Abstract 116P
Background
The preoperative localization of pulmonary nodules that extend 2cm beyond the surface of the pleura or are obstructed by thoracic bones is limited. There is an urgent need for an effective auxiliary method to locate nodules that difficult to reach. We attempted to identify the supply area of pulmonary nodule drainage vessels through preoperative simulation staining model, in order to accurately locate the lesion location in virtual reality images.
Methods
We used the Vnet framework to oversample thin-layer CT to restore the distribution of pulmonary arteries, pulmonary veins, and capillary networks, and used morphological methods to convert annotated data into graded vascular data to separate cerebral vessels of different sizes. Predicting watershed boundaries by simulating fluid perfusion, performing nodule localization and preoperative planning based on the boundaries, and assisting navigation during surgery through virtual reality. During the surgery, we injected indocyanine green from the peripheral vein after blocking the target blood vessels and stained the target area where the nodules were located. And by comparing the corners in actual surgery, the accuracy of simulating watershed boundaries was verified. Then we calculate the impact of this auxiliary localization method on the surgical time and precise resection of pulmonary nodules.
Results
A total of 231 patients underwent lung wedge resection and participated in preoperative and intraoperative target vessel watershed comparisons. 220 (95.24%) watershed boundary showed consistent comparisons between simulated and Image under the endoscope. The average tumor diameter is 11.3 (8-18.6) mm, average depth of nodules is 29.4(18.2-38.4) mm, and the average operation time is 92(44-132) minutes.
Conclusions
AI simulated dyeing model can reliably simulate the indocyanine green staining area after pulmonary vascular occlusion during surgery, so as to achieve intraoperative three-dimensional positioning of nodules, give visual guidance to surgeons, speed up and improve the safety in surgery. It is an effective supplement to traditional localization methods for dealing with difficult nodules.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.