Comprehensive Summary
Chen et al explores the creation of an artificial intelligence laparoscopic exploration system (AiLES) to recognize intra-abdominal metastasis (IAM) lesions for gastric cancer. AiLES was developed based on a dataset of 5111 frames based on varying severities and locations of metastasis. The AiLES received the best DICE score out of all the other models of 0.76 (±0.17). for IAM recognition. It also has the fastest real-time interference speed of 11 frames per second. For lesions with different extents of metastasis, AiLES got good segmentation performance with Dice scores of 0.76 (±0.18), 0.74 (±0.16), and 0.76 (±0.14) for single, multiple, and extensive metastases. It was also successful in looking at lesions in different locations and during laparoscopic surgery. AiLES was consistent with novice surgeons, sometimes detecting tiny and occult IAM lesions novice surgeons could not.
Outcomes and Implications
AiLES’ success is important in being used for intraoperative diagnosis and treatment making, especially in patients with gastric cancer presenting with IAM lesions. Gastric cancer is the fifth most common malignant tumor and causes many cancer-related deaths because CT scans are not able to capture accurate tumor staging and metastasis. However, with the help of AiLES during laparoscopic exploration, it can identify and evaluate lesions pre-surgery. To implement this into the clinic, there is work that needs to be done on forming an appropriate annotation approach for a high-quality dataset. In gastric cancer patients, peritoneal metastasis has predominant IAM’s that are typically missed, so hopefully refining the AiLES will help identify tiny lesions missed during the exploration. With the AiLES being at a comparable level to novice surgeons, there is potential for it to be a valuable tool during surgical training to help them work on their IAM lesion identification skills. This can help decrease unnecessary gastrectomies and reduce healthcare costs. AiLES can also be expanded to analyze images from other abdominal surgeries and do more multicenter validation to be implemented in the clinic.