Objective To evaluate the reliability of three dimensional spiral fast spin echo pseudo-continuous arterial spin labeling(3 D pc-ASL) in measuring cerebral blood flow(CBF) with different post-labeling delay time(PLD) ...Objective To evaluate the reliability of three dimensional spiral fast spin echo pseudo-continuous arterial spin labeling(3 D pc-ASL) in measuring cerebral blood flow(CBF) with different post-labeling delay time(PLD) in the resting state and the right finger taping state.Methods 3 D pc-ASL and three dimensional T1-weighted fast spoiled gradient recalled echo(3 D T1-FSPGR) sequence were applied to eight healthy subjects twice at the same time each day for one week interval. ASL data acquisition was performed with post-labeling delay time(PLD) 1.5 seconds and 2.0 seconds in the resting state and the right finger taping state respectively. CBF mapping was calculated and CBF value of both the gray matter(GM) and white matter(WM) was automatically extracted. The reliability was evaluated using the intraclass correlation coefficient(ICC) and Bland and Altman plot.Results ICC of the GM(0.84) and WM(0.92) was lower at PLD 1.5 seconds than that(GM, 0.88; WM, 0.94) at PLD 2.0 seconds in the resting state, and ICC of GM(0.88) was higher in the right finger taping state than that in the resting state at PLD 1.5 seconds. ICC of the GM and WM was 0.71 and 0.78 for PLD 1.5 seconds and PLD 2.0 seconds in the resting state at the first scan, and ICC of the GM and WM was 0.83 and 0.79 at the second scan, respectively.Conclusion This work demonstrated that 3 D pc-ASL might be a reliable imaging technique to measure CBF over the whole brain at different PLD in the resting state or controlled state.展开更多
In the field of medical images,pixel-level labels are time-consuming and expensive to acquire,while image-level labels are relatively easier to obtain.Therefore,it makes sense to learn more information(knowledge)from ...In the field of medical images,pixel-level labels are time-consuming and expensive to acquire,while image-level labels are relatively easier to obtain.Therefore,it makes sense to learn more information(knowledge)from a small number of hard-to-get pixel-level annotated images to apply to different tasks to maximize their usefulness and save time and training costs.In this paper,using Pixel-Level Labeled Images forMulti-Task Learning(PLDMLT),we focus on grading the severity of fundus images for Diabetic Retinopathy(DR).This is because,for the segmentation task,there is a finely labeled mask,while the severity grading task is without classification labels.To this end,we propose a two-stage multi-label learning weakly supervised algorithm,which generates initial classification pseudo labels in the first stage and visualizes heat maps at all levels of severity using Grad-Cam to further provide medical interpretability for the classification task.A multitask model framework with U-net as the baseline is proposed in the second stage.A label update network is designed to alleviate the gradient balance between the classification and segmentation tasks.Extensive experimental results show that our PLDMLTmethod significantly outperforms other stateof-the-art methods in DR segmentation on two public datasets,achieving up to 98.897%segmentation accuracy.In addition,our method achieves comparable competitiveness with single-task fully supervised learning in the DR severity grading task.展开更多
当前,深度主动学习(Deep Active Learning,DAL)在分类数据标注工作中获得成功,但如何筛选出最能提升模型性能的样本仍是难题.本文提出基于弱标签争议的半自动分类数据标注方法(Dispute about Weak Label based Deep Active Learning,DWL...当前,深度主动学习(Deep Active Learning,DAL)在分类数据标注工作中获得成功,但如何筛选出最能提升模型性能的样本仍是难题.本文提出基于弱标签争议的半自动分类数据标注方法(Dispute about Weak Label based Deep Active Learning,DWLDAL),迭代地筛选出模型难以区分的样本,交给人工进行准确标注.该方法包含伪标签生成器和弱标签生成器,伪标签生成器是在准确标注的数据集上训练而成,用于生成无标签数据的伪标签;弱标签生成器则是在带伪标签的随机子集上训练而成.弱标签生成器委员会决定哪些无标签数据最有争议,则交给人工标注.本文针对文本分类问题,在公开数据集IMDB(Internet Movie DataBase)、20NEWS(20NEW Sgroup)和chnsenticorp(chnsenticorp_htl_all)上进行实验验证.从数据标注和分类任务的准确性2个角度,对3种不同投票决策方式进行评估.DWLDAL方法中数据标注的F1分数比现有方法Snuba分别提高30.22%、14.07%和2.57%,DWLDAL方法中分类任务的F1分数比Snuba分别提高1.01%、22.72%和4.83%.展开更多
