Revolutionizing spinal interventions: a systematic review of artificial intelligence technology applications in contemporary surgery

Abstract

Leveraging its ability to handle large and complex datasets, artificial intelligence can uncover subtle patterns and correlations that human observation may overlook. This is particularly valuable for understanding the intricate dynamics of spinal surgery and its multifaceted impacts on patient prognosis. This review aims to delineate the role of artificial intelligence in spinal surgery. A search of the PubMed database from 1992 to 2023 was conducted using relevant English publications related to the application of artificial intelligence in spinal surgery. The search strategy involved a combination of the following keywords: "Artificial neural network," "deep learning," "artificial intelligence," "spinal," "musculoskeletal," "lumbar," "vertebra," "disc," "cervical," "cord," "stenosis," "procedure," "operation," "surgery," "preoperative," "postoperative," and "operative." A total of 1,182 articles were retrieved. After a careful evaluation of abstracts, 90 articles were found to meet the inclusion criteria for this review. Our review highlights various applications of artificial neural networks in spinal disease management, including (1) assessing surgical indications, (2) assisting in surgical procedures, (3) preoperatively predicting surgical outcomes, and (4) estimating the occurrence of various surgical complications and adverse events. By utilizing these technologies, surgical outcomes can be improved, ultimately enhancing the quality of life for patients.

Introduction

Spinal surgery, as a significant specialty in the field of medicine, is at a turning point. Thanks to unprecedented medical data and advanced computational technology, medicine is entering a new era of precision healthcare. With the development of the information age, medical data is rapidly accumulating in digital form, injecting powerful momentum into the technological advancement and innovation in the medical field. By combining massive and diverse medical data with efficient computational methods, we are harnessing the potential of artificial intelligence (AI) to accelerate progress in the field of spinal surgery [1, 2].

AI is a potent tool in today's technological landscape, significantly enhancing the accuracy of prediction, prognosis, diagnosis, and biomechanical evaluation of spinal diseases [1]. Compared to traditional statistical methods, AI not only demonstrates superior capabilities but also holds tremendous promise in the medical domain. With the continuous advancement of hardware, software, and computational scientific technologies, AI is poised to gain broader recognition and application in the near future.

However, we have not yet achieved fully automated intelligence. Nevertheless, we have made significant strides, being able to develop systems using AI tools that can simulate human intelligence features—learning from extensive data, making decisions, and providing recommendations [3]. Through artificial neural networks and powerful machine learning techniques, we can dynamically learn from data, providing valuable insights for future actions and decisions. AI represents a vast technological domain, and machine learning, as a vital branch of AI, leverages computer algorithms to learn from data and past experiences, constructing intelligent models, contributing to the development and progress of the medical field [2, 4].

AI has made remarkable strides across various medical fields, each benefiting from advancements in machine learning, neural networks, deep learning, and computer vision. In radiology, AI has significantly enhanced the accuracy and efficiency of image interpretation. Machine learning algorithms, particularly convolutional neural networks (CNNs), have been developed to detect anomalies such as tumors, fractures, and other pathologies in imaging studies including X-rays, CT scans, and MRIs [5]. These AI systems can analyze vast amounts of imaging data rapidly, providing radiologists with precise diagnostic insights and reducing the likelihood of human error [6].

AI applications in cardiology have focused on predictive analytics and personalized medicine. Neural networks have been employed to predict the occurrence of cardiovascular events such as heart attacks and strokes by analyzing patient data, including electronic health records (EHRs), imaging, and genetic information. These predictive models assist clinicians in identifying high-risk patients and tailoring preventive strategies accordingly [7]. Furthermore, AI has been used in the interpretation of echocardiograms and electrocardiograms (ECGs), enhancing the detection of cardiac abnormalities and improving patient management [8].

In oncology, AI has revolutionized cancer diagnosis, treatment planning, and prognosis. Deep learning models, especially those based on CNNs, have demonstrated high accuracy in detecting and classifying various types of cancers from histopathological images, radiographs, and genomic data [9]. AI algorithms can identify subtle patterns that may be missed by human observers, leading to earlier and more accurate diagnoses [10]. AI algorithms are used to detect and monitor conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration through retinal imaging [9, 11]. AI assists pathologists in analyzing biopsy samples, automating the detection of abnormal cells, and providing quantitative assessments of tissue samples [12]. AI triage systems help prioritize patients based on the severity of their conditions, optimizing resource allocation and improving patient outcomes [13]. Table 1 summarizes the applications of artificial intelligence in various medical fields.

