Brain Informatics最新文献

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that primarily affects memory, thinking, and behavior, leading to dementia, a severe cognitive decline. While no cure currently exists, recent advancements in preventive drug trials and therapeutic management have increased interest in developing clinical algorithms for early detection and biomarker identification. Electroencephalography (EEG) is noninvasive, cost-effective, and has high temporal resolution, making it a promising tool for automated AD detection. However, conventional machine learning approaches often fall short in accurately detecting AD due to their limited architectures. We also need to investigate the impact of EEG signal segment length on classification accuracy. To address these issues, a deep learning-based framework is proposed to detect AD using EEG data, focusing on determining the optimal segment length for classification. This framework contains EEG data collection, pre-processing for noise removal, temporal segmentation, convolutional neural network (CNN) model training and classification, and finally, evaluation. We have tested different segment lengths to test the impact on AD detection. We have used both 10-fold and leave-one-out cross-validation techniques and obtained accuracy of 97.08% and 96.90%, respectively, on a publicly available dataset from AHEPA General University Hospital of Thessaloniki. We have also tested the generalizability of the proposed model by testing it to detect frontotemporal dementia and obtained better results than existing studies. Furthermore, we have validated our proposed CNN model using several ablation studies and layer-wise extracted feature visualization. This study will establish a pioneering direction for future researchers and technology experts in the field of neurodiseases.

In real-world psychiatric practice, patients may experience complex treatment journeys, including various diagnoses and lines of therapy. Insurance claims databases could potentially provide insight into outcomes of psychiatric treatment processes, but the diversity of event sequences restricts analyses with currently available methods. Here, we developed a novel kernel k-means clustering algorithm for event sequences that can accommodate highly diverse event types and sequence lengths. The approach, Divisive Optimized Clustering using Kernel K-means for Event Sequences (DOCKKES), also leverages a novel performance metric, the transition score, which measures sequence coherence in individual clusters. The performance of DOCKKES was evaluated in the context of bipolar I disorder, which is characterized by heterogeneous treatment journeys. We conducted a retrospective, observational analysis of a large sample (n = 31,578) of patients with bipolar I disorder from the MarketScan® Commercial Database. Using insurance claims, bipolar episode diagnoses and mental health-related lines of therapy were identified as events of interest for patient clustering. The dataset included 202,122 events; 75% of the cohort experienced unique treatment journeys. Based on an optimal run, DOCKKES identified 16 treatment journey clusters, which were evenly split for initial manic/mixed or depressive episodes (8 clusters each) and varied in sequence length and early lines of therapy. Variability across clusters was also observed for demographics, comorbidities, and mental health-related healthcare resource utilization and cost. This proof-of-concept study demonstrated the use of DOCKKES for integrating information from large datasets, enabling comparisons between patient clusters and evaluation of real-world treatment journeys in the context of evidence-based guidelines.

Recent advancements in cloud-based machine learning (ML) now allow for the rapid and remote identification of neural-biomarkers associated with common neuro-developmental disorders from neuroimaging datasets. Due to the sensitive nature of these datasets, secure deep learning (DL) algorithms are essential. Although, fully homomorphic encryption (FHE)-based methods have been proposed to maintain data confidentiality and privacy, however, existing FHE deep convolutional neural network (CNN) models still face some issues such as low accuracy, high encryption/decryption latency, energy inefficiency, long feature extraction times, and significant cipher-image expansion. To address these issues, this study introduces the HoRNS-CNN model, which integrates the energy-efficient features of the residue number system FHE scheme (RNS-FHE scheme) with the high accuracy of pre-trained deep CNN models in the cloud for efficient, privacy-preserving predictions and provide some proofs of its energy efficiency and homomorphism. The RNS-FHE scheme's FPGA implementation includes embedded RNS pixel-bitstream homomorphic encoder/decoder circuits for encrypting 8-bit grayscale pixels, with cloud CNN models performing remote classification on the encrypted images. In the HoRNS-CNN architecture, the ReLU activation functions of deep CNNs were initially trained for stability and later adapted for homomorphic computations using a Taylor polynomial approximation of degree 3 and batch normalization to achieve high accuracy. The findings show that the HoRNS-CNN model effectively manages cipher-image expansion with an asymptotic complexity of $O\left(n^3\right)$ , offering better performance and faster feature extraction compared to its peers. The model can predict 400,000 neural-biomarker features in one hour, providing an effective tool for analyzing neuroimages while ensuring privacy and security.

