Over several decades, solid-state electrodes in which reversible intercalation (insertion) and deintercalation (extraction) of cationic guest atoms occur along with accompanying electron flow without any change of their crystal structure, have attracted great interest in fundamental and practical perspectives for improving the performance of rechargeable batteries. This chapter provides comprehensive reviews of principle and recent advances especially in thermodynamic and kinetic approaches to lithium intercalation into, and deintercalation from, transition metals oxides and carbonaceous materials. Thermodynamic properties such as chemical potential, entropy and enthalpy of lithium intercalation/deintercalation are first discussed, based on a lattice gas model with various approximations. Lithium intercalation/deintercalation involving an order-disorder transition or a two-phase coexistence caused by strong interaction of lithium ions in solid-state electrodes is explained, based on the lattice gas model and with the help of computational methods. Second, the kinetics of lithium intercalation/ deintercalation is treated in detail on the basis of a cell-impedance-controlled model. Anomalous features of potentiostatic current transients obtained experimentally from transitional metal oxide and carbonaceous electrodes, which are hardly explained under a diffusion control model, are readily analyzed by the cell-impedancecontrolled lithium transport concept, with the aid of computational methods. IntroductionWhen cationic guest atoms such as lithium, hydrogen, and sodium reversibly enter or leave the host oxide crystal, along with an accompanying electron flow but without any change in crystal structure, the reaction is referred to as intercalation/ deintercalation as follows [1,2]::1Þ Solid State Electrochemistry I: Fundamentals, Materials and their Applications. Edited by Vladislav V. Kharton

Background The heterogeneity of Alzheimer’s disease is important theme for studying the epidemiology of the disease. Practically, understanding disease phenotypes helps deciding treatment before proceeding to dementia. Generally, it is known that AD’s subtype can be divided into three. Method We used 660 mild cognitive impairment (MCI) patients’ T1‐weighted magnetic resonance images (MRI) from Alzheimer’s Disease Neuroimaging Initiative (ADNI). Inputs are atrophy vectors which mean excluding the positive part and resigned W‐scores calculated from 81,924 vertex‐level cortical‐thicknesses extracted from T1. Unlike existing clustering‐only or separated clustering‐classification models, we used a model which is jointly optimized with respect to both clustering and classification. The model is comprised of three parts: 1) feature extractor, 2) clustering module, and 3) classifier. In first part, we extract important features using simple neural network. Then, we do clustering using the deep‐features and obtain pseudo‐labels in latent space. Finally, we implement classification at last layer with the answers which are assigned from clustering module. It is end‐to‐end training being able to find the solution of combined objective function. Results Results were represented from 10 fold cross‐validation. Average log‐likelihood of Gaussian mixture and normalized mutual information of labels between adjacent iterations are used as clustering performance metric, cross‐entropy loss and accuracy as classification. It is difficult for such models to converge according to referenced studies. But, in this study, all the metrics were well‐converged. Atrophy maps show group comparison results of cortical thicknesses between each subtype and cognitive normal (CN) patient with FDR correction and W‐score averaged in each subtype. The demographic table about Aβ positivity and dementia conversion percentages without missing information and neuropsychological test scores table of each subtypes showed that subtype‐I has highest degree of disease severity, then subtype‐II and finally III. External validation with 662 aMCI patients’ data from Samsung Medical Center (SMC) used in similar study represented bigger gap of dementia conversion percentages between subtype‐I and II. Conclusions In this study, we made clustering‐classification model for individualized subtype prediction. It is end‐to‐end training method optimizing clustering and classification loss simultaneously. The results showed our model exceeds existing subtype clustering model.

Background Alzheimer’s disease (AD) usually appears in mid‐60s (Late‐Onset AD, LOAD), but also rarely appears in patients younger than 65 (Early‐Onset AD, EOAD), while showing distinct features. Tau shows high correlation with AD prognosis and can be detected through positron emission tomography (PET) scan. Tau data mapped into a low dimensional latent space will help detect longitudinal change pattern along AD spectrum. Furthermore, fitting tensors using individual line segments will enable us to track individual tau trajectory in the latent space. Method We recruited 74 healthy controls, 75 mild cognitive impairment patients, and 38 AD patients with two PET scans ([18F]Flortaucipir for tau and [18F]Florbetaben for Aβ) with two‐year follow‐up. We used standardized uptake value ratios (SUVRs) for tau PET data. All patients’ tau data went through autoencoder and principal component analysis, pretrained using baseline (BL) tau data. Using only amyloid positive AD patients’ BL – follow‐up (FU) line segments, we fit tensors and tracked individual trajectories in the latent space using a tensor tracing algorithm. Result In the 2‐dimensional latent space the points dispersed toward the upper‐ and right‐side as the disease deteriorated. Reconstructed SUVR maps showed tau accumulation pattern of EOAD through x‐axis, such as tau spreading in temporal, prefrontal, and parietal lobes. Reconstructed map of 1st quartile group in y‐axis showed tau spreading dominance in parietal lobe and 4th quartile group showed dominance in prefrontal, temporal lobe. Each patient’s trajectory, predicted based on the latent space, showed reconstructed map following individual tau accumulation pattern. In addition, EOAD tau spreading trajectory tend to precede LOAD tau spreading trajectory. Conclusion We decomposed high dimensional tau accumulation pattern into 2 independent components through deep‐learning methods. X‐axis showed EOAD tau accumulation pattern and y‐axis was divided into LOAD tau accumulation pattern and non‐prefrontal dominant pattern. In addition, tensor fitting of the longitudinal vector field enables us to find individual tau spreading trajectories which can be used to predict future and past spreading. Even though this study was done using only one factor, tau, it has the potential to be expanded to other factors such as amyloid or neurodegeneration data.