Table 1 Applications of artificial intelligence in various medical fields

In the context of spinal surgery, AI offers solutions to several specific challenges that traditional methods struggle to address. Spinal surgery demands extremely high precision, as millimeter-level errors can lead to severe consequences. Traditional methods often struggle to achieve the desired accuracy within the complex anatomical structures involved. AI, through advanced image recognition and real-time navigation technologies, can significantly enhance surgical precision and reduce human errors. Accurately predicting individual patient outcomes is also challenging with traditional methods. AI can analyze vast amounts of historical data, combined with individual patient characteristics, to provide more precise prognostic predictions, aiding doctors and patients in making more informed decisions. Additionally, the complication rate in spinal surgery is relatively high. AI can analyze patient risk factors and surgical technique details to identify potential complication risks and offer personalized prevention strategies, thereby reducing the incidence of complications. Selecting the optimal surgical approach is crucial for the success of the surgery. AI can recommend the best approach based on the patient's specific conditions, such as anatomical variations, lesion locations, and previous surgical history, balancing surgical effectiveness and minimal invasiveness. Therefore, AI demonstrates significant potential in enhancing precision, predicting outcomes, reducing complications, and selecting surgical approaches in spinal surgery.

Additionally, AI offers solutions beyond traditional methods in several key areas. In preoperative planning, AI can generate precise 3D models and simulate different surgical plans, helping doctors choose the best strategy [14]. During surgery, AI-driven augmented reality technologies can provide real-time navigation and visualization of critical structures to surgeons, enhancing intraoperative assistance [15]. Postoperatively, AI can personalize rehabilitation plans and monitor patient recovery in real-time through wearable devices, ensuring tailored and effective rehabilitation [16]. Furthermore, AI systems can learn from each surgery, continuously optimizing algorithms to improve future surgical success rates [17].

To date, there has been no comprehensive review summarizing these developments. This paper aims to explore the full potential of AI technology in driving progress and transformation within the field of spinal surgery. AI not only aids physicians in making more accurate diagnoses and formulating treatment plans but also propels the healthcare system towards becoming more intelligent, efficient, and personalized. We will examine the applications of AI in spinal surgery, discussing its potential benefits, challenges, and future directions. By integrating AI technology with medical practice, our goal is to provide patients with higher quality and more precise medical services, thereby contributing to the advancement of the entire medical field. At this pivotal moment, we foresee AI having a positive impact on the future of spinal surgery, positioning it as a pioneer in the era of precision medicine within the healthcare sector.

Method

Search strategy

A comprehensive search of original articles was conducted through the PubMed search engine to identify the application of Artificial Neural Networks (ANN) in spinal surgery. This review aimed to encompass all full-text publications in English biomedical journals. The following keyword combinations were searched in titles and abstracts: "Artificial neural network," "deep learning," "artificial intelligence," "spinal," "musculoskeletal," "lumbar," "vertebra," "disc," "cervical," "cord," "stenosis," "procedure," "operation," "surgery," "preoperative," "postoperative," and "operative." These keywords were selected as they might be mentioned in the titles or abstracts of relevant articles. A comprehensive search was conducted for the period between 1992 and 2023. The search was carried out in August 2023.

Inclusion and exclusion criteria

All retrieved literature underwent a screening process within the database. Each article was independently reviewed by two reviewers. In case of discrepancies, the article was referred to another reviewer for resolution. Only literature that specifically emphasized relevant research and advancements of AI in spinal disease surgical treatment was included.

Literature related to other diseases, animal studies, and review publications were excluded from consideration. Animal studies were excluded to focus on directly relevant human clinical applications, thereby enhancing the clinical relevance of the study results. Review articles were also excluded to avoid double-counting original studies and ensure our analysis is based on primary data. Only English literature was included, considering the language proficiency of the research team and the fact that most high-impact studies tend to be published in English journals. The time range of 1992 to 2023 was chosen to capture the early applications of artificial neural networks in medical imaging. The search was limited to PubMed due to its comprehensive coverage of biomedical literature.

