Integration of stem cell therapy and artificial intelligence: From promise to clinical reality

Abstract

In the recent issue of World Journal of Stem Cells, a comprehensive analysis by Choudhery et al of artificial intelligence (AI) applications in stem cell-based therapy, emphasizing both the extraordinary potential and the critical challenges in the field’s development. With advances in AI and its use in medicine, a combinatorial approach to AI in stem cell-based therapy will shift it from traditional trial-and-error methods to data-driven, predictive systems that can analyze complex, multidimensional datasets. Recent advancements demonstrate that AI-controlled bioreactors can sustain consistent, industrial-scale stem cell production while maintaining therapeutic quality, as exemplified by systems such as deep learning-based automated cell-tracking technology that significantly reduces cell evaluation time. Data from clinical studies indicate that AI-optimized mesenchymal stem cell-based therapies, in which AI is used for cell isolation, culture, expansion, and quality control, as well as for patient selection, delivery method selection, follow-up assessment, and data interpretation, are safe and effective in achieving the desired outcome. Availability of Ryoncil, the first Food and Drug Administration-approved mesenchymal stem cell-based therapy for graft-vs-host disease, reflects increasing regulatory confidence in stem cell-based therapies. Nevertheless, considerable challenges persist, including difficulties with algorithm validation, variability in data quality across laboratories, and potential algorithmic bias. A combinatorial approach integrating AI and stem cell-based therapy will substantially advance the development of standardized protocols that ensure greater efficacy and safety.

Key Words: Artificial intelligence; Clinical translation; Machine learning; Regenerative medicine; Stem cell therapy

Core Tip: Over the past two and a half decades, despite remarkable advancements in cell-based therapy, inherent limitations have hampered its full potential as a clinical therapeutic modality. These limitations should be addressed to facilitate its routine application in clinical settings. As artificial intelligence (AI) continues to transform medicine, the combinatorial approach of integrating AI into stem cell-based therapies offers an excellent opportunity to achieve the best of both worlds. This combinatorial approach promises to revolutionize stem cell research by shifting from traditional trial-and-error methodologies to sophisticated, data-driven predictive systems that analyze intricate, multidimensional datasets. These datasets encompass a wide array of biological information, including genomic, proteomic, metabolomic, and transcriptomic data, facilitating a more comprehensive understanding of treatment outcomes. Moving forward, integrating AI with stem cell-based therapy is expected to significantly improve multiple aspects while ensuring stringent safety standards and more reliable outcomes.

This editorial refers to “Applications of artificial intelligence in stem cell therapy” by Choudhery et al, 2025; https://doi.org/10.4252/wjsc.v17.i8.106086.

---

INTRODUCTION

A paradigm shift in regenerative medicine

Since the inception of stem/progenitor cell-based therapy, experimental animal studies and clinical trials have demonstrated that it is a feasible and safe treatment strategy for diverse pathological conditions[1,2]. Furthermore, preclinical experiments and the results of the discussed clinical trials demonstrate that bone marrow-derived mesenchymal stem cells (MSCs), in addition to their safety, have the potential to improve cardiac function beyond what is achievable with current interventional and medical therapies[3]. Nonetheless, the efficacy remains questionable and warrants further study across a broader range of homogeneous, rigorous designs. The editorial delves into recent advances in stem cell-based therapy, its potential for integration with artificial intelligence (AI) to develop a combinatorial approach, and the limitations of such integration.

