Neuroimaging and deep learning for brain stroke detection - A review of recent advancements and future prospects.

Research progress on selective catalytic reduction (SCR) catalysts for NO removal from coal-fired flue gas

In this study, we propose a three-dimensional Medical image classifier using Multi-plane and Multi-slice Trans-former (M3T) network to classify Alzheimer's disease (AD) in 3D MRI images. The proposed network synergically com-bines 3D CNN, 2D CNN, and Transformer for accurate AD classification. The 3D CNN is used to perform natively 3D representation learning, while 2D CNN is used to utilize the pre-trained weights on large 2D databases and 2D repre-sentation learning. It is possible to efficiently extract the lo-cality information for AD-related abnormalities in the local brain using CNN networks with inductive bias. The trans-former network is also used to obtain attention relationships among multi-plane (axial, coronal, and sagittal) and multi-slice images after CNN. It is also possible to learn the ab-normalities distributed over the wider region in the brain using the transformer without inductive bias. In this ex-periment, we used a training dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) which contains a total of 4,786 3D T1-weighted MRI images. For the validation data, we used dataset from three different institutions: The Australian Imaging, Biomarker and Lifestyle Flagship Study of Ageing (AIBL), The Open Access Series of Imaging Studies (OASIS), and some set of ADNI data indepen-dent from the training dataset. Our proposed M3T is compared to conventional 3D classification networks based on an area under the curve (AUC) and classification accuracy for AD classification. This study represents that the pro-posed network M3T achieved the highest performance in multi-institutional validation database, and demonstrates the feasibility of the method to efficiently combine CNN and Transformer for 3D medical images.

M3T: three-dimensional Medical image classifier using Multi-plane and Multi-slice Transformer

A series of MnO2 (with different crystal types) doping CeO2 catalysts were synthesized by combining hydrothermal method and impregnation method, and the SCR activity and N2O formation of the catalysts were investigated. The results showed that NO conversion rate of CeO2 @ α-MnO2, CeO2 @ β-MnO2 and CeO2 @ γ-MnO2 catalysts from 75 to 250 °C was always more than 95%, which showed excellent low-temperature NH3-SCR activity. N2O amount produced by CeO2 @ α-MnO2, CeO2 @ β-MnO2 and CeO2 @ γ-MnO2 catalysts during denitrification all reduced at 75–125 °C comparing with pure MnO2, and the N2O in denitration process mainly came from SCR reaction, but not from direct oxidation of NH3. Besides, the reducibility and acidity of the complex oxides catalyst were improved, which was more beneficial to the reaction of adsorbed NH3 with NO, thus reducing the direct oxidation of adsorbed NH3 by O2 and inhibiting the formation of N2O. What’s more, the effect of Ce and different crystal types of MnO2 also influenced N2O generation behavior. Moreover, the SCR reaction of CeO2 @ α-MnO2 and CeO2 @ β-MnO2 catalysts mainly followed Eley-Rideal (E-R) mechanism, while CeO2 @ γ-MnO2 catalyst followed Langmuir-Hinshelwood (L-H) mechanism. 

Effect of MnO2 crystal types on CeO2@MnO2 oxides catalysts for low-temperature NH3-SCR

A series of Cu-exchanged zeolite NaX (derived from blast furnace slag) catalysts were synthesized by an ion-exchanged method. The low-temperature catalytic activity and SO2 or/and H2O resistance of the catalysts were studied as well. The results found that the Cu(N)X catalyst exhibited excellent low-temperature selective catalytic reduction (SCR) activity, whose NO conversion was nearly 100% at ca. 175 °C and remained stable, while the Cu(C)X catalyst showed the poorest catalytic performance. When SO2 and H2O was introduced, the NO conversion of the Cu(N)X catalyst declined from 100% to 60%, and it could recover to ca. 63% after cutting off SO2 and H2O. More isolated Cu2+ species and surface chemisorbed oxygen species on the Cu(N)X catalyst could provide more active sites for SCR reaction, and more surface oxygen vacancies accelerated the oxidation of absorbed NH3 species and NO species. Besides, the higher surface acidity and redox ability on the Cu(N)X catalyst promoted the SCR reaction performance. In addition, after the poisoning of SO2/H2O, the surface acidity decreased, redox ability became poor, and the Cu active species was less than that of the Cu(N)X catalyst, indicating that the Cu active sites were consumed during the reaction. Furthermore, it could also be found that the generated sulfate species occupied the main active sites and reduced the varieties of nitrate species and intermediates, thereby reducing the SCR activity of the catalyst. 

