Robotic surgery began in the 1980s with telemanipulation systems that allowed surgeons to operate through remotely controlled mechanical arms. The landmark FDA approval of the da Vinci Surgical System in 2000 marked a turning point, establishing robotic surgery as a clinical reality and setting the stage for its integration with artificial intelligence.
Compared to traditional open or laparoscopic surgery, robotic systems offer enhanced dexterity, improved 3D visualization, and tremor reduction. These features translate into smaller incisions, less blood loss, faster recovery times, and reduced postoperative pain for patients -- particularly in prostate, colorectal, and cardiac procedures.
The next frontier is the integration of artificial intelligence (AI) directly into robotic surgical platforms. Early AI applications focused on automating discrete tasks like suturing or tissue dissection. Modern systems are moving toward real-time image analysis, autonomous motion planning, and haptic (touch) feedback augmentation.
This review synthesizes evidence from 103 peer-reviewed publications to map the current landscape of AI-driven robotic surgery, covering applications across urology, cardiology, neurosurgery, orthopedics, and gastrointestinal surgery.
A widely used framework classifies surgical robot autonomy on a scale from Level 0 to Level 5. At Level 0 (exemplified by the standard da Vinci system), the surgeon controls every robotic movement directly with no AI assistance. At Level 1, the robot provides virtual fixtures -- soft guardrails that guide the surgeon's movements and prevent unsafe actions.
Level 2 robots can execute specific pre-defined tasks autonomously under physician oversight -- for example, controlling a colonoscopy scope's tip retroflection. Level 3 robots perceive their environment and plan task execution independently, such as flexible endoscopes that navigate the colon without manual guidance.
At Level 4, robots interpret pre-operative and intraoperative data to create intervention plans and adapt in real-time. A potential Level 4 application in cancer surgery would involve intelligent tissue removal that minimizes damage to healthy tissue while targeting tumor margins -- an area of active research.
Level 5 (fully autonomous surgery without any human presence) remains aspirational and has not yet been achieved. Current efforts are focused on advancing systems through Levels 2 to 4, with increasing AI capability at each stage.
The Smart Tissue Autonomous Robot (STAR) is a Level 3 system designed for bowel anastomosis (reconnecting cut ends of intestine). STAR has demonstrated the ability to match or surpass human surgeons in suture consistency and error reduction. It assesses tissue thickness to devise a suture plan, executes autonomously after surgeon approval, and continuously adapts to tissue deformation during the procedure.
The TSolution One system performs autonomous bone carving for hip and knee replacement according to a pre-established plan, improving implant positioning precision. The CyberKnife robotic radiosurgery system delivers stereotactic radiation to brain and spine tumors while automatically adjusting in real-time for even minor patient position changes, ensuring millimeter-level accuracy.
In urology, Aquablation is an FDA-approved semi-autonomous system that uses robotic waterjet technology under real-time ultrasound guidance to treat prostate conditions. In vascular surgery, augmented reality systems like those used in kidney transplantation can overlay 3D CT-derived vascular maps onto the surgical field, helping surgeons navigate complex anatomy.
These systems illustrate that AI in surgery is not a distant concept -- selective autonomy is already in clinical use, with capabilities continuing to expand across specialties.
Robotic-assisted radical prostatectomy (RARP) is one of the most common applications of surgical robotics. In prostate cancer, nerve-sparing during surgery is critical for preserving sexual and urinary function post-operatively. AI is being developed to improve intraoperative identification of the neurovascular bundles that surround the prostate, reducing the risk of nerve damage.
A major challenge in RARP is the absence of haptic feedback -- the robotic system does not transmit the sense of touch to the surgeon. This can lead to applying too much force (damaging nerves) or too little (poor suture retention). To address this, systems like the one developed by Dai et al. incorporate biaxial shear detection to provide vibrotactile feedback that warns surgeons before suture tension reaches its breaking point.
Computer vision (CV) algorithms are being applied to CT imaging data for accurate kidney stone localization and in laparoscopic video analysis for real-time identification of urological anatomy. Machine learning is also being used for patient selection, predicting who will benefit most from robotic surgery, and for automated tracking of surgical anatomy during procedures.
The field of urology has also seen an increase in automated biopsy sampling systems and ML-based analysis of perioperative imaging data to improve diagnostic precision and surgical planning before a single incision is made.
The benefits of AI integration in robotic surgery are substantial. Enhanced precision through motion prediction and instrument path optimization allows procedures to be more accurate and consistent. Reduced surgeon fatigue from automated repetitive tasks means the surgeon can focus mental energy on critical decision points. Real-time AI monitoring can alert the surgical team to potential complications like bleeding events or instrument collisions before they become harmful.
However, significant limitations exist. High development and implementation costs limit access, particularly at smaller hospitals and in low- and middle-income countries. A da Vinci system itself costs over a million dollars, and AI-integrated versions add further expense for hardware, software maintenance, and staff training.
AI effectiveness depends critically on training data quality. If the data used to train a system is biased (for example, derived mainly from one patient population or procedure type), the AI may perform poorly or unfairly for patients who differ from that training set. This raises concerns about healthcare disparities if AI tools are validated only on certain demographics.
Ethical and legal questions about AI autonomy in surgery remain unresolved. When an AI-assisted robot makes a decision that leads to patient harm, who is responsible -- the surgeon, the hospital, or the AI developer? Clear regulatory frameworks are still being developed to address this question.
Telesurgery -- where a surgeon operates on a patient from a remote location via a robotic system -- is a promising application that could bring specialized surgical expertise to underserved geographic areas. AI would play a central role in compensating for communication latency and ensuring safe remote operation.
Personalized surgical planning using AI could one day analyze a patient's anatomy, medical history, and genetic profile to generate customized surgical approaches. Rather than using standardized techniques, surgeons would follow AI-recommended steps tailored to the individual -- improving outcomes and reducing complications.
Micro-robotics -- millimeter-sized robots guided by external magnets -- are being explored for targeted drug delivery and minimally invasive procedures in previously inaccessible anatomical locations. Research in this area focuses on miniaturization, propulsion mechanisms, imaging integration, and remote manipulation.
AI-powered surgical training simulations integrated with virtual reality could allow surgeons to practice complex procedures in realistic digital environments, get feedback on their technique, and track skill development over time -- potentially reducing the learning curve for adopting new robotic procedures.
The review concludes that AI integration into robotic surgery holds transformative potential but requires careful, evidence-based implementation. Despite numerous promising examples, the authors emphasize that no surgical AI system can yet independently recognize the critical decision points that determine patient outcomes -- human oversight remains essential.
Validation studies on large, diverse datasets and external institutional validation are necessary before AI algorithms can be trusted for clinical decision-making. The growing autonomy of robotic systems has the potential to standardize surgical outcomes, reducing dependence on individual surgeon skill variation -- but this benefit must be balanced against the risks of over-reliance on technology.
Regulatory bodies, surgical societies, and technology developers need to collaborate on clear governance frameworks that define acceptable levels of AI autonomy, establish testing standards, address liability, and ensure cybersecurity of AI-integrated surgical platforms. Responsible adoption, rather than rapid uncritical deployment, is the path to realizing the full potential of AI-driven surgical care.