Stem cell-based therapy has emerged as a significant development in contemporary biomedical science, offering transformative prospects for the treatment and regeneration of tissues affected by diverse pathological conditions and traumatic injuries. This transformative advancement has led to a paradigm shift in treatment strategies, from symptomatic treatment to a permanent cure, particularly for diseases that were, until recently, considered incurable[4,5]. This is attributed to stem cells’ unlimited potential for self-renewal, their capacity to overcome lineage restrictions and differentiate into an array of specialized cells during postnatal growth and development, and their paracrine release of soluble and insoluble factors[6]. They are instrumental in the intrinsic repair process in various tissues and organs, including muscles, the heart, the liver, and the kidneys[7]. Given the reparative and regenerative potential of stem cells, research on their preclinical and clinical use has remained a focus of intense investigation over the past 2.5 decades. The groundbreaking reprogramming protocol for pluripotency induction in terminally differentiated somatic cells has given new impetus to its diverse applications, including regenerative medicine[8]. Stem/progenitor cells have been categorized in various ways: Anatomically, functionally, and by surface marker expression. Based on their tissue of origin, stem cells can be classified as extraembryonic, embryonic, or adult tissue-derived stem/progenitor cells. They have been extensively characterized in vitro and in vivo, with adult stem cells advancing into clinical trials for safety and efficacy assessment, and, in some cases, as in the case of MSCs, they have been launched as living biodrugs after approval from the drug agencies, as discussed in the subsequent sections[9]. On the other hand, induced pluripotent stem cells (iPSCs)-based platforms have been increasingly applied in the study of neurodegenerative disorders, facilitating disease modelling and the evaluation of targeted interventions. For instance, in 2024, a clinical trial was initiated to evaluate the transplantation of iPSC-derived dopaminergic neurons into patients with Parkinson’s disease, a condition characterized by striatal dopamine deficiency[10]. The phase I/II trial was conducted at Kyoto University Hospital in seven patients who received bilateral transplantation of iPSC-derived dopaminergic progenitors as an adjunct to maintenance anti-Parkinsonian medications, with slight adjustments as needed. The early results indicate that the treatment is safe, with no adverse events. However, secondary outcomes assessed motor symptom changes and dopamine production at a 24-month follow-up. Magnetic resonance imaging showed no graft overgrowth, while four patients showed improvements in the Movement Disorder Society Unified Parkinson’s Disease Rating Scale part III OFF score, and five showed improvements in the ON scores. Another study has reported that the usage of stem cells from cord blood has successfully led to the remission of human immunodeficiency virus in a female patient with leukemia[11]. Taken together, these diverse applications emphasize the dynamic expansion of stem cell-based modalities in contemporary clinical practice, enabling increasingly precise, strategically targeted therapeutic interventions with the potential to cure as-yet incurable diseases.

Limitations of stem cell-based therapy from a clinical perspective

Despite advances in stem cell-based therapy, it is not without limitations, and many caveats remain, raising concerns about its routine clinical use. These limitations encompass the time-consuming and painstaking conventional protocols for large-scale production, quality control, targeted delivery to the site of action, and post-treatment safety and efficacy[12]. For example, one of the more concerning long-term risks of stem cell-based therapies is the potential for uncontrolled cell growth that may culminate in tumorigenesis. This is particularly observed with pluripotent stem cells, i.e., embryonic stem cells and iPSCs. Similarly, specific patient populations may experience immune rejection and compatibility-related issues, as well as widespread infections. Another serious issue concerns the ethical implications of using embryonic stem cells derived from human embryos, which necessitate the destruction of the embryos despite the cells’ remarkable pluripotency. Given that stem cell-based therapies are a relatively recent area of clinical development, long-term safety data remain limited, underscoring the need for cautious implementation. Numerous studies predominantly assess short-term outcomes, leaving significant gaps in knowledge about long-term effects, including insufficient follow-up periods, lack of standardization, and the risk of delayed complications.