Low-Cost CuX Catalyst from Blast Furnace Slag Waste for Low-Temperature NH3-SCR: Nature of Cu Active Sites and Influence of SO2/H2O

Effect of Mn and Ce oxides on low-temperature NH3-SCR performance over blast furnace slag-derived zeolite X supported catalysts

Diffusion-weighted magnetic resonance imaging (DWI) is sensitive to acute ischemic stroke and is a common diagnostic method for the stroke. However, the diagnostic result relies on the visual observation of neurologists which may vary from doctor to doctor under different circumstance. And manual segmentation is often a time-consuming and subjective process. The time from onset to thrombus removal has a significant impact on the prognosis of patients with acute ischemic stroke. The shorter the time, the better the prognosis. For this purpose we present a novel framework to quickly and automatically segment the ischemic stroke lesions on DWI. We mainly have three contributions: firstly, we design a detection and segmentation network (DSN) to solve the two kinds of data imbalance; secondly, we propose a triple-branch DSN architecture, used for extracting different plane feature respectively; thirdly, we propose a multi-plane fusion network (MPFN), which aims to make final prediction more accurate. Extensive experiments on ISLES2015 SSIS DWI sequence dataset demonstrate the superiority of our proposed segmentation method. The dice reached 62.2% and the sensitivity reached 71.7%.

/pdf/ischemic-stroke-lesion-segmentation-using-multi-plane-gygm5so71u.pdf

Ischemic Stroke Lesion Segmentation Using Multi-Plane Information Fusion

Time-resolved in situ DRIFTS study on NH3–SCR of NO on CeO2/TiO2 catalyst

The invention implements a method and a device for detecting ischemic stroke focus based on brain CT images. In this method, brain CT images were preprocessed and region growing based on region growing method, which increased the difference of image characteristics between lesion region and normal brain. Then, based on the symmetry of the brain, The images of the left and right brain regions are divided into a plurality of pixel squares, and the characteristic differences of two pixel squares with symmetry positions in the left and right brain are compared one by one, so that it is possible todetermine which pixel square contains the focus of ischemic stroke according to the characteristic differences. Through the above steps, the purpose of effectively detecting ischemic stroke lesions by brain CT images is realized.

Method and device for detecting ischemic stroke focus based on brain CT image

Intracranial hemorrhage (ICH) classification from computed tomography (CT) scans is important for accurate diagnosis of stroke in emergency centers. However, it is challenged by low image contrast and the complex appearance of the lesion. Therefore, the diagnosis results may vary from doctor to doctor under different situations. Existing auxiliary diagnosis algorithms treat the features of each location as equally important, and they do not impose any constraints on the learned features. The learned features used for classification in the model contain noise or features that have no diagnostic significance, which greatly limits the performance and reliability of systems. To facilitate the ICH treatment, we propose to use the auxiliary weak-segmentation-label supervision in multiscale CNN to classify ICH lesions on brain CT images, where the ICH lesions are segmented and used to guide the attention of the classification network. The weak segmentation labels we use are generated by an unsupervised algorithm and do not require additional annotations. We conduct experiments on the RSNA ICH detection data set, and the results demonstrate that our method can achieve an average ACC of 98.1% and an average $F1$ score of 74.6% on the test set of nearly 80 000 CT images. The proposed weakly supervised-guided attention mechanism can accurately activate those neurons related to diagnosis, while inhibiting the activation of irrelevant areas and noise, thus achieving good performance.

Lian Liu

Papers

Effect of Mn and Ce oxides on low-temperature NH3-SCR performance over blast furnace slag-derived zeolite X supported catalysts

Ischemic Stroke Lesion Segmentation Using Multi-Plane Information Fusion

Time-resolved in situ DRIFTS study on NH3–SCR of NO on CeO2/TiO2 catalyst

Method and device for detecting ischemic stroke focus based on brain CT image

A Weakly Supervised-Guided Soft Attention Network for Classification of Intracranial Hemorrhage