An explainable Artificial Intelligence approach to study MCI to AD conversion via HD-EEG processing

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.

MSE-VGG: A Novel Deep Learning Approach Based on EEG for Rapid Ischemic Stroke Detection

Ischemic stroke is a type of brain dysfunction caused by pathological changes in the blood vessels of the brain which leads to brain tissue ischemia and hypoxia and ultimately results in cell necrosis. Without timely and effective treatment in the early time window, ischemic stroke can lead to long-term disability and even death. Therefore, rapid detection is crucial in patients with ischemic stroke. In this study, we developed a deep learning model based on fusion features extracted from electroencephalography (EEG) signals for the fast detection of ischemic stroke. Specifically, we recruited 20 ischemic stroke patients who underwent EEG examination during the acute phase of stroke and collected EEG signals from 19 adults with no history of stroke as a control group. Afterwards, we constructed correlation-weighted Phase Lag Index (cwPLI), a novel feature, to explore the synchronization information and functional connectivity between EEG channels. Moreover, the spatio-temporal information from functional connectivity and the nonlinear information from complexity were fused by combining the cwPLI matrix and Sample Entropy (SaEn) together to further improve the discriminative ability of the model. Finally, the novel MSE-VGG network was employed as a classifier to distinguish ischemic stroke from non-ischemic stroke data. Five-fold cross-validation experiments demonstrated that the proposed model possesses excellent performance, with accuracy, sensitivity, and specificity reaching 90.17%, 89.86%, and 90.44%, respectively. Experiments on time consumption verified that the proposed method is superior to other state-of-the-art examinations. This study contributes to the advancement of the rapid detection of ischemic stroke, shedding light on the untapped potential of EEG and demonstrating the efficacy of deep learning in ischemic stroke identification.

SHAP value-based ERP analysis (SHERPA): Increasing the sensitivity of EEG signals with explainable AI methods

Conventionally, event-related potential (ERP) analysis relies on the researcher to identify the sensors and time points where an effect is expected. However, this approach is prone to bias and may limit the ability to detect unexpected effects or to investigate the full range of the electroencephalography (EEG) signal. Data-driven approaches circumvent this limitation, however, the multiple comparison problem and the statistical correction thereof affect both the sensitivity and specificity of the analysis. In this study, we present SHERPA - a novel approach based on explainable artificial intelligence (XAI) designed to provide the researcher with a straightforward and objective method to find relevant latency ranges and electrodes. SHERPA is comprised of a convolutional neural network (CNN) for classifying the conditions of the experiment and SHapley Additive exPlanations (SHAP) as a post hoc explainer to identify the important temporal and spatial features. A classical EEG face perception experiment is employed to validate the approach by comparing it to the established researcher- and data-driven approaches. Likewise, SHERPA identified an occipital cluster close to the temporal coordinates for the N170 effect expected. Most importantly, SHERPA allows quantifying the relevance of an ERP for a psychological mechanism by calculating an "importance score". Hence, SHERPA suggests the presence of a negative selection process at the early and later stages of processing. In conclusion, our new method not only offers an analysis approach suitable in situations with limited prior knowledge of the effect in question but also an increased sensitivity capable of distinguishing neural processes with high precision.

Editorial Topical Collection: "Explainable and Augmented Machine Learning for Biosignals and Biomedical Images"

Machine learning (ML) is a well-known subfield of artificial intelligence (AI) that aims at developing algorithms and statistical models able to empower computer systems to automatically adapt to a specific task through experience or learning from data [...].

An explainable deep-learning model to stage sleep states in children and propose novel EEG-related patterns in sleep apnea

Automatic deep-learning models used for sleep scoring in children with obstructive sleep apnea (OSA) are perceived as black boxes, limiting their implementation in clinical settings. Accordingly, we aimed to develop an accurate and interpretable deep-learning model for sleep staging in children using single-channel electroencephalogram (EEG) recordings. We used EEG signals from the Childhood Adenotonsillectomy Trial (CHAT) dataset (n = 1637) and a clinical sleep database (n = 980). Three distinct deep-learning architectures were explored to automatically classify sleep stages from a single-channel EEG data. Gradient-weighted Class Activation Mapping (Grad-CAM), an explainable artificial intelligence (XAI) algorithm, was then applied to provide an interpretation of the singular EEG patterns contributing to each predicted sleep stage. Among the tested architectures, a standard convolutional neural network (CNN) demonstrated the highest performance for automated sleep stage detection in the CHAT test set (accuracy = 86.9% and five-class kappa = 0.827). Furthermore, the CNN-based estimation of total sleep time exhibited strong agreement in the clinical dataset (intra-class correlation coefficient = 0.772). Our XAI approach using Grad-CAM effectively highlighted the EEG features associated with each sleep stage, emphasizing their influence on the CNN's decision-making process in both datasets. Grad-CAM heatmaps also allowed to identify and analyze epochs within a recording with a highly likelihood to be misclassified, revealing mixed features from different sleep stages within these epochs. Finally, Grad-CAM heatmaps unveiled novel features contributing to sleep scoring using a single EEG channel. Consequently, integrating an explainable CNN-based deep-learning model in the clinical environment could enable automatic sleep staging in pediatric sleep apnea tests.