COMBINING AI WITH STEM CELL THERAPY: GETTING THE BEST OF BOTH WORLDS

The integration of AI with stem cell therapy embodies one of the most pioneering advancements in contemporary regenerative medicine. Choudhery et al[13] recently published a study in World Journal of Stem Cells provide a comprehensive review of the supportive use of AI applications in stem cell-based therapy to streamline clinical practice in this rapidly developing field. Although the integration of the two approaches is still in its infancy, the authors have provided ample evidence from published data that, as in other fields, AI has significant transformative potential to advance cell-based therapy from bench to bedside as a routine treatment option for diverse pathological conditions. Nevertheless, algorithm validation, data quality, accessibility, and ethical considerations remain significant challenges for this combinatorial approach as it advances toward clinical applications. As of the writing of this manuscript and using the search terms “stem cells” and “cell-based therapy”, more than 1685 clinical trials have been registered on Clinicaltrials.gov, investigating stem cell-based therapeutic approaches at both the early and advanced stages of clinical assessment for different pathologies. Moreover, more than 10 stem cell-based drugs have been launched in the market for diverse clinical applications after regulatory approval in different countries, including one each for heart failure and myocardial infarction, connective tissue disorders, rheumatoid arthritis, and spinal cord injury[9]. On the same note, regulatory agencies are supporting the integration of AI in both worlds. For example, the Food and Drug Administration (FDA, United States) plans to deploy generative AI platforms across all its research centers to transform the drug review process[4]. Such an initiative, combined with rapid technological advancements, positions AI-enhanced stem cell therapy at the forefront of widespread clinical implementation.

REVOLUTIONIZING THERAPEUTIC DEVELOPMENT THROUGH DATA-DRIVEN APPROACHES

Choudhery et al[13] have been methodical in outlining how AI can overcome some of the fundamental hurdles that have hampered the progress of cell-based therapy toward clinical settings. Despite the immense progress made in the field, there are as yet no optimized or standardized protocols for stem cell selection and characterization, nor for their quality control. Similarly, the protocol optimization has been labor-intensive, subjective, and prone to inconsistencies, hindering the scalability and reproducibility of cell preparations for clinical use. This heterogeneity in cell products has led to divergent published data, hampering interpretable analysis and hindering the attainment of a workable, definitive conclusion[14]. Indeed, the authors have rightly raised numerous valid points in the research and development context for the use of stem cells as living biodrugs, where AI’s impact will be substantial. The possible involvement of AI-driven tools may be, but is not restricted to: (1) Enhance understanding of stem cell behavior and characterize cell phenotypes; (2) Isolate and expand the cells in vitro; (3) Optimize culture and differentiation protocols; (4) Forecast risks such as graft failure or adverse events in patients; (5) Perform imaging-based classification of cells and colonies; (6) Develop in silico models of stem cell dynamics; (7) Optimize experimental in vitro and in vivo models and data interpretation; (8) Contribute to the development of off-the-shelf availability of stem cell-based living drugs; (9) Accelerate drug discovery to improve regenerative outcomes; and (10) Define and refine delivery methods for targeted cell therapies.

By providing both qualitative and quantitative insights at each procedural step, AI can facilitate the transformation of stem cell science into a more reproducible engineering discipline. For instance, computer vision algorithms, often employing convolutional neural networks, can analyze microscopic images to identify subtle morphological features associated with pluripotency or differentiation potential, thereby enhancing the selection process for high-quality iPSC colonies for subsequent development. Similarly, AI can integrate data regarding media components, growth factors, and culture parameters to identify optimal conditions that maximize cell yield and functionality. AI is therefore revolutionizing the development of stem cell therapies by infusing robust, data-driven decision-making into processes that were once predominantly guided by experience and conjecture[15].

