Abstract
OBJECTIVE: This study introduces PulmoSimulatedReality (Pulmo-SR), a novel technique combining artificial intelligence, finite element method, 3-dimensional (3D) visualization, and 4-dimensional (4D) interaction for preoperative imaging and intraoperative surgical guidance in pulmonary resections, such as lobectomy and segmentectomy. The clinical applicability of this 3D modeling approach is evaluated through a preliminary validation protocol. METHODS: A deep learning algorithm was employed to generate 3D segmentations of patient anatomy. 3D models were created for 30 patients undergoing pulmonary resection, and 4D models were developed using the Pulmo-SR platform, incorporating finite element methods for dynamic deformation. Clinical validation was conducted by assessing accuracy, precision, and sensitivity using retrospective intraoperative video recordings alongside dynamic 4D models. Latency and 3D model reconstruction time were also measured. RESULTS: Validation of 30 cases yielded high average scores for accuracy, precision, and sensitivity, respectively: artery (0.987 ± 0.047, 0.993 ± 0.037, and 0.994 ± 0.031), vein (0.976 ± 0.099, 0.976 ± 0.099, and 1.00 ± 0.00), and bronchus (1.00 ± 0.00, 1.00 ± 0.00, and 1.00 ± 0.00). Latency was 0.23 ± 0.06 seconds, and 4D model reconstruction was completed in 8.47 seconds. CONCLUSIONS: Pulmo-SR integrates artificial intelligence, finite element method, and 3D modeling to provide a 4D deformable reconstruction of patient anatomy, offering realistic simulations for complex lung resections. Clinical validation demonstrated high accuracy, precision, and sensitivity, indicating the potential as a valuable tool in preoperative and intraoperative workflows for anatomical lung resections.