Abstract
BACKGROUND: We have developed an innovative artificial intelligence (AI)-assisted surgical simulation system to enhance surgical planning and navigation for thoracoscopic pulmonary segmentectomies using computed tomography (CT) images. Traditional preoperative planning methods are often time-consuming, labor-intensive, and lack the necessary precision, which can negatively impact surgical outcomes. Our system, in contrast, enables intelligent nodule analysis and precise preoperative planning, thereby improving intraoperative navigation accuracy and contributing to better postoperative recovery for patients. METHODS: The novel AI-assisted surgical simulation system (LungDimensionGo V1.0, SunKun, Beijing, China) adopted EfficientDet method for detecting lung nodules and used an image segmentation algorithm based on Mamba-Unet and SegRefiner to reconstruct lung three-dimensional (3D) model using CT images. We assessed the clinical value of this novel AI system by comparing it with traditional methods across preoperative, intraoperative, and postoperative phases. The study included data from retrospective (n=125) and prospective (n=38) cohorts of patients who underwent segmentectomy at our institution. RESULTS: Patient and tumor characteristics, as well as postoperative pathology, showed no significant differences between the two groups. However, the AI-assisted group exhibited several advantages over the traditional method group. These included shorter model reconstruction times, higher accuracy of anatomical structures, reduced operative times, less intraoperative blood loss, shorter postoperative chest tube durations, reduced postoperative hospital stays, shorter total hospital stays, and fewer postoperative complications. CONCLUSIONS: We found that the AI-assisted system significantly outperforms traditional methods in preoperative preparation, intraoperative guidance, and postoperative patient recovery.