suchen
Generic filters

Artificial Intelligence System Outperforms Guideline Predictive Criteria for Lymph Node Metastasis in T1 Cancers

Douglas K. Rex, MD, MASGE, reviewing Kudo SE, et al. Gastroenterology 2020 Sep 23.

In a study from 7 centers in Japan, a deep learning artificial intelligence (AI) program trained on 3134 T1 colorectal cancers (CRCs) from 6 Japanese hospitals was tested in 939 T1 CRCs from a seventh hospital as the validation cohort. The model learned from 8 factors: age, sex, tumor size, location, morphology, lymphatic invasion, vascular invasion, and histologic grade.

Compared to U.S. guideline predictors for lymph node metastasis (histologic grade and lymphovascular invasion), the model in the training cohort had an area under the curve of 0.78 versus 0.67 for the guideline. In the validation cohort, the AUC was 0.83 with the model compared to 0.73 with the guideline (P<.001). The model was also superior in the group undergoing endoscopic resection only.

Similarly, the model outperformed the Japanese guidelines, which include factors such as tumor budding and submucosal invasion depth.

Douglas K. Rex, MD, FASGE

COMMENT

These data are encouraging, suggesting that AI can lead to better selection of patients for adjuvant surgical therapy after endoscopic resection of T1 cancers. U.S. guidelines also include factors such as involvement of or proximity to the resection line and the use of piecemeal resection for nonpedunculated lesions, which remains common in the U.S. Therefore, validation of the U.S. cohort will be important, as will the value placed by patients and clinicians on sensitivity versus specificity of any predictive tool.

Note to readers: At the time we reviewed this paper, its publisher noted that it was not in final form and that subsequent changes might be made.

CITATION(S)

Kudo SE, Ichimasa K, Villard B, et al. Artificial intelligence system to determine risk of T1 colorectal cancer metastasis to lymph node. Gastroenterology 2020 Sep 23. (Epub ahead of print) (https://doi.org/10.1053/j.gastro.2020.09.027)

Ähnliche Beiträge

Scroll to Top