Dampened Slow Oscillation Connectivity Anticipates Amyloid Deposition in the PS2APP Mouse Model of Alzheimer's Disease

Performance Evaluation of Deep, Shallow and Ensemble Machine Learning Methods for the Automated Classification of Alzheimer's Disease

Artificial intelligence (AI)-based approaches are crucial in computer-aided diagnosis (CAD) for various medical applications. Their ability to quickly and accurately learn from complex data is remarkable. Deep learning (DL) models have shown promising results in accurately classifying Alzheimer's disease (AD) and its related cognitive states, Early Mild Cognitive Impairment (EMCI) and Late Mild Cognitive Impairment (LMCI), along with the healthy conditions known as Cognitively Normal (CN). This offers valuable insights into disease progression and diagnosis. However, certain traditional machine learning (ML) classifiers perform equally well or even better than DL models, requiring less training data. This is particularly valuable in CAD in situations with limited labeled datasets. In this paper, we propose an ensemble classifier based on ML models for magnetic resonance imaging (MRI) data, which achieved an impressive accuracy of 96.52%. This represents a 3-5% improvement over the best individual classifier. We evaluated popular ML classifiers for AD classification under both data-scarce and data-rich conditions using the Alzheimer's Disease Neuroimaging Initiative and Open Access Series of Imaging Studies datasets. By comparing the results to state-of-the-art CNN-centric DL algorithms, we gain insights into the strengths and weaknesses of each approach. This work will help users to select the most suitable algorithm for AD classification based on data availability.

Interpreting artificial intelligence models: a systematic review on the application of LIME and SHAP in Alzheimer's disease detection

Explainable artificial intelligence (XAI) has gained much interest in recent years for its ability to explain the complex decision-making process of machine learning (ML) and deep learning (DL) models. The Local Interpretable Model-agnostic Explanations (LIME) and Shaply Additive exPlanation (SHAP) frameworks have grown as popular interpretive tools for ML and DL models. This article provides a systematic review of the application of LIME and SHAP in interpreting the detection of Alzheimer's disease (AD). Adhering to PRISMA and Kitchenham's guidelines, we identified 23 relevant articles and investigated these frameworks' prospective capabilities, benefits, and challenges in depth. The results emphasise XAI's crucial role in strengthening the trustworthiness of AI-based AD predictions. This review aims to provide fundamental capabilities of LIME and SHAP XAI frameworks in enhancing fidelity within clinical decision support systems for AD prognosis.

Toward the Identification of Neurophysiological Biomarkers for Alzheimer's Disease in Down Syndrome: A Potential Role for Cross-Frequency Phase-Amplitude Coupling Analysis

Cross-frequency coupling (CFC) mechanisms play a central role in brain activity. Pathophysiological mechanisms leading to many brain disorders, such as Alzheimer's disease (AD), may produce unique patterns of brain activity detectable by electroencephalography (EEG). Identifying biomarkers for AD diagnosis is also an ambition among research teams working in Down syndrome (DS), given the increased susceptibility of people with DS to develop early-onset AD (DS-AD). Here, we review accumulating evidence that altered theta-gamma phase-amplitude coupling (PAC) may be one of the earliest EEG signatures of AD, and therefore may serve as an adjuvant tool for detecting cognitive decline in DS-AD. We suggest that this field of research could potentially provide clues to the biophysical mechanisms underlying cognitive dysfunction in DS-AD and generate opportunities for identifying EEG-based biomarkers with diagnostic and prognostic utility in DS-AD.