Data synthesis

Results of all identified studies were summarized in a descriptive table. The table included author names, publication year, treatment stage, study sample, primary mentioned artificial intelligence models, whether compared with traditional non-artificial intelligence models, research focus, and key findings or conclusions. All results were organized in chronological order.

Results

Statistics

The reviewer identified and screened 1,182 articles. After screening, it was determined that 1,082 articles were not relevant. Subsequently, a review was conducted on the remaining 100 articles, and a thorough assessment of eligibility criteria was performed on the full texts. Among these, 5 reviews, 3 animal studies, and 1 retracted article were excluded. Finally, 91 articles were included in this review. The flowchart of the literature review process is illustrated in Fig. 1. AI technology has been applied before, during and after spinal surgery. A summary of all literature is shown in Table 2.

Fig. 1

Flow chart of literature search, review, and selection

Table 2 A list of papers on artificial intelligence used in spinal surgery

Preoperative

Some literature describes the application of AI in preoperative settings for spinal surgeries. One focus of a study is utilizing AI to assess the suitability of specific surgical techniques for particular spinal disease patients, enhancing patient selection and ensuring safer surgeries through preoperative simulations. Another study utilizes a hybrid AI model, combining data-driven machine learning with expert models to provide surgical recommendations for patients, showcasing the potential of AI to improve surgical decision accuracy. AI further simplifies patient triage for spinal surgeries by integrating questionnaire data and MRI reports, aiming to identify suitable candidates and optimize outpatient efficiency. Machine learning techniques are applied to predict the likelihood of patients undergoing selective lumbar surgeries based on clinical and radiological features, demonstrating hope for improving surgical referrals and healthcare system efficiency. AI-based 3D imaging technologies become valuable tools for evaluating the safety zone of spinal surgeries and understanding variations in relationships between spinal skeletal and neural tissues, highlighting the practicality of AI in enhancing surgical safety. Additionally, utilizing machine learning algorithms to analyze patient data and predict the invasiveness of specific spinal surgeries demonstrates the potential of machine learning in personalized patient care and surgical planning. The document also emphasizes research focused on predicting bone health using machine learning models, particularly in the context of spinal reconstruction surgeries, demonstrating high accuracy in identifying osteoporosis and underscoring the need for further exploration in this field. Collectively, these studies underscore the transformative potential of AI and machine learning in optimizing preoperative preparations for spinal surgeries, ultimately paving the way for wiser and more efficient interventions in spinal surgeries.

Intraoperative

AI has also found widespread applications during surgical procedures. One study focuses on a unique case of a Jehovah's Witness patient refusing the use of blood products during spinal surgery, effectively managing intraoperative hypotension using an AI tool called the Hypotension Prediction Index (HPI). Another study emphasizes the application of deep learning and CNN to enhance the accuracy and efficiency of pedicle screw insertion during spinal surgeries. Additionally, research explores the automatic identification of neuro-monitoring documents within Electronic Health Records (EHR) using deep learning models and Natural Language Processing (NLP) techniques, aiding in quality improvement efforts. Accurate segmentation of vertebrae is also a focus, with the aim of reducing surgery time and risks in robot-assisted surgeries. The role of AI in assisting surgeons during Percutaneous Transforaminal Endoscopic Discectomy (PTED) is investigated, highlighting the potential of CNN in tissue recognition and improving surgical accuracy. Another innovative study explores the use of Augmented Reality (AR) and AI-guided systems during percutaneous vertebroplasty, providing precise navigation and reducing radiation exposure. Additionally, machine learning algorithms are utilized to predict bone cement leakage during vertebroplasty, aiding in making informed surgical decisions. Lastly, the integration of machine learning and Natural Language Processing (NLP) is explored to predict and identify vascular injuries during anterior lumbar surgery, demonstrating the potential to enhance surgical quality and safety reporting. Overall, these studies collectively demonstrate the transformative potential of AI and machine learning in improving the surgical processes of spinal surgeries.

Postoperative

Numerous studies highlight the applications of AI in postoperative scenarios related to spinal surgeries. Key areas of focus include predictive models for surgical outcomes, particularly complications and readmission rates. To address the opioid crisis, machine learning aids in predicting long-term opioid use postoperatively. The role of machine learning in predicting resource utilization (including hospital costs and length of stay) and in image recognition of spinal implants is crucial. Furthermore, some researchers use machine learning to predict the survival rates of patients with spinal metastases, achieving significant advancements in validating predictive algorithms. In summary, machine learning demonstrates tremendous potential in predicting postoperative outcomes, with the potential to enhance patient treatment outcomes and efficient resource allocation in spinal surgeries.