The availability of AI-based machine learning algorithms enables analysis of large, multidimensional datasets, including multi-mic information (genomic, proteomic, transcriptomic, and epigenomic), whose complexity far exceeds human analytical capacity multi-mic information (genomic, proteomic, transcriptomic, and epigenomic), whose complexity far exceeds human analytical capacity[16]. Some of the main machine learning algorithms in regenerative medicine include logistic regression, K-means clustering, K-nearest neighbors, Decision trees/Random Forests, etc.[16]. These algorithms may also play a crucial role in predictive modeling, cellular analysis, image matching and interpretation, and the optimization of treatment protocols by enhancing the practicality of these advancements, as evidenced in the field. For example, Zhang et al[17] reported an automated mitosis-detection system for studying cell proliferation activity. They have reported a mitosis detection system that uses a graph neural network complemented by a differentiable optimization layer. Using time-lapse microscopy sequences of the cells cultured under four different conditions, the authors reported that the layer substantially improved detection performance compared with graph neural network-based link prediction alone. These results indicate the importance of explicitly integrating biological knowledge into deep learning models. Padovani et al[18] have developed Cell-ACDC, an open-source GUI-based framework for segmentation, tracking, and cell cycle studies. Hirose et al[19] at Tokyo Medical and Dental University have reported an AI-based system that rapidly identifies stem cells that can be grown into skin grafts. Such identification of stem cells based on their developmental potential is otherwise a demanding and time-consuming process. They have reported a deep learning-based automated cell-tracking technology that facilitates the non-invasive identification of a healthy population of human stem cells using phase-contrast images and distinguishes them from cells of non-stem-cell origin[19]. These findings are significant and expected to support label-free, non-invasive automation for stem cell quality control and high-throughput screening for stemness and functionality, and are gaining popularity for extrapolating laboratory methods to industrial-scale biomanufacturing[20].

Another aspect in this regard is the development of AI-controlled bioreactors, which represent a leap forward toward ensuring uninterrupted clinical-grade stem/progenitor cell production to meet large-scale demands while maintaining their stemness and therapeutic quality[21]. This advancement in stem cell manufacturing automation is expected to ease scalability (both scale-up and scale-down) at low cost, with improved quality and error-free performance. Lim et al[22] have developed a fully automated bioreactor system (fABS) for precisely controlling the proliferation and differentiation of human bone marrow-derived MSCs. The authors observed that the shear stress induced by the fABS primarily enhanced their proliferation and osteogenic differentiation. Moreover, hypoxia induced by fABS also promoted chondrogenic differentiation of the cells. The development of fABS and similar systems not only enhances mass production of stem/progenitor cells but may also facilitate their differentiation into the desired cell lineages. Recently, Herbst et al[23] reported the AutoCRAT platform for the end-to-end manufacturing of iPSC-derived iMSCs and iPSC-derived chondrocytes for the treatment of osteoarthritis, acute graft-vs-host disease, and chronic kidney disease. The system has integrated various analytical techniques, including microscopy, cell count, cell viability quantitative polymerase chain reaction, endotoxin assays, and the ChemStress Fingerprinting assay, for in-process continuous quality control of the product cells and for scale-up production. Similar studies reporting on similar platforms are[24,25]. For further reading, please refer to the following articles[26,27]. The European Innovation Council-funded AiPSC project (Grant agreement ID: 101071188) began in September 2022 and aims to develop a cost-efficient, rapid, and standardized AI-based microfluidic device to mass-produce personalized iPSCs for adoptive cancer immunotherapies, hematopoietic stem cell transplantation, and tissue regeneration[28]. To be completed in August 2026, it will also perform leading-edge single-cell genomics and bioinformatics research on iPSCs.

It is expected that the AI-driven smart bioprocessing approach will reduce heterogeneity in cell culture processes and, in turn, in cell products, accelerate the development of cell-based therapies, and optimize scaling-up, making it less laborious, more cost-effective, and of clinical-grade quality. These advancements in AI for cell-based treatment are transforming the landscape of stem cell bioprocessing and controlled differentiation into desired cell lineages, and doing so in a cost-effective manner.

