Platelets play a crucial role in the maintenance of normal blood clotting and the formation of blood clots. The process of platelet aggregation is a significant factor in the development of various arterial blockages such as coronary artery disease, strokes, and peripheral arterial disease. Consequently, the use of medications that inhibit platelet function, known as antiplatelet therapy, is essential in preventing recurring events in individuals affected by these conditions [1,2,3,4].

There is a lack of comprehensive guidelines regarding the use of antiplatelet agents in elective neurosurgery. The decision to continue antiplatelet therapy during surgery presents a risk due to increased bleeding, while discontinuing the medication prior to surgery carries the risk of thrombosis. In general, the discontinuation of antiplatelet therapy is typically the recommended course of action in neurosurgical settings [5], but this carries an elevated risk of ischemic events, including potentially life-threatening stent thrombosis, myocardial infarction, and stroke. Conversely, maintaining antithrombotic therapy may heighten the chances of bleeding and the need for blood transfusions, both of which are recognized factors that contribute to a poor prognosis [6]. It is possible that these difficult decisions regarding when to use antiplatelet therapy could be determined by artificial intelligence (AI) through utilizing technology to monitor individual antiplatelet compliance and platelet function.

Artificial intelligence (AI) is increasingly shaping medical decision-making, with notable applications in neurosurgery and antiplatelet therapy. In neurosurgery, AI aids in assessing the delicate balance between ischemia and bleeding risks post-stent placement or -WEB embolization. Although most AI research has historically focused on cardiovascular medicine, recent studies have developed AI models for predicting ischemic and bleeding risks in patients undergoing drug-eluting stent implantation. Additionally, AI can enhance medication adherence through smartphone applications, ensuring that patients follow their prescribed regimens, including antiplatelet therapies. The integration of AI in medical decision-making holds promise for optimizing individualized treatment plans in the complex realm of neuroendovascular procedures.

It is crucial to assess the balance between the risks of ischemia and bleeding after stent placement or WEB embolization in neurosurgery and to determine the optimal duration of DAPT accordingly. With varying degrees of adherence to their medication regimen and different levels of platelet inhibition within individuals, it may be possible to use artificial intelligence (AI) to make important decisions regarding when to use antiplatelet therapy. Currently, most research that exists that has studied the use of AI in the prediction of ischemic/bleeding risk exists in the field of cardiovascular medicine, but many of these AI models can potentially be applied to the field of neurosurgery. AI’s predictive capabilities can significantly impact the management of antiplatelet therapy in neurosurgery. The following is a summary of the various ways in which AI can impact the management of antiplatelet therapy in neurosurgery.

Predicting Ischemic and Bleeding Risks in Neurosurgery: AI models have been successfully used in cardiovascular medicine to predict ischemic and bleeding risks following procedures like drug-eluting stent (DES) implantation [73,74,75,76,77,78,79]. These models take into account various patient factors, including age, diagnoses, medications, procedures, and DAPT continuation status. AI can adapt and extend these models to neurosurgery settings, helping to predict the risks of ischemic events and bleeding after stent placement or WEB embolization in neuroendovascular procedures. These predictions can guide decisions regarding the timing and duration of antiplatelet therapy.

Superior Performance of AI Models: Studies have shown that AI models can outperform traditional clinical tools like the DAPT score in predicting outcomes. For example, AI models demonstrated a higher area under the receiver operating characteristic (AUC) values, a performance metric that is used to evaluate classification models, for predicting ischemia and bleeding compared to the DAPT score [73]. In addition, these models accounted for scenarios that were not addressed by the DAPT score. This improved accuracy can aid neurosurgeons in making more informed decisions about antiplatelet therapy, especially in the 12–30-month window following stent placement.

Personalized Treatment Plans: AI models can help create personalized treatment plans based on individual patient profiles and risks. For example, in one study, a machine learning tool known as the PRAISE (Prediction of Adverse Events Following an Acute Coronary Syndrome) predicted death, myocardial infarctions, and major bleeding following an acute coronary syndrome using 25 clinical features that are incorporated during discharge. Based on this information, patients can be categorized into different risk groups (low, intermediate, high), and those at a higher risk can receive closer monitoring and potentially shorter durations of DAPT. This personalized approach enhances patient care and minimizes the risk of complications [80,81].