Discussion

The application of AI technology in various surgical disciplines is rapidly evolving, bringing revolutionary changes to patient care. AI-assisted methods can reconstruct 3D liver models for virtual hepatectomy, allowing for the analysis of total liver volume, tumor-involved segments, and tumor-free segments [109]. Traditionally, decisions for transplant candidates have heavily relied on subjective assessments, but AI can also aid in making complex liver transplantation decisions [110]. In gastrointestinal surgery, AI plays a crucial role in early cancer detection and predicting lymph node metastasis. Because the diagnostic ability and accuracy of endoscopists for early gastrointestinal cancers vary, AI combined with the expertise of endoscopists can improve the accuracy of early gastrointestinal cancer diagnosis [111]. Additionally, AI can predict lymph node metastasis preoperatively, thereby minimizing the need for additional surgeries after endoscopic resection of colorectal cancer [112]. In breast surgery, AI assists in breast reconstruction and breast cancer screening. In the field of breast reconstruction, AI improves surgical procedures, enhances outcomes, and simplifies decision-making [113]. Additionally, AI aids human experts in breast cancer screening [114]. Predicting individual mortality rates for congenital heart disease patients undergoing cardiac surgery and estimating mortality risk for patients receiving left ventricular assist device therapy are critical applications of AI [115, 116]. Comprehensive preoperative understanding of lung anatomy is crucial for accurate surgical planning and case selection. Identifying intersegmental planes on CT scans is extremely challenging. AI enables the 3D visualization of complex anatomical structures, such as lung segment partitions, vascular branches, and bronchial anatomy, aiding in the preoperative planning of segmental lung resection [117]. In the field of neurosurgery, artificial intelligence can assist in planning surgical pathways using preoperative MRI images [118]. Additionally, it can analyze biopsy samples during surgery to distinguish between healthy tissue and benign or malignant tumor tissue [119]. In urological surgery, AI contributes to the early diagnosis of renal cell carcinoma and predicts recurrence following prostate cancer resection. Using NMR-based serum metabolomics and self-organizing maps, AI can facilitate the early diagnosis of renal cell carcinoma [120]. Additionally, AI can automatically predict early recurrence in patients following robot-assisted prostatectomy [121]. AI technology continues to enhance surgical precision, improve patient outcomes, and facilitate personalized care across these specialized fields. The application of artificial intelligence technology in various surgical fields is partially illustrated in Table 3.

Table 3 Application of artificial intelligence technology in various surgical fields

This article provides a comprehensive overview of various research efforts utilizing the power of AI in the field of spinal surgery. These studies aim to develop and validate predictive models related to spinal surgery, aiming to enhance surgical outcomes, predict postoperative complications, and ultimately improve patient prognosis. Researchers employ a range of AI methods to analyze extensive datasets. These models, once trained, can assess surgical indications, assist in the surgical process, and predict preoperative surgical efficacy, postoperative complications, readmission rates, and non-home discharge outcomes. Additionally, most articles emphasize the importance of external validation, with many studies highlighting the necessity of validation using independent datasets. Furthermore, the document advocates for standardized reporting and validation methods, aligning with the broader scientific concern for the repeatability and reliability of AI-based research.

Artificial intelligence has preoperative applications, specifically in identifying surgical indications and detecting pedicle screws from previous spinal surgeries. Several researchers are leveraging AI to determine surgical indications and evaluate patient suitability for spinal surgery. This involves humans augmented by AI [21, 30], decision making guided by AI [39, 72, 79], and fully autonomous AI systems [64]. They also seek to choose appropriate surgical approaches for patients who have already decided on surgery [51, 60]. AI can analyze a large number of cases to identify patient populations suitable for specific types of surgery and predict postoperative outcomes. This helps both doctors and patients make more informed surgical choices, ultimately improving the success rate of surgeries and the quality of life for patients. It also alleviates the burden on doctors, saves time and costs, and enhances the efficiency of healthcare services, allowing more patients to receive timely and personalized medical care. Artificial intelligence and machine learning technologies can evaluate whether patients are suitable for surgical treatment based on imaging examination results and their medical history; however, the final decision still requires clinical judgment by physicians.