AI also facilitates the optimization and standardization of processes that were previously difficult to manage. A recent review published by the European Medicines Agency in 2024 highlights the transformative potential of AI for advanced medicinal therapy products, focusing on predicting optimal conditions for generating specific cell types and enhancing the efficiency and consistency of regenerative treatments[29]. This report is based on current uses and potential applications of AI and machine learning technologies across the medical product lifecycle, as well as on the regulatory opportunities and challenges identified in published data. For example, machine learning models can precisely discriminate between undifferentiated and differentiated cells and predict their potential to develop into specialized 2D cells or 3D organoids[30]. More interestingly, deep learning was used to develop a model that could predict, from images of organoids, whether differentiation was progressing appropriately during organoid development. These advancements fundamentally alter the approach to developing cell-based therapies, thus enabling researchers to anticipate differentiation outcomes and adjust culture conditions accordingly, rather than relying solely on trial-and-error.

AI INTEGRATION WITH CELL-BASED THERAPY FROM A CLINICAL PERSPECTIVE

It is pretty evident from the clinical landscape of 2025 that AI-enhanced stem cell therapies have successfully transitioned from theoretical potential to real patient outcomes. Recent meta-analyses and clinical trials report impressive success rates across diverse applications[1,2]. A meta-analysis by Rahmadian et al[31] found that intra-articular injection of low-dose (< 25 million cells) MSCs significantly improved symptoms in patients with knee osteoarthritis at the 12-week follow-up. The meta-analysis included six randomized clinical trials with eight independent treatment arms (300 patients) and reported a pooled standardized mean difference in functional improvement of approximately -1.35 (95% confidence interval: -1.97 to -0.74), indicating a moderate-to-substantial treatment effect[31]. Notably, the authors inferred that arbitrarily assigned low doses (≤ 25 million MSCs) yielded statistically significant improvements with no additional benefit observed at higher doses. Incidentally, most studies used allogenic cells for treatment. These data not only highlight that treatment with the minimum effective dose could improve cost-effectiveness, expand access, and simplify large-scale implementation, but also raise the possibility that AI can optimize therapeutic protocols by extracting insights from patient outcome datasets that are challenging to analyze conventionally. Alborzi and Abadi[32] have recently exploited the data-analytic potential of various AI models (reported between 2018 and 2025) to assess their performance in early detection and prediction of osteoarthritis and osteoporosis. The authors used 33 studies employing imaging modalities (i.e., X-rays, magnetic resonance imaging, dual X-ray absorptiometry). They evaluated the accuracy, reliability, and clinical applicability of the models, finding that the top-performing models achieved 97% accuracy. The authors claimed that AI-driven models outperformed conventional diagnostic tools in terms of accuracy and early disease detection.

Another example concerns the role of AI in stem cell therapy applications for cardiovascular-related pathologies. In patients with chronic heart failure, a recent assessment of bone marrow mononuclear cell therapy showed improved left ventricular ejection fraction when AI-guided patient selection and monitoring protocols were employed[33]. In essence, AI algorithms assisted in identifying patients most likely to respond, such as by stratifying patients based on biomarkers, including inflammation, or specific genetic profiles, and by adjusting monitoring in real time, thereby augmenting the therapy’s efficacy. This also highlights the importance of protocol optimization and patient stratification, and how AI can facilitate it by uncovering subtle patterns in patient data that correlate with improved responses. An interesting new development in the use of AI-integrated stem cell-based therapy in experimental settings is the creation of an AI-generated virtual beating heart to test the performance of human and mouse cells in myocardial repair. The Jackson lab has reported a project, Cardioverse, to reduce the use of large-animal studies, streamline FDA approval, and ensure the safer, faster delivery of new therapies[34]. The application of AI to outcome prediction has also proven invaluable for managing complex neurological disorders such as Parkinson’s disease. Recent phase I/II studies utilizing AI-optimized autologous iPSC-derived dopaminergic neural progenitor cells for Parkinson’s disease demonstrate promising early efficacy[35], providing hope that AI can assist in refining cell replacement therapies for neurodegenerative diseases through the optimization of cell type, dosage, and timing tailored to individual patients. For further reading, please refer to the review by Suresh et al[36], which provides an overview of AI-driven approaches to stem cell-based therapy for diseases and disorders.

