A Deep Learning-Based Model for Classification of Different Subtypes of Subcortical Vascular Cognitive Impairment With FLAIR
基于深度学习的 FLAIR 皮质下血管性认知障碍不同亚型分类模型
- 作者列表："Qi Chen","Yao Wang","Yage Qiu","Xiaowei Wu","Yan Zhou","Guangtao Zhai
Deep learning methods have shown their great capability of extracting high-level features from image and have been used for effective medical imaging classification recently. However, training samples of medical images are restricted by the amount of patients as well as medical ethics issues, making it hard to train the neural networks. In this paper, we propose a novel end-to-end three-dimensional (3D) attention-based residual neural network (ResNet) architecture to classify different subtypes of subcortical vascular cognitive impairment (SVCI) with single-shot T2-weighted fluid-attenuated inversion recovery (FLAIR) sequence. Our aim is to develop a convolutional neural network to provide a convenient and effective way to assist doctors in the diagnosis and early treatment of the different subtypes of SVCI. The experiment data in this paper are collected from 242 patients from the Neurology Department of Renji Hospital, including 78 amnestic mild cognitive impairment (a-MCI), 70 nonamnestic MCI (na-MCI), and 94 no cognitive impairment (NCI). The accuracy of our proposed model has reached 98.6% on a training set and 97.3% on a validation set. The test accuracy on an untrained testing set reaches 93.8% with robustness. Our proposed method can provide a convenient and effective way to assist doctors in the diagnosis and early treatment.
深度学习方法已经显示出其从图像中提取高级特征的巨大能力，并最近被用于有效的医学成像分类。然而，医学图像的训练样本受到患者数量和医学伦理问题的限制，使得神经网络的训练变得困难。在本文中，我们提出了一种新颖的端到端三维 (3D) 基于注意的残差神经网络 (ResNet) 架构用单次激发 T2-weighted 液体衰减反转恢复 (FLAIR) 序列对皮质下血管性认知障碍 (SVCI) 的不同亚型进行分类。我们的目的是开发一种卷积神经网络，为帮助医生诊断和早期治疗不同亚型的 SVCI 提供一种方便有效的方法。本文的实验数据来自仁济医院神经内科的 242 例患者，其中遗忘型轻度认知障碍 (a-MCI) 78 例，非遗忘型 MCI (na-MCI) 70 例,94 无认知功能障碍 (NCI)。我们提出的模型在训练集上的准确性达到了 98.6%，在验证集上的准确性达到了 97.3%。在未训练的测试集上的测试精度达到 93.8%，具有鲁棒性。我们提出的方法可为医生提供方便有效的辅助诊断和早期治疗方法。
METHODS:BACKGROUND:People with stroke are not meeting recommended levels of physical activity. The modifiable factors associated with post-stroke physical activity levels need to be identified to develop targeted interventions. OBJECTIVE:The objective of this study was to investigate the factors at discharge from inpatient rehabilitation that are associated with physical activity levels at 3 months following discharge. DESIGN:This was a prospective cohort study. METHODS:Sixty-four people with stroke completed baseline assessments at discharge from inpatient rehabilitation and 55 completed the follow-up 3 months later. The candidate factors (i.e. gait speed, balance, strength, cognition, mood and motivation) were measured at discharge. The primary outcome measure at follow-up was walking related activity (measured by wrist-worn accelerometer). Secondary outcome measures were physical activity participation (Activity Card Sort) and intensity of physical activity (International Physical Activity Questionnaire - Short 7 days). Adjusted separate multivariable linear regression models or proportional odds regression models were used to evaluate the associations between candidate factors and physical activity. RESULTS:Gait speed and balance were associated with all aspects of physical activity. Higher level of intrinsic motivation was also associated with higher physical activity participation. Anxiety demonstrated a significant non-linear relationship with physical activity participation. LIMITATIONS:Inclusion of fatigue and individual muscle strength could have provided further insights into associations with steps per day. CONCLUSION:The results demonstrated that better physical function at discharge from inpatient rehabilitation was associated with future increased levels of physical activity. Additionally, higher levels of motivation impacted on increased physical activity participation. The influence of anxiety on physical activity participation requires further exploration. Mixed-method study designs can be utilized to further understand the factors associated with post-stroke physical activity.
METHODS:Cerebral ischemia-reperfusion (I/R) is characterized by initial transient cerebral ischemia followed by reperfusion. Various pathophysiological processes are involved in brain injury and functional recovery during cerebral I/R. There are few studies on dynamic metabolic process after cerebral I/R. The present study was to observe dynamic alteration of brain injury, functional recovery, and metabolites after cerebral I/R in rats and discover potential metabolic markers. The cerebral I/R model was established by middle cerebral artery occlusion (MCAO) for 90 min, following reperfusion in rats. The results of cerebral infarction area, cerebral edema, and behavior test showed that there were dynamic changes in brain injury and functional recovery at different periods after cerebral I/R. Further analysis showed that the brain injury was severe on the first day of cerebral I/R, and there was a significant functional recovery from the 7th day of cerebral I/R, followed by an aggravation trend of brain injury from the days 7 to 28. Furthermore, Matrix-assisted laser desorption ionization mass spectrometry imaging analysis showed that the expression of ATP, glucose, and citric acid on 7th day was the highest during cerebral I/R, which indicated that energy metabolism and oxidative phosphorylation played important roles during cerebral I/R. In addition, the untargeted metabolomic results showed that the level of isocitric acid, the ratio of oxyglutaric acid/glutamic acid, and the level of pyruvic acid associated with the TCA cycle were also the highest on the 7th day during cerebral I/R, which indicated that the transient spontaneous recovery of ischemic brain on the 7th day after ischemia-reperfusion might be related to oxidative phosphorylation and energy metabolism in the brain in this period. In conclusion, the results suggest that some small molecule metabolites participate in the brain injury and functional recovery during cerebral I/R, which is of great significance to the development of therapeutic drugs and diagnostic markers.
METHODS:The aims of this study were to study the effects of miR-2 on cerebral ischemia–reperfusion rats and to explore its further mechanism. Rats were assigned into sham, model, miR-22 control and miR-22 groups. Observation of neurological behaviors at 24 h after operation found that neurological functions were severely damaged in the model and miR-22 control groups and these damages were improved by miR-22. RT-PCR indicated that miR-22 mRNA level in the brain tissue was significantly decreased in the model and miR-22 control groups, but increased in the miR-22 group. TTC staining showed increased percentage of cerebral infarction volume in the model and miR-22 control groups and this increase was reduced by miR-22. Immunohistochemistry showed increased densities of CD34^+ and VEGF^+ microvessels in the cortex in the model and miR-22 control groups, which were further increased in the miR-22 group. ELISA showed increased serum VEGF and Ang-1 levels in the model and miR-22 control groups, which were also further increased in the miR-22 group. Western blot analysis showed increased phosphorylation level of PI3K and Akt in brain tissue in the model and miR-22 control groups, which were further increased in the miR-22 group. Administration of LY294002, a specific PI3K pathway inhibitor, significantly reversed all the effects of miR-22 on rats in the model group. miR-22 exerts its neuroprotective and angiogenic functions via the PI3K/Akt signaling pathway, at least partly, in rats under cerebral ischemia–reperfusion.