Rescue of astrocyte activity by the calcium sensor STIM1 restores long-term synaptic plasticity in female mice modelling Alzheimer's disease

Calcium dynamics in astrocytes represent a fundamental signal that through gliotransmitter release regulates synaptic plasticity and behaviour. Here we present a longitudinal study in the PS2APP mouse model of Alzheimer's disease (AD) linking astrocyte Ca²⁺ hypoactivity to memory loss. At the onset of plaque deposition, somatosensory cortical astrocytes of AD female mice exhibit a drastic reduction of Ca²⁺ signaling, closely associated with decreased endoplasmic reticulum Ca²⁺ concentration and reduced expression of the Ca²⁺ sensor STIM1. In parallel, astrocyte-dependent long-term synaptic plasticity declines in the somatosensory circuitry, anticipating specific tactile memory loss. Notably, we show that both astrocyte Ca²⁺ signaling and long-term synaptic plasticity are fully recovered by selective STIM1 overexpression in astrocytes. Our data unveil astrocyte Ca²⁺ hypoactivity in neocortical astrocytes as a functional hallmark of early AD stages and indicate astrocytic STIM1 as a target to rescue memory deficits.

Preclinical Models for Alzheimer's Disease: Past, Present, and Future Approaches

A robust preclinical disease model is a primary requirement to understand the underlying mechanisms, signaling pathways, and drug screening for human diseases. Although various preclinical models are available for several diseases, clinical models for Alzheimer's disease (AD) remain underdeveloped and inaccurate. The pathophysiology of AD mainly includes the presence of amyloid plaques and neurofibrillary tangles (NFT). Furthermore, neuroinflammation and free radical generation also contribute to AD. Currently, there is a wide gap in scientific approaches to preventing AD progression. Most of the available drugs are limited to symptomatic relief and improve deteriorating cognitive functions. To mimic the pathogenesis of human AD, animal models like 3XTg-AD and 5XFAD are the primarily used mice models in AD therapeutics. Animal models for AD include intracerebroventricular-streptozotocin (ICV-STZ), amyloid beta-induced, colchicine-induced, etc., focusing on parameters such as cognitive decline and dementia. Unfortunately, the translational rate of the potential drug candidates in clinical trials is poor due to limitations in imitating human AD pathology in animal models. Therefore, the available preclinical models possess a gap in AD modeling. This paper presents an outline that critically assesses the applicability and limitations of the current approaches in disease modeling for AD. Also, we attempted to provide key suggestions for the best-fit model to evaluate potential therapies, which might improve therapy translation from preclinical studies to patients with AD.

ABOT: an open-source online benchmarking tool for machine learning-based artefact detection and removal methods from neuronal signals

Brain signals are recorded using different techniques to aid an accurate understanding of brain function and to treat its disorders. Untargeted internal and external sources contaminate the acquired signals during the recording process. Often termed as artefacts, these contaminations cause serious hindrances in decoding the recorded signals; hence, they must be removed to facilitate unbiased decision-making for a given investigation. Due to the complex and elusive manifestation of artefacts in neuronal signals, computational techniques serve as powerful tools for their detection and removal. Machine learning (ML) based methods have been successfully applied in this task. Due to ML’s popularity, many articles are published every year, making it challenging to find, compare and select the most appropriate method for a given experiment. To this end, this paper presents ABOT (Artefact removal Benchmarking Online Tool) as an online benchmarking tool which allows users to compare existing ML-driven artefact detection and removal methods from the literature. The characteristics and related information about the existing methods have been compiled as a knowledgebase (KB) and presented through a user-friendly interface with interactive plots and tables for users to search it using several criteria. Key characteristics extracted from over 120 articles from the literature have been used in the KB to help compare the specific ML models. To comply with the FAIR (Findable, Accessible, Interoperable and Reusable) principle, the source code and documentation of the toolbox have been made available via an open-access repository.