It is also essential to recognize that AI’s advantages include the ability to understand failure modes and subtle effects. By aggregating patient data from clinical trials, AI can facilitate the early detection of rare adverse events and explain why specific therapies are effective for some patient groups but not others. These clinical examples underscore a fundamental point: The integration of AI and stem cell therapy is now delivering tangible benefits to patients across various medical fields, including orthopedics, cardiology, and neurology, thereby transforming the discipline from potential to reality.

ADDRESSING CRITICAL CHALLENGES AND LIMITATIONS

While recognizing these advancements, the authors provide a balanced assessment of the primary challenges that remain, with the foremost issues being algorithm validation and the “black box” nature of many AI systems[13]. Current state-of-the-art AI and deep learning models can frequently produce accurate predictions or classifications; however, they do so without providing transparent reasoning processes[37]. This lack of transparency poses significant obstacles to regulatory approval and clinical acceptance, especially in life-critical therapeutic contexts, where understanding the rationale for decisions is as essential as the decisions themselves. When an AI recommends a specific cell line for a patient or identifies a culture batch as high-risk, clinicians need to understand the AI’s justification, underscoring the importance of advancing explainable AI in healthcare.

One significant challenge concerns data quality for developing AI algorithms, as the efficiency of these algorithms and the quality of their outputs depend on the quality of the training data[38]. Given the heterogeneity of experimental and clinical data across multiple laboratories worldwide engaged in stem cell research, using different protocols, and serving patient populations with mixed demographics, there is an apparent dearth of standardized datasets[39]. This renders it challenging for an algorithm trained in one laboratory to achieve comparable performance in different settings. It is widely recognized that standardized reference datasets and benchmarking criteria are needed to facilitate comparisons among AI methodologies. An important step in this regard will be the establishment of an international consortium to develop open-access databases that comprise cell images, genomic profiles, or bioprocessing metrics, and to establish consensus on performance evaluation and assessment. Without such standards, it will be challenging to validate AI models trained on diverse datasets for meaningful interpretation[40]. On the same note, in the absence of comprehensive methodological details and reports of negative results, it will be a futile effort[41]. Resolving these issues will substantially improve the reproducibility and reliability of AI-based data interpretation.

An additional ethical concern pertains to data governance: Patient data used to train AI algorithms has considerable value, yet current frameworks governing data sharing, intellectual property rights, and patient consent remain evolving[42,43]. Key questions regarding data ownership (including that of AI models derived from these data), safeguarding patient privacy while facilitating data pooling, and managing incidental findings will require robust guidelines. Stem cell researchers should establish comprehensive policies and adopt procedures to protect patient rights while enabling data sharing necessary to realize the full potential of AI.