Some studies aimed to use deep learning algorithms to accurately identify pedicle screws implanted during spinal surgery based on radiographic images (anterior–posterior and lateral views of the spine) [59, 71]. This could assist clinicians in evaluating previous surgeries, especially revision surgeries. The research compared the performance of different deep learning algorithms, including standard deep neural network models using transfer learning, and two automated platforms: Google Auto ML and Apple Create ML. The evaluation was based on accuracy and recall metrics for identifying the implants. By demonstrating the effectiveness of deep learning in identifying surgical implants, these studies could aid in patient assessment and decision-making during spinal surgery. Some researchers have deployed this algorithm as an accessible smartphone application for further evaluation, improvement, and eventual widespread use, which will serve as a valuable supplement to clinical practice.

Artificial intelligence is utilized intraoperatively to assist and enhance the surgical process during operations. The integration of AI technology into spinal surgery to assist during the surgical process is a prominent research area [27, 43, 45, 50, 57, 82, 89]. It's worth mentioning a particular study that discussed a case in which a patient, due to religious beliefs, declined a blood transfusion. During the spinal surgery, AI and machine learning were employed to implement a fluid-restricted treatment plan. The patient underwent a major spinal surgery, and an innovative monitoring tool called the Hypotension Prediction Index (HPI) was utilized to predict and manage intraoperative hypotension [27]. AI technology provides crucial information for surgical planning by analyzing a vast amount of patient imaging data and considering the patient's anatomical structure and medical condition. It assists in planning surgical pathways, such as screw pathways, and detects intraoperative damage or anomalies, predicting potential issues and challenges in advance. AI also enables real-time analysis of patients' physiological data, surgical images, and monitoring information, providing surgeons with real-time feedback and aiding in decision-making, including adjustments to the use of medications during the surgery. This assistance helps tailor the surgical strategy, minimizing surgical risks to the greatest extent. The application of artificial intelligence can significantly reduce the risks associated with surgery. However, the formulation of surgical strategies should ultimately be decided by clinical physicians.

Artificial intelligence is used postoperatively to predict surgical complications, outcomes, discharge status, hospitalization costs, opioid use, and survival in patients with spinal metastases or tumors. The majority of researchers focus their studies on postoperative complications. These include postoperative pain [19], functional decline [73], the need for blood transfusions after surgery [26], postoperative infections [26, 34, 47, 80, 83, 84], delirium [32], urinary retention [48, 55], post-cervical surgery C5 palsy [52], and even patient mortality after surgery [25, 102, 104]. Some patients may experience neurological symptoms postoperatively, such as swelling, numbness, limp, and weakness of the lower limb [58, 100]. Certain patients may experience extended hospital stays [20, 26, 37, 87], recurrence of symptoms after surgery [44, 54, 61, 77, 108], the need for reoperation [25, 26, 48], readmission after discharge [23, 26, 56, 62, 67, 98], and various other complications and adverse events [90, 91, 102, 104]. Studying these complications can identify potential issues or risk factors, helping physicians understand in advance the risks patients may face. This understanding enables better communication with patients regarding potential risks and postoperative recovery, assisting patients in making informed decisions about whether to undergo surgery. Physicians can also adjust surgical plans based on predictive results, choosing the most suitable treatment path for the patient and taking appropriate measures to minimize potential risks. This information provides a basis for improving medical practices and enhancing patient safety. Most authors believe that AI and machine learning technologies can assist clinicians in decision-making by incorporating relevant risk factors, thereby aiding in the prevention and treatment of various postoperative complications. However, the overall clinical efficacy remains insufficient at present, necessitating further validation and refinement.