Accelerated Aging Characterizes the Early Stage of Alzheimer's Disease

For Alzheimer's disease (AD), aging is the main risk factor, but whether cognitive impairments due to aging resemble early AD deficits is not yet defined. When working with mouse models of AD, the situation is just as complicated, because only a few studies track the progression of the disease at different ages, and most ignore how the aging process affects control mice. In this work, we addressed this problem by comparing the aging process of PS2APP (AD) and wild-type (WT) mice at the level of spontaneous brain electrical activity under anesthesia. Using local field potential recordings, obtained with a linear probe that traverses the posterior parietal cortex and the entire hippocampus, we analyzed how multiple electrical parameters are modified by aging in AD and WT mice. With this approach, we highlighted AD specific features that appear in young AD mice prior to plaque deposition or that are delayed at 12 and 16 months of age. Furthermore, we identified aging characteristics present in WT mice but also occurring prematurely in young AD mice. In short, we found that reduction in the relative power of slow oscillations (SO) and Low/High power imbalance are linked to an AD phenotype at its onset. The loss of SO connectivity and cortico-hippocampal coupling between SO and higher frequencies as well as the increase in UP-state and burst durations are found in young AD and old WT mice. We show evidence that the aging process is accelerated by the mutant PS2 itself and discuss such changes in relation to amyloidosis and gliosis.

Simultaneous Two-Photon Voltage or Calcium Imaging and Multi-Channel Local Field Potential Recordings in Barrel Cortex of Awake and Anesthetized Mice

Neuronal population activity, both spontaneous and sensory-evoked, generates propagating waves in cortex. However, high spatiotemporal-resolution mapping of these waves is difficult as calcium imaging, the work horse of current imaging, does not reveal subthreshold activity. Here, we present a platform combining voltage or calcium two-photon imaging with multi-channel local field potential (LFP) recordings in different layers of the barrel cortex from anesthetized and awake head-restrained mice. A chronic cranial window with access port allows injecting a viral vector expressing GCaMP6f or the voltage-sensitive dye (VSD) ANNINE-6plus, as well as entering the brain with a multi-channel neural probe. We present both average spontaneous activity and average evoked signals in response to multi-whisker air-puff stimulations. Time domain analysis shows the dependence of the evoked responses on the cortical layer and on the state of the animal, here separated into anesthetized, awake but resting, and running. The simultaneous data acquisition allows to compare the average membrane depolarization measured with ANNINE-6plus with the amplitude and shape of the LFP recordings. The calcium imaging data connects these data sets to the large existing database of this important second messenger. Interestingly, in the calcium imaging data, we found a few cells which showed a decrease in calcium concentration in response to vibrissa stimulation in awake mice. This system offers a multimodal technique to study the spatiotemporal dynamics of neuronal signals through a 3D architecture in vivo. It will provide novel insights on sensory coding, closing the gap between electrical and optical recordings.

Calcium Signaling and Mitochondrial Function in Presenilin 2 Knock-Out Mice: Looking for Any Loss-of-Function Phenotype Related to Alzheimer's Disease

Alzheimer's disease (AD) is the most common age-related neurodegenerative disorder in which learning, memory and cognitive functions decline progressively. Familial forms of AD (FAD) are caused by mutations in amyloid precursor protein (APP), presenilin 1 (PSEN1) and presenilin 2 (PSEN2) genes. Presenilin 1 (PS1) and its homologue, presenilin 2 (PS2), represent, alternatively, the catalytic core of the γ-secretase complex that, by cleaving APP, produces neurotoxic amyloid beta (Aβ) peptides responsible for one of the histopathological hallmarks in AD brains, the amyloid plaques. Recently, PSEN1 FAD mutations have been associated with a loss-of-function phenotype. To investigate whether this finding can also be extended to PSEN2 FAD mutations, we studied two processes known to be modulated by PS2 and altered by FAD mutations: Ca2+ signaling and mitochondrial function. By exploiting neurons derived from a PSEN2 knock-out (PS2-/-) mouse model, we found that, upon IP3-generating stimulation, cytosolic Ca2+ handling is not altered, compared to wild-type cells, while mitochondrial Ca2+ uptake is strongly compromised. Accordingly, PS2-/- neurons show a marked reduction in endoplasmic reticulum-mitochondria apposition and a slight alteration in mitochondrial respiration, whereas mitochondrial membrane potential, and organelle morphology and number appear unchanged. Thus, although some alterations in mitochondrial function appear to be shared between PS2-/- and FAD-PS2-expressing neurons, the mechanisms leading to these defects are quite distinct between the two models. Taken together, our data appear to be difficult to reconcile with the proposal that FAD-PS2 mutants are loss-of-function, whereas the concept that PS2 plays a key role in sustaining mitochondrial function is here confirmed.