REGULATORY INNOVATION AND COMMERCIAL VIABILITY

As mentioned earlier, drug regulatory agencies in different countries have approved multiple stem cell-based living biodrugs for use in patients with a range of pathologies, including acute myocardial infarction, osteoarthritis, perinatal fistula in Crohn’s disease, type 1 diabetes, critical limb ischemia, etc.[9]. These products primarily use MSCs from various tissue sources, whereas a few, such as Omiserge (Omidubicel), use hematopoietic stem cells[44]. In December 2024, the FDA approved Ryoncil (remestemcel-L) as the first MSC-based therapy for pediatric patients with steroid-refractory acute graft-vs-host disease[45]. The product is for children aged ≤ 18 years and shows a 70% response rate and approximately 70% 6-month survival among treated patients. This is a landmark approval, as it not only addresses a life-threatening condition in children with poor prognoses but also validates the efficacy of an MSC product in a primary regulated market for the first time. As stated in the FDA’s announcement, the approval of Ryoncil exemplifies the agency’s commitment to supporting innovative cell-based therapies and encourages the development of more advanced regenerative products. Choudhery et al[13] appropriately highlight this achievement, as it provides a crucial confidence boost and a strategic framework for upcoming AI-enhanced cell therapies to secure approval. In addition to this pool of stem cell-based living drugs, pluripotent stem cell (PSC)-based products are elegantly reviewed by Kirkeby et al[46]. Despite their pluripotency, ethical issues, safety concerns, the need for immunosuppression for graft acceptance, and tumorigenesis, their regenerative potential has been hampered[47]. Since the report of the first phase I clinical trial by Geron Corporation to test a human PSC-derived product: The hESC-derived oligodendrocyte progenitor product (GRN-OPC1) for the treatment of thoracic spinal cord injury (NCT01217008), followed by testing of the MA09-hRPE product from Advanced Cell Technology, and at the RIKEN institute in Japan for testing the first autologous human iPSC-based product for treatment of age-related macular degeneration (UMIN000011929), a string of clinical trials has been launched to test PSC-based products. The authors have reported 115 clinical trials across 19 countries testing human PSC-derived products for 34 indications as of December 2024[46]. The majority of these trials are phase I/IIa, designed to test safety and feasibility, with only a few yet designed to test efficacy.

Representing one of the most complex biologics, adding a novel class of drugs to the already diverse array of pharmaceuticals and biopharmaceuticals, and requiring that the “drug molecule has to be a living entity”, it is bringing a dogmatic change to the official definition of drugs. Similarly, to adopt this new class of “living drug”, the quality control testing and regulatory requirements need to be revisited. The integration of AI into cell-based therapy has fundamentally reshaped the landscape of AI-enhanced stem cell therapies, making them a superior therapeutic option, with recent approvals setting important precedents in the field[48]. The FDA has shown significant support for next-generation living biodrug products that integrate AI in their development processes. This is exemplified by numerous new programs using iPSCs that have received clinical trial approval and incorporate AI optimization into their design. These include OpCT-001, intended for retinal degenerative diseases, and FT819, an off-the-shelf iPSC-derived chimeric antigen receptor T cell therapy for systemic lupus erythematosus[49].

For example, the FT819 program is unique due to its innovative approach, which employs AI-optimized methods in an allogeneic chimeric antigen receptor T therapy targeting an autoimmune disorder, integrating cell therapy, gene editing, and AI-driven personalization. Similarly, OpCT-001 is the first iPSC-based therapy for photoreceptor diseases to undergo clinical testing. These examples reflect regulatory agencies’ growing inclination to use AI to enhance product quality and consistency, enabling more precise patient targeting and better prognoses. In addition to the United States FDA, other regulatory agencies worldwide are making significant progress. A typical example is China, where the first MSC therapy was approved in 2023, supported by strong government initiatives in AI for biotechnology[50]. Overall, AI has transitioned from a supplementary tool to a core component of competitive strategies in regenerative medicine.

ETHICAL FRAMEWORK AND REGULATORY EVOLUTION

As with any other technology, the use of AI-enhanced stem cell therapy in patients requires proactive consideration. With advances in stem cell-based therapies in regenerative medicine, the general concerns raised by the research community about ethical issues must be addressed; accordingly, a supplementary working framework should be established and updated regularly to address emerging challenges. This working framework should be established on the four basic principles of ethics: Autonomy, beneficence, non-maleficence, and justice[51]. The basic principles should be considered to address issues of patient data privacy, algorithmic bias, transparency to overcome black-box AI decisions, and accountability for potential treatment errors, which may have long-lasting impacts on the patient[52]. A crucial immediate measure is to establish transparent data governance and patient consent protocols suitable for the AI era, as the AI system is expected to handle patient-specific biological and genetic data, along with their demographic information. In this regard, informed consent from the patient should be mandatory to ensure that she knows that AI algorithms will handle her data or biological samples. Hence, data-sharing policies safeguard patients’ interests as we enter a new era of fostering innovation. In other words, embracing innovation should not come at the expense of patients’ right to know or the privacy of their sensitive information[53]. Although some disjointed regulatory approaches are developing such policies for the use of cell therapies with AI, it will require a more serious and concerted effort to establish a meaningful framework. Some nations have established expedited pathways for regenerative medicine, such as Japan’s Sakigake and adaptive licensing frameworks. In contrast, guidelines for AI in medicine are still under development globally[54]. Regulatory agencies are beginning to collaborate; for instance, the FDA, the European Medicines Agency, and Health Canada jointly issued guiding principles for machine-learning-enabled medical devices in 2023, highlighting transparency, performance oversight, and human supervision[52]. Such initiatives should be expanded to address AI in cell and gene therapy products specifically.