In the field of surgical advancements, a group of renowned researchers is turning to the use of machine learning and AI to predict surgical outcomes. By utilizing comprehensive datasets that encompass patient clinical data, imaging data, medical history, and the specific type of surgery conducted, these researchers aim to develop predictive models. This involves humans augmented by AI [35, 44, 46, 63, 76, 91], decision making guided by AI [33, 78, 107], and autonomous AI systems [29, 38, 40, 65, 68, 82, 93, 94]. The goal is to provide a detailed understanding of potential outcomes following spinal surgery. Many spinal surgeries involve relatively significant trauma, and the majority of patients undergoing spinal surgery are elderly. Therefore, predicting surgical outcomes in advance, assessing risks and benefits associated with the surgery, and performing surgery only on patients for whom the benefits outweigh the risks are critical. This approach enhances treatment outcomes and paves the way for wiser and more precise medical interventions. The recovery period after spinal surgery is inherently lengthy, involving a series of complex processes, including rehabilitation, physical therapy, and meticulous follow-ups. Predicting outcomes in advance is also beneficial in preparing healthcare professionals, patients, and their families for postoperative care. This is crucial for a patient's recovery, ensuring a smooth recovery, and overall effectiveness of the surgical intervention. Artificial intelligence and machine learning technologies have demonstrated value in predicting adverse events such as surgical site infections, pneumonia, and more. Some studies suggest that their predictive capabilities surpass those of traditional models. These tools can improve the acquisition and usability of prognostic information in clinical practice. However, their effectiveness and practicality still require further validation. Ultimately, these tools will aid in facilitating decision-making and managing patient expectations.

When a patient is designated as "non-home discharge" upon leaving a medical facility, there is a significant deviation from the traditional post-discharge trajectory. These patients are directed to alternative care facilities or institutions based on their unique medical needs and individual circumstances, rather than returning to their familiar homes. Older patients undergoing spinal fusion surgery are more likely to be discharged to a rehabilitation center. This targeted transition to a non-home environment indeed carries significant financial implications, leading to a notable increase in healthcare insurance expenditures. On the other hand, compared to patients discharged directly home, those sent to non-home environments experience higher rates of postoperative complications, readmissions, and other adverse events. Some forward-thinking researchers are striving to use AI methods to predict postoperative non-home discharge. These AI-driven predictive models are documented in various studies [18, 22,23,24, 41, 75, 87, 95, 101], aiming to streamline the decision-making process regarding patient discharge. The primary objective of this emerging research is to enable healthcare professionals to accurately predict potential discharge trajectories. By doing so, they can optimize the postoperative transition for each patient, ultimately reducing complications and improving healthcare outcomes. Additionally, by optimizing the allocation of patients to appropriate care settings, these AI-driven predictions can help manage healthcare resources wisely, potentially reducing medical costs without compromising the quality of care. Integrating artificial intelligence into healthcare systems holds the promise of fundamentally transforming patient care and shaping more efficient healthcare approaches for those in need. Researchers employ rigorous methodologies, and the algorithms used demonstrate satisfactory classification performance in predicting patient post-discharge outcomes. However, due to factors such as the patient samples possibly not being fully representative of all patients, and the data not including all specific factors that may be related to the risk of adverse postoperative outcomes, these technologies can only serve as auxiliary tools and cannot completely replace clinical judgment.

The financial aspects of surgery and hospitalization have always been a significant concern for patients. To address this issue, some researchers are focusing on accurately predicting the overall hospitalization costs for surgical patients [31, 75, 87]. This predictive model holds great significance for both patients and healthcare institutions. For healthcare institutions, accurately predicting the overall cost of patients is crucial. It aids in proactive financial planning, budgeting, and effective resource allocation. With precise cost predictions, healthcare organizations can design cost structures in advance, ensuring the sustainability and affordability of healthcare services without compromising quality. Equally important is the impact on patients. Having a clear understanding of the expected costs associated with surgery and its subsequent treatments enables patients to make informed medical decisions. Transparent cost information bridges the knowledge gap, allowing patients to comprehend the financial implications and plan accordingly. It creates a collaborative environment for effective communication and shared decision-making between patients and healthcare professionals. Ultimately, this informed and collaborative approach enhances the efficiency and effectiveness of healthcare resource allocation, aligns patient expectations with the reality of healthcare costs, and fosters a more equitable and patient-centered healthcare system. In the realm of predicting hospitalization and surgical costs, AI and ML technologies not only play a pivotal role in forecasting but also in identifying additional variables that are crucial in determining the cost of care. This further enhances cost efficiency and alleviates financial burdens.

