Search

PurposePresence of a long remnant sigmoid colon after left hemicolectomy with inferior mesenteric vein (IMV) ligation for distal transverse and descending colon cancers may be a risk factor for venous ischemia. This study aimed to evaluate the clinical impact of IMV preservation in patients who underwent left hemicolectomy with inferior mesenteric artery (IMA) preservation.MethodsWe included 155 patients who underwent left hemicolectomy with IMA preservation for distal transverse and descending colon cancers from 2003 to 2020. Technical success of IMV preservation was determined by assessing pre- and post-operative patency of the IMV on computed tomography (CT) by an abdominal radiologist. Intestinal complications comprising ulceration, stricture, venous engorgement, and colitis in remnant colon were compared between the IMV preservation and ligation groups.ResultsIMV was preserved in 22 (14.2%) and ligated in 133 (85.8%) patients. Surgical time, postoperative recovery outcomes, and number of harvested lymph nodes were similar in both groups. The technical success of IMV preservation was 81.8%. Intestinal complications were less common in the preservation group than in the IMV ligation group (4.5% vs. 23.3%, P=0.048). The complications in the IMV ligation group were anastomotic ulcer (n=2), anastomotic stricture (n=4), venous engorgement of the remnant distal colon (n=4), and colitis in the distal colon (n=21).ConclusionsIMV preservation may be beneficial after left hemicolectomy with IMA preservation for distal transverse and descending colon cancers. We suggest that IMV preservation might be considered when long remnant sigmoid colon is expected during left hemicolectomy with low ligation of IMA.

Integrated gradients are widely employed to evaluate the contribution of input features in classification models because it satisfies the axioms for attribution of prediction. This method, however, requires an appropriate baseline for reliable determination of the contributions. We propose a compensated integrated gradients method that does not require a baseline. In fact, the method compensates the attributions calculated by integrated gradients at an arbitrary baseline using Shapley sampling. We prove that the method retrieves reliable attributions if the processes of input features in a classifier are mutually independent, and they are identical like shared weights in convolutional neural networks. Using three electroencephalogram datasets, we experimentally demonstrate that the attributions of the proposed method are more reliable than those of the original integrated gradients, and its computational complexity is much lower than that of Shapley sampling., Comment: Machine Learning for Health (ML4H) Workshop at NeurIPS 2018 arXiv:1811.07216