FUTURE DIRECTIONS AND RESEARCH IMPERATIVES

It is generally believed that future AI-driven approaches and advanced algorithms will offer exciting solutions to elucidate the underlying molecular and cellular mechanisms and to help fill these knowledge gaps. Given the promise of AI and stem cell-based therapies, it is anticipated that integrating these novel, highly sophisticated fields will help realize the full potential of stem cell-based therapy in regenerative medicine, which currently appears to be stagnating due to challenges in translating bench-to-bedside. Table 1 summarizes areas in which AI may deliver the best outcomes in stem cell-based therapy.

Table 1 Some of the benefits of artificial intelligence integration with stem cell therapy.

Benefits of artificial intelligence integration with stem cell therapy
In vitro expansion of stem/progenitor cells and their bioprocessing
In vitro characterization of stem cells
Quality control of the purified and expanded cells for experimental and clinical uses
Spotting healthy cells with reduced DNA damage to avoid the chances of tumorigenesis
Priming the cells for enhanced stemness and functions
Using AI and CRISPR to enhance stem cell control
Enhancement of paracrine activity with the desired profile
Development of AI-generated virtual models to reduce dependence on experimental animal models
Extrapolation of experimental data for clinical application
Designing of the clinical trials
Clinical outcome prediction

More importantly, all these can be achieved in a time- and cost-efficient manner while maintaining accuracy and reliability. AI: Artificial intelligence.

For instance, researchers are concerned about advancing pluripotent stem cells to clinical use due to inherent safety risks. However, recent advances in machine learning and deep learning for refining iPSC classification, monitoring cell function, and genetic analysis have led to significant improvements across diverse applications of iPSCs, including drug development, disease modelling, and transplantation therapy. A recent systematic scoping review by Vo et al[55], which included 79 studies, concludes that AI technologies have laid a foundation for future iPSC-based technologies. Al Khafaji et al[56] highlight the potential of AI to design cardiac patches and 4D bioprinting to create biomaterials for cardiac tissue engineering, thereby enabling rapid treatment of heart disease. These ideas are currently under discussion in medicine, and an investigation is underway to assess their effectiveness[56]. However, significant challenges exist in the evolution of AI in this field. Obstacles such as limited access to datasets and inadequate, incomplete reporting standards hinder overall progress. Furthermore, as AI-driven innovations in stem cell therapy evolve, the ability to evaluate treatment-related risks remains essential to ensure safe, sustained, and clinically meaningful prognosis. Beyond assessing treatment risks with AI, ethical and regulatory considerations must be addressed and integrated into stem cell therapy.

CONCLUSION

In conclusion, the appropriate and ethical use of AI can accelerate scientific advancement and enable the development of regenerative and personalized medicines that, overall, can improve patient prognosis. By using AI to address essential research imperatives, we can optimize our future treatment strategies. However, a high standard of scientific rigor is needed to ensure that AI-implemented stem cell methodologies and protocols produce reliable data, as AI models remain in their infancy. Strict criteria must be developed to evaluate the quality and reliability of datasets while ensuring patient confidentiality. A multidisciplinary collaborative approach is imperative to validate guidelines to ensure patient safety and efficacy.