Opioids are a class of potent analgesic medications that work by binding to opioid receptors in the brain, effectively blocking pain signals and relieving moderate to severe pain. While the use of opioids in pain management is highly effective, the inherent risks of addiction, tolerance, and potential side effects necessitate careful monitoring and regulation. Many studies delve into the prescription patterns of opioids after spinal surgery, aiming to investigate the continued use of opioids post-surgery and identify influencing factors [49, 86, 92, 97]. These studies consider key factors such as preoperative opioid use, antidepressant usage, tobacco use, insurance status, and preoperative medication therapies. The ultimate goal is to develop a stratified predictive algorithm for preoperative patients, designed to minimize long-term opioid use through informed decisions and personalized care. Despite advancements in predictive analytics, the application of these medications still necessitates stringent human oversight to ensure safe and effective use.

In the field of spinal care, some literature is dedicated to developing postoperative survival and death prediction models, specifically tailored for patients with spinal metastases or spinal tumors [28, 36, 99, 103, 105]. These predictive models play a crucial role in providing information for clinical decisions regarding surgical interventions and appropriate surgical strategies for patients in this unique medical condition. Decisions regarding surgical interventions for spinal tumors are often based on the estimated life expectancy of the patients, with the overall goal being to maintain or improve their quality of life. The development and implementation of postoperative survival and death prediction models have been instrumental in this regard. These models offer valuable insights into the expected postoperative survival time and the occurrence of potential adverse outcomes for clinical practitioners and healthcare providers. These predictive models fundamentally enhance end-of-life care for patients with spinal metastases or tumors. By anticipating survival trajectories and the likelihood of postoperative adverse events, healthcare professionals can adjust treatment plans accordingly, optimizing the quality of care provided to patients in their remaining time. Moreover, these predictive models enable patients and their families to make informed decisions regarding treatment choices, end-of-life planning, and overall care strategies, forming an approach that is patient-centered and prioritizes individual well-being and comfort. Integrating these predictive models into clinical practice will significantly improve the overall management and outcomes of patients facing challenging spinal health conditions. It is worth mentioning that some researchers have focused on the validation and evaluation of the SORG (Skeletal Oncology Research Group) machine learning algorithm, which is used to predict the 90-day and 1-year mortality of patients with spinal metastatic disease [49, 66, 70, 81, 88]. They conducted validation on an independent population to assess the accuracy and practicality of these algorithms in predicting survival outcomes, especially in the context of spinal metastatic disease. They believe that the SORG machine learning algorithm has the potential to be a valuable tool for predicting survival in spinal metastatic diseases. This algorithm has shown good predictive capability in assessing the 1-year survival rate of patients with spinal metastases and has undergone external validation, confirming its effectiveness in practical applications. This has led to its widespread use in spinal tumor research.

Researchers believe that applying AI technology to spinal surgery can enhance the safety, precision, and efficiency of the surgical process. However, the use of AI in assisting spinal surgery requires caution. Its limitations and risks must be thoroughly considered, and the professional judgment of the surgeon should always remain at the core of decision-making.

Firstly, the quality and representativeness of the data used to train AI models are critical. The performance of these models heavily depends on the diversity and accuracy of the training data. For instance, in tumor diagnosis, if the training data predominantly comes from a specific race or region, the AI's diagnostic accuracy may decrease when applied to other populations. This highlights the risk of bias and reduced efficacy in certain demographic groups or rare cases.

Another challenge is the generalization capability of AI models. These models may struggle to handle scenarios not present in the training data. In complex or rare cases, such as congenital heart anomalies in cardiac surgery, AI's performance may fall short of expectations. Additionally, the interpretability of AI decisions remains a significant hurdle. Many high-performing AI models, like deep learning algorithms, operate as "black boxes," making it difficult to explain their decision-making processes. This lack of transparency can undermine trust among doctors and patients and pose legal and ethical concerns. For example, in surgical planning, if an AI cannot justify its recommended surgical path, it may be challenging for surgeons to adopt its suggestions.

Real-time performance is another critical limitation. Some sophisticated AI models require substantial computational resources, making real-time processing challenging. In fast-paced surgical environments, this can restrict AI's applicability. For instance, during liver tumor surgery, if the AI cannot update resection plans in real-time, it may hinder surgical flexibility. Additionally, AI faces difficulties in handling unstructured data, such as surgical notes and medical records. This limitation can prevent AI from fully leveraging all available information, potentially impacting its ability to predict postoperative complications accurately.

