| Strategic International Session (Symposium)3(JSGS・JSGE・JGES) |
| Sat. November 7th 9:00 - 11:00 Room 9: Portopia Hotel Main Building Kairaku 3 |
| Treatment strategy for early colorectal cancer with artificial intelligence | |||
| Katsuro Ichimasa1, Shin-ei Kudo1, Hideyuki Miyachi1 | |||
| 1Digestive Diseases Center, Showa University Northern Yokohama Hospital | |||
|
|||
|
|||
| Introduction: Most T1 colorectal cancers (CRCs) undergo surgical resection despite the low incidence (10%) of lymph node metastasis (LNM), leading many unnecessary surgeries. The aim was to investigate whether artificial intelligence (AI) can predict LNM presence of T1 CRCs before surgery. In a substudy, we investigated LNM presence of T2 CRCs with AI which showed about 25% of LNM. Methods: Data on 690 consecutive patients with T1 CRCs that were surgically resected in 2001–2016 were retrospectively analyzed. We used 590 patients for training the AI model, and the remaining 100 patients for validating the model. The AI model analyzed 44 clinicopathological factors and then predicted positivity or negativity for LNM. The AI model was validated by calculating the AUC of ROC, comparing with Japanese guidelines. As for T2 CRCs, we used data on 511 consecutive patients as well. Results: The rate of LNM-positivity was 9% in T1 and 28% in T2, respectively. The AUCs were T1 CRCs －the AI model: 0.82 vs. Japanese guideline: 0.53. The AUC of the AI model in T2 CRCs was 0.96. Conclusion: AI will help in making decisions as to whether additional surgery is indicated after endoscopic resection of T1 CRCs. AI can also accurately predict the presence of LNM in T2 CRCs, which may enable the patients, especially for super elderly to choose the local resection such as full-thickness endoscopic resection. |
|||
|
Index Term 1: Colorectal cancer Index Term 2: Artificial intelligence |
|||
| Page Top | |||