基金Supported by the Foundation for Medical and Health Sci&Tech Innovation Project of Sanya(2016YW37)the Special Financial Grant from China Postdoctoral Science Foundation(2014T70960)
文摘Objective To evaluate the reliability of three dimensional spiral fast spin echo pseudo-continuous arterial spin labeling(3 D pc-ASL) in measuring cerebral blood flow(CBF) with different post-labeling delay time(PLD) in the resting state and the right finger taping state.Methods 3 D pc-ASL and three dimensional T1-weighted fast spoiled gradient recalled echo(3 D T1-FSPGR) sequence were applied to eight healthy subjects twice at the same time each day for one week interval. ASL data acquisition was performed with post-labeling delay time(PLD) 1.5 seconds and 2.0 seconds in the resting state and the right finger taping state respectively. CBF mapping was calculated and CBF value of both the gray matter(GM) and white matter(WM) was automatically extracted. The reliability was evaluated using the intraclass correlation coefficient(ICC) and Bland and Altman plot.Results ICC of the GM(0.84) and WM(0.92) was lower at PLD 1.5 seconds than that(GM, 0.88; WM, 0.94) at PLD 2.0 seconds in the resting state, and ICC of GM(0.88) was higher in the right finger taping state than that in the resting state at PLD 1.5 seconds. ICC of the GM and WM was 0.71 and 0.78 for PLD 1.5 seconds and PLD 2.0 seconds in the resting state at the first scan, and ICC of the GM and WM was 0.83 and 0.79 at the second scan, respectively.Conclusion This work demonstrated that 3 D pc-ASL might be a reliable imaging technique to measure CBF over the whole brain at different PLD in the resting state or controlled state.
文摘In the field of medical images,pixel-level labels are time-consuming and expensive to acquire,while image-level labels are relatively easier to obtain.Therefore,it makes sense to learn more information(knowledge)from a small number of hard-to-get pixel-level annotated images to apply to different tasks to maximize their usefulness and save time and training costs.In this paper,using Pixel-Level Labeled Images forMulti-Task Learning(PLDMLT),we focus on grading the severity of fundus images for Diabetic Retinopathy(DR).This is because,for the segmentation task,there is a finely labeled mask,while the severity grading task is without classification labels.To this end,we propose a two-stage multi-label learning weakly supervised algorithm,which generates initial classification pseudo labels in the first stage and visualizes heat maps at all levels of severity using Grad-Cam to further provide medical interpretability for the classification task.A multitask model framework with U-net as the baseline is proposed in the second stage.A label update network is designed to alleviate the gradient balance between the classification and segmentation tasks.Extensive experimental results show that our PLDMLTmethod significantly outperforms other stateof-the-art methods in DR segmentation on two public datasets,achieving up to 98.897%segmentation accuracy.In addition,our method achieves comparable competitiveness with single-task fully supervised learning in the DR severity grading task.
文摘当前,深度主动学习(Deep Active Learning,DAL)在分类数据标注工作中获得成功,但如何筛选出最能提升模型性能的样本仍是难题.本文提出基于弱标签争议的半自动分类数据标注方法(Dispute about Weak Label based Deep Active Learning,DWLDAL),迭代地筛选出模型难以区分的样本,交给人工进行准确标注.该方法包含伪标签生成器和弱标签生成器,伪标签生成器是在准确标注的数据集上训练而成,用于生成无标签数据的伪标签;弱标签生成器则是在带伪标签的随机子集上训练而成.弱标签生成器委员会决定哪些无标签数据最有争议,则交给人工标注.本文针对文本分类问题,在公开数据集IMDB(Internet Movie DataBase)、20NEWS(20NEW Sgroup)和chnsenticorp(chnsenticorp_htl_all)上进行实验验证.从数据标注和分类任务的准确性2个角度,对3种不同投票决策方式进行评估.DWLDAL方法中数据标注的F1分数比现有方法Snuba分别提高30.22%、14.07%和2.57%,DWLDAL方法中分类任务的F1分数比Snuba分别提高1.01%、22.72%和4.83%.
