Transmural migrated retained sponges usually impact at the level of the ileo-cecal valve leading to a small bowel obstruction.Once passed through the ileo-cecal valve,a retained sponge can be propelled forward by peri...Transmural migrated retained sponges usually impact at the level of the ileo-cecal valve leading to a small bowel obstruction.Once passed through the ileo-cecal valve,a retained sponge can be propelled forward by peristaltic activity and eliminated with feces.We report the case of a 52-year-old female with a past surgical history and recurrent episodes of abdominal pain and constipation.On physical examination,a generalized resistance was observed with tenderness in the right flank.Contrast-enhanced multi-detector computed tomography findings were consistent with a perforated right colonic diverticulitis with several out-pouchings at the level of the ascending colon and evidence of free air in the right parieto-colic gutter along with an air-fluid collection within the mesentery.In addition,a ring-shaped hyperdense intraluminal material was also noted.At surgery,the ascending colon appeared irregularly thickened and folded with a focal wall interruption and a peri-visceral abscess at the level of the hepatic flexure,but no diverticula were found.A right hemi-colectomy was performed and on dissection of the surgical specimen a retained laparotomy sponge was found in the bowel lumen.展开更多
针对嵌入式平台下卷积神经网络运行速度慢,无法快速手势检测的问题,提出一种基于SSD的卷积神经网络的嵌入式手势检测算法,该算法显著提高了手势检测速度,并保持了高精度。首先通过一种预处理方法,对原来的手势数据库进行5倍扩展;然后对...针对嵌入式平台下卷积神经网络运行速度慢,无法快速手势检测的问题,提出一种基于SSD的卷积神经网络的嵌入式手势检测算法,该算法显著提高了手势检测速度,并保持了高精度。首先通过一种预处理方法,对原来的手势数据库进行5倍扩展;然后对SSD算法的基础神经网络层进行卷积因子分解,使用MobileNet神经网络获得了在CPU下的3倍加速;最后通过改变输入图片大小同时改变网络结构,减少了算法的计算复杂度。实验结果表明所提算法在两个数据集上的平均精度均值(Mean Average Precision,mAP)下降2.7%,但是在Qualcomm SnapDragon820平台下检测一张图片时间可达到0.233 s,检测速度提高40倍以上。展开更多
文摘Transmural migrated retained sponges usually impact at the level of the ileo-cecal valve leading to a small bowel obstruction.Once passed through the ileo-cecal valve,a retained sponge can be propelled forward by peristaltic activity and eliminated with feces.We report the case of a 52-year-old female with a past surgical history and recurrent episodes of abdominal pain and constipation.On physical examination,a generalized resistance was observed with tenderness in the right flank.Contrast-enhanced multi-detector computed tomography findings were consistent with a perforated right colonic diverticulitis with several out-pouchings at the level of the ascending colon and evidence of free air in the right parieto-colic gutter along with an air-fluid collection within the mesentery.In addition,a ring-shaped hyperdense intraluminal material was also noted.At surgery,the ascending colon appeared irregularly thickened and folded with a focal wall interruption and a peri-visceral abscess at the level of the hepatic flexure,but no diverticula were found.A right hemi-colectomy was performed and on dissection of the surgical specimen a retained laparotomy sponge was found in the bowel lumen.
文摘针对嵌入式平台下卷积神经网络运行速度慢,无法快速手势检测的问题,提出一种基于SSD的卷积神经网络的嵌入式手势检测算法,该算法显著提高了手势检测速度,并保持了高精度。首先通过一种预处理方法,对原来的手势数据库进行5倍扩展;然后对SSD算法的基础神经网络层进行卷积因子分解,使用MobileNet神经网络获得了在CPU下的3倍加速;最后通过改变输入图片大小同时改变网络结构,减少了算法的计算复杂度。实验结果表明所提算法在两个数据集上的平均精度均值(Mean Average Precision,mAP)下降2.7%,但是在Qualcomm SnapDragon820平台下检测一张图片时间可达到0.233 s,检测速度提高40倍以上。
