Artificial Intelligence Assisted Topographic Mapping System for Endoscopic Submucosal Dissection Specimens

BACKGROUND: Endoscopic submucosal dissection (ESD), a minimally invasive surgery used to treat early gastrointestinal malignancies, has been widely embraced around the world. The gross reconstruction of ESD specimens can facilitate a more precise pathological diagnosis and allow endoscopists to explore lesions thoroughly. The traditional method of mapping is time-consuming and inaccurate. We aim to design a topographic mapping system via artificial intelligence to perform the job automatically. METHODS: The topographic mapping system was built using computer vision techniques. We enrolled 23 ESD cases at the Peking Union Medical College Hospital from September to November 2019. The reconstruction maps were created for each case using both the traditional approach and the system. RESULTS: Using the system, the time saved per case ranges from 34 to 3,336 s. Two approaches revealed no significant variations in the shape, size, or tumor area. CONCLUSION: We developed an AI-assisted system that would help pathologists complete the ESD topographic mapping process rapidly and accurately.