Decoding movement kinematics from EEG using an interpretable convolutional neural network

Continuous decoding of hand kinematics has been recently explored for the intuitive control of electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs). Deep neural networks (DNNs) are emerging as powerful decoders, for their ability to automatically learn features from lightly pre-processed signals. However, DNNs for kinematics decoding lack in the interpretability of the learned features and are only used to realize within-subject decoders without testing other training approaches potentially beneficial for reducing calibration time, such as transfer learning. Here, we aim to overcome these limitations by using an interpretable convolutional neural network (ICNN) to decode 2-D hand kinematics (position and velocity) from EEG in a pursuit tracking task performed by 13 participants. The ICNN is trained using both within-subject and cross-subject strategies, and also testing the feasibility of transferring the knowledge learned on other subjects on a new one. Moreover, the network eases the interpretation of learned spectral and spatial EEG features. Our ICNN outperformed most of the other state-of-the-art decoders, showing the best trade-off between performance, size, and training time. Furthermore, transfer learning improved kinematics prediction in the low data regime. The network attributed the highest relevance for decoding to the delta-band across all subjects, and to higher frequencies (alpha, beta, low-gamma) for a cluster of them; contralateral central and parieto-occipital sites were the most relevant, reflecting the involvement of sensorimotor, visual and visuo-motor processing. The approach improved the quality of kinematics prediction from the EEG, at the same time allowing interpretation of the most relevant spectral and spatial features.

An eXplainability Artificial Intelligence approach to brain connectivity in Alzheimer's disease

The advent of eXplainable Artificial Intelligence (XAI) has revolutionized the way human experts, especially from non-computational domains, approach artificial intelligence; this is particularly true for clinical applications where the transparency of the results is often compromised by the algorithmic complexity. Here, we investigate how Alzheimer's disease (AD) affects brain connectivity within a cohort of 432 subjects whose T1 brain Magnetic Resonance Imaging data (MRI) were acquired within the Alzheimer's Disease Neuroimaging Initiative (ADNI). In particular, the cohort included 92 patients with AD, 126 normal controls (NC) and 214 subjects with mild cognitive impairment (MCI). We show how graph theory-based models can accurately distinguish these clinical conditions and how Shapley values, borrowed from game theory, can be adopted to make these models intelligible and easy to interpret. Explainability analyses outline the role played by regions like putamen, middle and superior temporal gyrus; from a class-related perspective, it is possible to outline specific regions, such as hippocampus and amygdala for AD and posterior cingulate and precuneus for MCI. The approach is general and could be adopted to outline how brain connectivity affects specific brain regions.

From local counterfactuals to global feature importance: efficient, robust, and model-agnostic explanations for brain connectivity networks

BACKGROUND

Explainable artificial intelligence (XAI) is a technology that can enhance trust in mental state classifications by providing explanations for the reasoning behind artificial intelligence (AI) models outputs, especially for high-dimensional and highly-correlated brain signals. Feature importance and counterfactual explanations are two common approaches to generate these explanations, but both have drawbacks. While feature importance methods, such as shapley additive explanations (SHAP), can be computationally expensive and sensitive to feature correlation, counterfactual explanations only explain a single outcome instead of the entire model.

METHODS

To overcome these limitations, we propose a new procedure for computing global feature importance that involves aggregating local counterfactual explanations. This approach is specifically tailored to fMRI signals and is based on the hypothesis that instances close to the decision boundary and their counterfactuals mainly differ in the features identified as most important for the downstream classification task. We refer to this proposed feature importance measure as Boundary Crossing Solo Ratio (BoCSoR), since it quantifies the frequency with which a change in each feature in isolation leads to a change in classification outcome, i.e., the crossing of the model's decision boundary.

RESULTS AND CONCLUSIONS

Experimental results on synthetic data and real publicly available fMRI data from the Human Connect project show that the proposed BoCSoR measure is more robust to feature correlation and less computationally expensive than state-of-the-art methods. Additionally, it is equally effective in providing an explanation for the behavior of any AI model for brain signals. These properties are crucial for medical decision support systems, where many different features are often extracted from the same physiological measures and a gold standard is absent. Consequently, computing feature importance may become computationally expensive, and there may be a high probability of mutual correlation among features, leading to unreliable results from state-of-the-art XAI methods.

A self-driven approach for multi-class discrimination in Alzheimer's disease based on wearable EEG

Early detection is critical to control Alzheimer's disease (AD) progression and postpone cognitive decline. Traditional medical procedures such as magnetic resonance imaging are costly, involve long waiting lists, and require complex analysis. Alternatively, for the past years, researchers have successfully evaluated AD detection approaches based on machine learning and electroencephalography (EEG). Nonetheless, these approaches frequently rely upon manual processing or involve non-portable EEG hardware. These aspects are suboptimal regarding automated diagnosis, since they require additional personnel and hinder portability. In this work, we report the preliminary evaluation of a self-driven AD multi-class discrimination approach based on a commercial EEG acquisition system using sixteen channels. For this purpose, we recorded the EEG of three groups of participants: mild AD, mild cognitive impairment (MCI) non-AD, and controls, and we implemented a self-driven analysis pipeline to discriminate the three groups. First, we applied automated artifact rejection algorithms to the EEG recordings. Then, we extracted power, entropy, and complexity features from the preprocessed epochs. Finally, we evaluated a multi-class classification problem using a multi-layer perceptron through leave-one-subject-out cross-validation. The preliminary results that we obtained are comparable to the best in literature (0.88 F1-score), what suggests that AD can potentially be detected through a self-driven approach based on commercial EEG and machine learning. We believe this work and further research could contribute to opening the door for the detection of AD in a single consultation session, therefore reducing the costs associated to AD screening and potentially advancing medical treatment.