人工智能干预对腹腔镜胃切除术解剖标志认知的影响
本研究开发了利用资深外科医生隐性知识可视化胰腺及凹陷线等解剖标志的胃癌/胃切除术人工智能(gAI),并探索其对外科医生解剖标志认知的影响。
要点
- gAI可可视化胰腺和凹陷线等解剖标志,辅助预防腹腔镜胃切除术中的胰瘘。
- gAI干预对解剖标志认知的影响因外科医生资历不同而异,新手医生变化率更高。
- 在部分探索性分析中,gAI使新手医生对某些解剖标志的识别评分显著提高。
核心结论
- gAI干预可能支持腹腔镜胃切除术中解剖标志识别的意识,且效果因资历水平而异,尤其对新手外科医生更明显。
主要发现
- 胃下区主观识别变化率:初级受训者52%,高级受训者38%,培训师9%;培训师显著低于两组受训者(p < 0.001)。
- 胰上区主观识别变化率:初级受训者38%,高级受训者17%,培训师4%;培训师显著低于初级受训者(p < 0.001),但与高级受训者差异无统计学意义(p = 0.054)。
- 在仅限于有评分变化的观察的探索性分析中,胰上区初级受训者(p < 0.01)和胃下区高级受训者(p = 0.03)对肠系膜与胰腺之间凹陷线的识别评分显著提高。
临床怎么用
- gAI可作为辅助工具帮助外科医生,尤其是新手,在腹腔镜胃切除术中更好地识别解剖标志,但需注意其效果因资历而异。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: We have developed gastric cancer/gastrectomy artificial intelligence (gAI) that leverages the tacit knowledge of experienced surgeons to visualize anatomical landmarks, including the pancreas and dimpling lines (DLs), to help prevent pancreatic fistula in laparoscopic gastrectomy for gastric cancer. However, whether gAI contributes to the awareness among surgeons remains unclear. Therefore, in the present study, we aimed to evaluate how gAI intervention influences anatomical landmark recognition, which reflects the awareness of these landmarks among surgeons, and how its effects vary by surgical expertise. METHODS: An external evaluation committee established the consensus reference standard and validated gAI. Twelve surgeons from four institutions were grouped by expertise as follows: junior trainee, senior trainee, and trainer. Each surgeon viewed videos of the infrapyloric and suprapancreatic areas without and with gAI, annotating images from each scene. Subjective recognition change rate and recognition score difference in anatomical landmark recognition compared to reference standards were assessed across expertise levels. RESULTS: In the infrapyloric area, subjective recognition change rates were 52% for junior trainees, 38% for senior trainees, and 9% for trainers, with trainers exhibiting significantly lower rates than both trainee groups (p < 0.001 for both). In the suprapancreatic area, the corresponding rates were 38, 17, and 4%, respectively; trainers exhibited significantly lower rates than junior trainees (p < 0.001) but not senior trainees (p = 0.054). In exploratory analyses limited to observations with measured score changes, recognition scores for the DL between the mesogastrium and pancreas increased significantly in junior trainees in the suprapancreatic area (p < 0.01) and senior trainees in the infrapyloric area (p = 0.03). CONCLUSIONS: This exploratory study suggests that gAI intervention could support awareness of anatomical landmark recognition in laparoscopic gastrectomy with differential effects across expertise levels, particularly among novice surgeons.
引用
- Hasegawa T, Etoh T, Nakano T, et al. Impact of artificial intelligence intervention on awareness of anatomical landmarks in laparoscopic gastrectomy: An exploratory study. Surg Endosc. 2026. PMID: 42387009. doi:10.1007/s00464-026-13043-6
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。