Background: Considerable evidence confirms the accumulation of fibrillar Amyloidβ(Aβ) and aggregates of hyperphosphorylated tau as early Alzheimer's Disease (AD) biomarker changes. However, the relationships between amyloid and tau within AD pathogenesis are not fully understood. Estimating the integrated order of these two biomarkers abnormality and identifying associations between two biomarkers in vivo could facilitate accurate prediction of patient-tailored prognosis in early disease stage. Method:We recruited 74 healthy controls, 75 MCI patients, and 38 AD patients with T1-weighted magnetic resonance image (MRI) and two PET scans ([ 18 F]Flortaucipir for tau and [ 18 F]Florbetaben for Aβ) with about two-year follow-up data. All regional Aβ and regional tau were aligned simultaneously in the order of posterior probabilities of biomarkers becoming abnormal using the DEBM algorithm. All participants were then labeled with disease staging score based on the integrated biomarker timelines. We then explored the relationship between the disease score at baseline and the patient's diagnosis in follow and annual change of CDR-SB scores.Result: Firstly, the integrated biomarker ordering shows a partially overlapped sequence of the Thal stage (in terms of the Aβ biomarker) and the Braak stage (in terms of the tau biomarker). Tau accumulates first within a region called the entorhinal cortex, then spreads out of that region. In the neocortical region (except the entorhinal cortex), the aggregation of amyloid precedes that of tau. Secondly, MCI patients' stage at baseline matches well with the diagnosis labels at the same time point, and also significantly correlates to the change of CDR-SB scores. Furthermore, MCI patients with higher disease stage scores at baseline have significantly higher chance to convert to AD at follow-up. Conclusion:In this study, we proposed in vivo temporal alterations of both regional amyloid and regional tau simultaneously. All biomarkers exhibited analogous trends to previous research. Amyloid aggregation generally occurs at early stages of the disease and seems to affect not tau seeding but initial tau spreading. Patient staging based on the integrated order of regional biomarker abnormality can help prediction of AD prognosis and future cognition change in early disease stage.

BackgroundAlzheimer's disease (AD) is a neurodegenerative disease characterized by the progressive loss of neurons. To predict such progression, deep neural network models with longitudinal data have been used recently. However, there is a limit that they didn't consider graph topology of brain network, while it is well‐known that neurodegeneration progresses along the large‐scale brain network. In this study, we proposed a graph neural network (GNN) based deep recurrent model to predict longitudinal progression of neurodegeneration in mild cognitive impairment (MCI) patients.MethodWe used 1,146 T1‐weighted magnetic resonance imaging from Alzheimer’s Disease Neuroimaging Initiative. The images were of 320 MCI subjects at baseline with longitudinal data at 2‐year intervals. The length of the data ranged from 3 to 7 years, and an average length is 3.58. We extracted cortical thicknesses using FreeSurfer v6.0 and Desikan‐Killiany atlas. Additionally, we acquired diffusion tensor imaging and resting‐state functional magnetic resonance imaging from Human Connectome Project to extract normative averaged structural connectivity (SC) and resting‐state functional connectivity (rsFC). We also used spatial connectivity of which the element for comparison. We adopted graph‐LSTM model in our problem setting: we used mean cortical thickness as node feature and connectivity as adjacency matrix. We performed 5‐fold cross‐validation and evaluated the performance using the test dataset.ResultWe compared performance of our model with two baseline models, which have been used to predict disease progression but couldn’t learn graph topology of brain, such as RNN and LSTM. The loss was significantly lower at our model than at both baseline models by about one seventh. We implemented additional studies using three distinct connectivities as mentioned above. When using SC and rsFC as adjacency matrix, model performances were increased by about one third compared to using spatial connectivity. In addition, raw cortical thicknesses between prediction and observation were not significantly different in most regions except right anterior cingulate in year 8. Conclusions: We predicted progression of neurodegeneration using graph‐LSTM considering brain networks. We found substituting GNN for embedding layers of LSTM improved performance. We also identified utilizing SC and rsFC as underlying structure of neurodegeneration decreased the prediction error.