introduction
In today's medical field, artificial intelligence (AI) is developing at an unprecedented pace and is showing its strong application potential in many fields. However, compared to other medical fields, the application of AI in surgery is still in its infancy, but its future development prospects are impressive. Surgery is a highly complex and experience-dependent field that encompasses preoperative diagnosis, clinical risk prediction, intraoperative decision support, and postoperative monitoring. With approximately 330 million surgeries underway worldwide each year, the potential of AI to optimize surgical processes and outcomes is emerging as the demand for surgeries continues to increase and waiting lists lengthen. On May 13, Nature Medicine published a review of progress in this field, "Artificial intelligence in surgery". Preoperative diagnosis is an important area of AI application, and AI can achieve high-precision detection of early diseases through deep learning of medical imaging data. For example, 3D Convolutional Neural Network (3D-CNN) has performed well in detecting pancreatic ductal adenocarcinoma, greatly improving the accuracy of early diagnosis. At the same time, the application of AI in clinical risk prediction and patient selection is becoming more and more extensive. Machine learning-based prediction models, such as the POTTER Calculator, perform well in surgical risk assessment, helping doctors to select indications more scientifically and optimize surgical planning. In intraoperative operations, the application of AI is mainly focused on real-time decision support and surgical automation. Through computer vision analysis of surgical videos, AI is able to identify surgical steps, assess the surgeon's skill level, and provide real-time feedback, significantly improving the accuracy and safety of surgery. In addition, AI tools such as the Hypotension Prediction Index (HPI) have shown significant benefits in clinical practice, which can predict and intervene in advance of intraoperative complications. Postoperative monitoring is another area where AI has a lot to offer. Through wearable devices and sensors, AI can realize continuous monitoring of patients' physiological parameters to predict and prevent postoperative complications in a timely manner. For example, the AI-based Prescience system is able to predict five minutes before a hypoxic event occurs and provides a real-time risk score, greatly improving the efficiency and safety of postoperative care. Overall, the application of AI in surgery is deepening, and it has shown great potential to improve surgical accuracy, optimize surgical processes, and improve patient outcomes. With the further development of AI technology, the application of multimodal AI methods will bring broader development prospects and more significant clinical benefits to surgery. In the future, AI is expected to play a more important role in all aspects of surgery and provide patients with better medical services.
The application of AI in preoperative preoperative diagnosisPreoperative diagnosis is an important part of the success of surgery. AI is particularly prominent in medical image recognition, especially in early disease diagnosis and surgical planning. For example, 3D Convolutional Neural Network (3D-CNN) can detect pancreatic ductal adenocarcinoma in preoperative magnetic resonance imaging (MRI) with extremely high accuracy (AUROC values of 0.97 and 0.90, respectively). This high-precision early diagnosis facilitates timely intervention, which can significantly improve the patient's prognosis.
The high-precision risk prediction model of clinical risk prediction and patient selection can help doctors better select surgical indications before surgery, reduce the ineffectiveness of surgery, and improve the quality of patients' informed consent. For example, the machine learning-based POTTER calculator achieved a prediction accuracy of 0.92 (internal validation) and 0.93 (external emergency surgery validation) in emergency surgery patients. This model improves the accuracy of risk prediction by comprehensively analyzing multiple data and provides strong support for clinical decision-making.
Preoperative OptimizationPreoperative optimization is a relatively new concept that aims to develop a personalized optimization plan through a comprehensive assessment of the patient's physical condition. For example, AI can optimize preoperative cardiovascular assessment by rapidly assessing cardiac function using a 12-lead electrocardiogram (ECG). In addition, the multimodal AI method can comprehensively analyze a variety of data, including genomics and microbiomics, to comprehensively evaluate patients for more accurate optimization before surgery.
人工智能驱动的数字干预措施在术中环境中的整合（Credit: Nature Medicine）
The application of AI in surgeryThe application of AI in intraoperative decision-making is mainly focused on real-time decision support. For example, the Intraoperative Hypotension Prediction Index (HPI) has shown significant clinical benefit in two randomized trials to predict hypotension in advance and intervene. By analyzing patients' physiological data in real time, AI can help doctors make more timely and accurate decisions during surgery and reduce the occurrence of intraoperative complications.
Computer vision and surgical automationComputer vision is another important area of AI application in surgery. By analyzing the surgical video, AI can identify surgical steps, assess the surgeon's skill level, and provide real-time feedback. For example, a unified surgical AI system was able to accurately identify and evaluate surgical steps such as needle extraction, handling, and actuation in external validation, showing high reliability (AUC value greater than 0.85). These techniques help to improve the precision and safety of the surgery.
围手术期和术后连续监测的传感器输入（Credit: Nature Medicine）
Application of AI in Postoperative MonitoringPostoperative monitoring is essential to improve the quality of patient rehabilitation. Currently, many hospitals still rely on nurse observation every four hours, which not only increases the workload of nursing staff, but can also lead to delayed response to patient conditions. Through wearable devices to achieve continuous monitoring of physiological parameters, AI can analyze the patient's recovery in real time, predict complications in advance, and intervene. For example, the explainability AI-based Prescience system is able to predict the risk of a hypoxic event five minutes before it occurs and provide a real-time risk score.
Complication prediction: Early detection of postoperative complications is essential to reduce postoperative mortality. For example, MySurgeryRisk, which uses machine learning algorithms to predict the risk of major postoperative complications and death, has shown good performance in single-center studies. With the development of sensors and wearable devices, the application of AI in complication prediction is promising.
The application of AI in patient education and informed consentLarge language models and large language models (LLMs) such as ChatGPT have demonstrated their potential in medical education and patient communication. These models can help physicians write informed consent forms for surgery and improve patients' understanding of the risks and processes of surgery. For example, informed consent documents generated by LLMs excel in readability, accuracy, and contextual awareness, even better than those written by surgeons.
Personalized patient consultation AI models can also be used to personalize patient consultations, providing customized risk assessment and post-operative recovery guidance by analyzing the patient's personal data. For example, an AI communication platform that integrates accurate deep learning risk prediction can provide patients with personalized risk assessments, answer questions about preoperative optimization and postoperative recovery, and guide patients through the entire surgical process.
With the continuous development of AI technology, its application in surgery has broad prospects. From preoperative diagnosis, clinical risk prediction, intraoperative decision support to postoperative monitoring, AI is playing an increasingly important role in all aspects. In the future, with the development of multimodal AI methods, the application of AI in surgery will be further expanded and deepened, bringing higher accuracy and safety to surgical procedures, and ultimately improving the prognosis and quality of life of patients.
Link to original article
https://www.nature.com/articles/s41591-024-02970-3
Editor-in-charge|Explore Jun
Typography |
Please indicate the source of reprinting【Biological Exploration】
End