AI for Subarachnoid Hemorrhage (SAH), Unruptured Intracranial Aneurysms (UIA), and Chronic Subdural Hematoma Outcomes: AI has also already been applied to the field of neurosurgery to predict outcomes in conditions like SAH. These predictive models can assess factors such as neurological severity, age, aneurysm location, and size to forecast outcomes [82,83,84,85,86,87,88,89,90,91,92,93,94] with AUC values ranging from 0.70 to 0.90. More recently, deep learning has been employed, which has resulted in an improved prediction accuracy of 0.90 with limited datasets for SAH outcomes [90,93,94]. Two widely recognized radiological scales, namely the Fisher computed tomography (CT) scale [95] and the modified Fisher scale [96], assess the extent of bleeding to forecast delayed cerebral ischemia (DCI) incidence. Additionally, several statistically derived scores [97,98,99] incorporating supplementary factors have been investigated, yielding an average AUC of approximately 0.70 for predicting DCI occurrence [100]. As for AI-based prediction models, they have demonstrated AUC values of around 0.80 [93,101,102,103,104,105] in forecasting DCI occurrence. If we were able to anticipate DCI, we could administer proactive and immediate treatment. Other studies have investigated the potential use of machine learning to predict clinical outcomes in the microsurgical treatment of unruptured intracranial aneurysms (UIA) [106,107,108]. One study offered personalized predictions at the patient level, estimating outcomes such as neurological recovery upon discharge, as well as the likelihood of complications and new neurological issues upon discharge. These predictions are based on readily available preoperative data. The study employed various scoring systems to estimate the risk of aneurysm rupture (PHASES) or growth (ELAPSS), or to directly assess the balance between potential risks and benefits (UIATS) [109]. Methodologies from these studies could be used to anticipate potential complications following UIAs to adjust antiplatelet therapies to respond to them. Additionally, ML models have been used to predict the recurrence risk of chronic subdural hematoma (cSDH) while withholding antiplatelet and anticoagulant agents. This study can also inform the adjustment of antiplatelet regimens according to ischemia/bleeding complication [110].

Medication Adherence Monitoring: AI can be used to monitor patient adherence to antiplatelet therapy. For instance, smartphone applications can use AI to confirm medication ingestion through the phone’s camera. Past studies using this technology demonstrated a 67% increase in absolute drug adherence compared to the control [111]. This technology can ensure that patients are adhering to their medication regimen, leading to better outcomes. Similar systems can be adapted for patients undergoing stent or WEB embolization placement, helping to ensure adherence to antiplatelet medications and facilitating communication among healthcare providers.

Individualized Treatment Plans with PRU and TEG Mapping: AI platforms can incorporate PRU and TEG mapping to provide individualized treatment plans [111]. This can help to promptly identify non-responders to antiplatelet therapy and adjust their medication regimens accordingly. Through the monitoring of labs, AI has the potential to identify those with suboptimal antiplatelet treatment.

Patient Engagement and Self-Efficacy: AI can be used to engage patients in their treatment plans through reminder systems, text messages, and other technologies, as other studies have done in the past [112]. This not only promotes medication adherence but also increases patient self-efficacy in managing their health.

Of course, the commonly discussed challenges associated with integrating AI into clinical practice remain and require attention before widespread clinical adoption [113]. One of these critical challenges revolves around patient privacy concerns, as AI relies on extensive data for algorithm training and sequencing [114]. Striking a balance between privacy protection and data accessibility is essential for sustaining long-term progress in neurosurgical AI [115]. Furthermore, ensuring the quality of data is paramount for meaningful results, emphasizing the importance of effective data implementation in machine learning training [115]. Another issue to consider is the risk of neurosurgeons becoming overly reliant on AI, potentially hindering their skill development, while hardware and software glitches pose the threat of incorrectly directing antiplatelet regimens if not addressed promptly [113,116]. Although artificial intelligence in healthcare has made significant strides, its potential for future advancements remains untapped. The existing innovations have already yielded benefits for patients, but it is crucial for regulatory frameworks to adapt to the rapidly evolving healthcare landscape to proactively address and mitigate potential risks.

In summary, AI’s predictive capabilities can revolutionize the management of antiplatelet therapy in neurosurgery by providing more accurate risk assessments and personalized treatment plans, as well as improved medication adherence. By adapting AI models from cardiovascular medicine and leveraging advanced technologies, neurosurgeons can optimize patient care and outcomes in neuroendovascular procedures (Table 1).