creation of microneedles, reinforcement learning and sensor feedback optimize 3D printing settings—nozzle velocity, layer thickness, UV curing—for accurate self-correction. These advancements improve output, reliability, and scalability, connecting laboratory prototypes to mass production.

4.4. Pillar 4: Biomedical Applications & Performance Enhancement

AI-based models associate microneedle characteristics—swelling ratio, pore dimensions—with pharmacokinetics to enhance bioavailability. Personalized medicine models utilize patient information (skin thickness, metabolism) and digital twins to tailor designs. Reasoning integration combines symbolic AI with neural networks for flexible dosing that responds to biomarkers such as glucose. Pharmacokinetic simulations speed up preclinical evaluations, decreasing the need for animals. Case studies on insulin delivery demonstrate improved accuracy and reduced off-target effects.

4.5. Pillar 5: Advanced AI Integration

This pillar proposes HMN enhanced with IoT for immediate monitoring and incorporation with healthcare systems. Embedded microsensors monitor drug release and send data to cloud platforms for clinician evaluation. Blockchain safeguards the creation and treatment documentation. Hybrid AI-physics developmental pathway improve forecasting in intricate biological systems, whereas federated learning facilitates distributed training among organizations, preserving privacy and adhering to regulations.

4.6. Synergistic Impact and Future Directions

The AIM-DO presents a cohesive AI/ML-based developmental pathway to enhance HMN development in material selection, structural design, manufacturing, and biomedical applications. Through the incorporation of predictive modeling, automation, and immediate feedback, it greatly reduces development timelines—turning iterative processes that previously took years into expedited workflows lasting months. This cross-disciplinary approach connects materials science, mechanical engineering, and biomedical research, improving both the pace of innovation and clinical effectiveness.

5. Challenges & Future Work

AI/ML play a crucial role in developing HMN systems into intelligent, customized, and versatile platforms. Nevertheless, advancement is impeded by fragmented data [47], insufficient standardization [48], reproducibility challenges [49], and ongoing ethical and regulatory issues [50] such as algorithmic bias, patient consent, and ownership of data [51]. These challenges, well-known in biomedical AI/ML, also limit HMN translation, highlighting the necessity for proactive regulatory alignment and collaboration across disciplines.

Initial evaluations highlighted the enhancement of mechanical and structural characteristics via AI, including hydrogels with better swelling and longevity [11,12], biomimetic designs influenced by predictive ML [52], and AI-driven polymer assessments for formulation improvement [13]. Although useful, these contributions are still disjointed, highlighting the necessity for a developmental pathway that unifies and enhances these methods.

The AIM-DO developmental pathway is designed to address these barriers by emphasizing harmonized data curation, AI explainability, and scalable deployment. Building interoperable datasets of physicochemical and performance properties of hydrogels and microneedle geometries will enable predictive models for drug release, mechanical stability, and patient-specific responses. Such models can be embedded in automated design tools and smart manufacturing systems, supporting real-time quality control and adaptive feedback.

Several recent studies illustrate progress toward this goal: [53] suggested cloud-connected AI microneedles for instant dosage modification; [54,55] created AI-based dose-release systems that respond to patient physiology; [56] investigated targeted delivery for rheumatoid arthritis; while [46,57,58] enhanced AI-enabled formulation design, optimization, and bioavailability enhancement.

Ultimately, the AIM-DO developmental pathway envisions AI-enhanced clinical deployment, where closed-loop microneedle systems integrate personalized therapy, diagnostics, and biosensing—driving progress from laboratory innovation to real-world healthcare.

6. Conclusions

This review compiled new applications of AI/ML in HMNs across five areas: material and microneedle development, diagnostics and treatment, drug administration, drug creation, and health/agricultural sensing. The AIM-DO development pathway details a step-by-step process from data curation and predictive modeling through automated design, intelligent manufacturing, and finally to closed-loop clinical implementation.

Recent research indicates that AI/ML can replicate hydrogel behavior, forecast drug release, automate manufacturing, and boost biosensing, thus reducing design cycles and increasing reproducibility. By connecting interoperable datasets with predictive models of HMN structure–function interactions, these tools facilitate adaptive design and scaling through intelligent manufacturing.

Advancement, nonetheless, is hindered by ongoing obstacles: disjointed data, absence of standards, restricted reproducibility, and unaddressed ethical and regulatory concerns. Tackling these issues necessitates synchronized efforts in data science, engineering, clinical sectors, and policy areas, underpinned by unified validation methods and clear governance.

Building on these foundations, the AIM-DO developmental pathway outlines a strategy for progressing AI-driven HMNs from experimental models to widespread biomedical applications. These systems show potential for targeted drug delivery, less invasive diagnostic methods, and continuous health surveillance, ultimately connecting lab advancements with customized healthcare options.