Application of deep learning in detecting neurological disorders from magnetic resonance images: a survey on the detection of Alzheimer's disease, Parkinson's disease and schizophrenia

Neuroimaging, in particular magnetic resonance imaging (MRI), has been playing an important role in understanding brain functionalities and its disorders during the last couple of decades. These cutting-edge MRI scans, supported by high-performance computational tools and novel ML techniques, have opened up possibilities to unprecedentedly identify neurological disorders. However, similarities in disease phenotypes make it very difficult to detect such disorders accurately from the acquired neuroimaging data. This article critically examines and compares performances of the existing deep learning (DL)-based methods to detect neurological disorders-focusing on Alzheimer's disease, Parkinson's disease and schizophrenia-from MRI data acquired using different modalities including functional and structural MRI. The comparative performance analysis of various DL architectures across different disorders and imaging modalities suggests that the Convolutional Neural Network outperforms other methods in detecting neurological disorders. Towards the end, a number of current research challenges are indicated and some possible future research directions are provided.

Presenilin-2 and Calcium Handling: Molecules, Organelles, Cells and Brain Networks

Presenilin-2 (PS2) is one of the three proteins that are dominantly mutated in familial Alzheimer's disease (FAD). It forms the catalytic core of the γ-secretase complex-a function shared with its homolog presenilin-1 (PS1)-the enzyme ultimately responsible of amyloid-β (Aβ) formation. Besides its enzymatic activity, PS2 is a multifunctional protein, being specifically involved, independently of γ-secretase activity, in the modulation of several cellular processes, such as Ca2+ signalling, mitochondrial function, inter-organelle communication, and autophagy. As for the former, evidence has accumulated that supports the involvement of PS2 at different levels, ranging from organelle Ca2+ handling to Ca2+ entry through plasma membrane channels. Thus FAD-linked PS2 mutations impact on multiple aspects of cell and tissue physiology, including bioenergetics and brain network excitability. In this contribution, we summarize the main findings on PS2, primarily as a modulator of Ca2+ homeostasis, with particular emphasis on the role of its mutations in the pathogenesis of FAD. Identification of cell pathways and molecules that are specifically targeted by PS2 mutants, as well as of common targets shared with PS1 mutants, will be fundamental to disentangle the complexity of memory loss and brain degeneration that occurs in Alzheimer's disease (AD).

Intracellular Calcium Dysregulation by the Alzheimer's Disease-Linked Protein Presenilin 2

Alzheimer's disease (AD) is the most common form of dementia. Even though most AD cases are sporadic, a small percentage is familial due to autosomal dominant mutations in amyloid precursor protein (APP), presenilin-1 (PSEN1), and presenilin-2 (PSEN2) genes. AD mutations contribute to the generation of toxic amyloid β (Aβ) peptides and the formation of cerebral plaques, leading to the formulation of the amyloid cascade hypothesis for AD pathogenesis. Many drugs have been developed to inhibit this pathway but all these approaches currently failed, raising the need to find additional pathogenic mechanisms. Alterations in cellular calcium (Ca2+) signaling have also been reported as causative of neurodegeneration. Interestingly, Aβ peptides, mutated presenilin-1 (PS1), and presenilin-2 (PS2) variously lead to modifications in Ca2+ homeostasis. In this contribution, we focus on PS2, summarizing how AD-linked PS2 mutants alter multiple Ca2+ pathways and the functional consequences of this Ca2+ dysregulation in AD pathogenesis.