Adaptability to dynamic environments is another concern. The surgical setting is inherently dynamic, and AI may struggle to adapt to unforeseen situations. In complex surgeries, unexpected events like sudden bleeding may require immediate adjustments that AI might not be able to provide. Ethical and legal issues also pose significant challenges. The use of AI involves complex questions around accountability, informed consent, and patient autonomy. These concerns can impede the widespread adoption of AI in high-risk areas. For example, if an AI-recommended treatment leads to adverse outcomes, determining responsibility can be problematic.

Finally, the continuous updating and maintenance of AI systems are essential but challenging. Medical knowledge and practices evolve rapidly, necessitating regular updates to AI systems to maintain their relevance and accuracy. This requires ongoing investment and expertise, which may be challenging for some healthcare institutions. If AI systems are not updated promptly to incorporate new diagnostic standards or treatment methods, they risk providing outdated recommendations.

Although this review provides a comprehensive overview of the advancements in AI applications in spinal surgery, it is important to acknowledge certain inherent limitations in our study. Our review exclusively includes literature published in English. This choice was made considering the language proficiency of our research team and the fact that most high-impact studies tend to be published in English-language journals. However, we recognize that this may introduce language bias, potentially excluding valuable research published in other languages. Our literature search was confined to PubMed. PubMed was chosen due to its extensive coverage of biomedical literature and its accessibility. However, we acknowledge that this focus might have led to the omission of significant studies, particularly those published in engineering or computer science journals, which are also relevant to the development and application of AI in spinal surgery. Additionally, one notable weakness in the current body of literature is the lack of high-quality neural networks. Although there have been significant advancements, many studies still rely on neural networks that may not be sufficiently robust or generalizable. High-quality neural networks require extensive, diverse datasets and significant computational resources, which are often limited. This limitation affects the reliability and applicability of AI models in clinical practice and should be addressed in future research.

In conclusion, while AI serves as a powerful supplemental tool aimed at enhancing the efficiency and accuracy of surgeons, it is unlikely to replace human surgeons entirely in the foreseeable future. The complex, dynamic nature of surgical procedures and the need for human intuition, real-time judgment, and ethical decision-making underscore the indispensable role of surgeons. Future research and publications should continue to explore these limitations and develop strategies to mitigate them, ensuring that AI technology can be safely and effectively integrated into spinal surgery and other medical fields. Specifically, more research is needed to address current limitations such as data quality, model interpretability, and integration into clinical workflows. Improving the diversity and representativeness of training data can help mitigate biases and enhance AI performance across different patient populations. Enhancing the interpretability of AI models is crucial for gaining the trust of both surgeons and patients, which can be achieved through the development of more transparent algorithms and explainable AI techniques. Integrating AI systems seamlessly into clinical workflows requires collaboration between AI developers and healthcare professionals to create user-friendly interfaces and ensure that AI recommendations can be easily understood and acted upon in real-time. Additionally, future research should focus on variable analysis for approach selection in spinal surgery, which involves identifying the key factors that influence surgical decisions and outcomes. By analyzing these variables, AI can provide more tailored and context-specific recommendations, ultimately improving patient care. Addressing these challenges and advancing AI research will pave the way for its broader adoption and greater impact on clinical practice in spinal surgery and beyond.

Conclusion

In recent years, AI technology has been widely applied in the field of spinal surgery, making significant contributions to the diagnosis, treatment, care of patients, and the development of surgical techniques. Through clever algorithms, doctors should be able to predict various postoperative outcomes before surgery. AI at the current stage is primarily a powerful supplemental tool aimed at enhancing the efficiency and accuracy of surgeons.AI technology can provide significant support in data analysis, surgical planning, intraoperative navigation, and postoperative monitoring, thereby improving patient outcomes and reducing surgical risks. However, we do not believe that AI will completely replace human surgeons in the foreseeable future. Surgical procedures involve many complex and dynamically changing situations, with many decisions relying on the surgeon's experience, intuition, and real-time judgment—factors that current AI technology is challenging to fully replicate. Surgeons are not merely technical executors; they also need to communicate with patients and their families, provide emotional support, and make ethical decisions, aspects that AI cannot replace. In the future, with the continued development of AI technology, more research and publications on its applications in spinal surgery are expected, and further studies will be necessary to explore and advance the field.