Despite significant advancements in brain tumor classification, many existing models suffer from complex structures that make them difficult to interpret. This complexity can hinder the transparency of the decision-making process, causing models to rely on irrelevant features or normal soft tissues. Besides, these models often include additional layers and parameters, which further complicate the classification process. Our work addresses these limitations by introducing a novel methodology that combines Explainable AI (XAI) techniques with a Convolutional Neural Network (CNN) architecture. The major contribution of this paper is ensuring that the model focuses on the most relevant features for tumor detection and classification, while simultaneously reducing complexity, by minimizing the number of layers. This approach enhances the model's transparency and robustness, giving clear insights into its decision-making process through XAI techniques such as Gradient-weighted Class Activation Mapping (Grad-Cam), Shapley Additive explanations (Shap), and Local Interpretable Model-agnostic Explanations (LIME). Additionally, the approach demonstrates better performance, achieving 99% accuracy on seen data and 95% on unseen data, highlighting its generalizability and reliability. This balance of simplicity, interpretability, and high accuracy represents a significant advancement in the classification of brain tumor.

Bipolar psychometric scale data play a crucial role in psychological healthcare and health economics, such as in psychotherapeutic profiling and setting standards. Creating an accurate psychological profile not only benefits the patient but also saves time and costs. The quality of psychotherapeutic measures directly impacts grant funding decisions, influencing managerial choices. Moreover, the accuracy of consumer data analyses affects costs, profits, and the long-term sustainability of decisions. Considering psychometric bipolar scale data as compositional data can enhance the statistical power of well-known paired and unpaired two-sample t-tests, supporting managerial decision-making and the development or implementation of health interventions. This increase in statistical power is observed when the central limit theorem (CLT) holds true in statistics. Through stochastic simulation, this study explores the impact of violating the CLT on statistical power of the unpaired t-test under heavy-tailed data generating processes (DGPs) with finite variance. The findings reveal a reduction in statistical power based on specific parameters like the psychometric limit of quantification, the number of items in a questionnaire, the response scale used, and the dispersion of the DGP.

Clinical diagnosis of Alzheimer's disease (AD) is usually made after symptoms such as short-term memory loss are exhibited, which minimizes the intervention and treatment options. The existing screening techniques cannot distinguish between stable MCI (sMCI) cases (i.e., patients who do not convert to AD for at least three years) and progressive MCI (pMCI) cases (i.e., patients who convert to AD in three years or sooner). Delayed diagnosis of AD also disproportionately affects underrepresented and socioeconomically disadvantaged populations. The significant positive impact of an early diagnosis solution for AD across diverse ethno-racial and demographic groups is well-known and recognized. While advancements in high-throughput technologies have enabled the generation of vast amounts of multimodal clinical, and neuroimaging datasets related to AD, most methods utilizing these data sets for diagnostic purposes have not found their way in clinical settings. To better understand the landscape, we surveyed the major preprocessing, data management, traditional machine-learning (ML), and deep learning (DL) techniques used for diagnosing AD using neuroimaging data such as structural magnetic resonance imaging (sMRI), functional magnetic resonance imaging (fMRI), and positron emission tomography (PET). Once we had a good understanding of the methods available, we conducted a study to assess the reproducibility and generalizability of open-source ML models. Our evaluation shows that existing models show reduced generalizability when different cohorts of the data modality are used while controlling other computational factors. The paper concludes with a discussion of major challenges that plague ML models for AD diagnosis and biomarker discovery.

Class incremental learning (CIL) is a specific scenario in incremental learning. It aims to continuously learn new classes from the data stream, which suffers from the challenge of catastrophic forgetting. Inspired by the human hippocampus, the CIL method for replaying episodic memory offers a promising solution. However, the limited buffer budget restricts the number of old class samples that can be stored, resulting in an imbalance between new and old class samples during each incremental learning stage. This imbalance adversely affects the mitigation of catastrophic forgetting. Therefore, we propose a novel CIL method based on multi-granularity balance strategy (MGBCIL), which is inspired by the three-way granular computing in human problem-solving. In order to mitigate the adverse effects of imbalances on catastrophic forgetting at fine-, medium-, and coarse-grained levels during training, MGBCIL introduces specific strategies across the batch, task, and decision stages. Specifically, a weighted cross-entropy loss function with a smoothing factor is proposed for batch processing. In the process of task updating and classification decision, contrastive learning with different anchor point settings is employed to promote local and global separation between new and old classes. Additionally, the knowledge distillation technology is used to preserve knowledge of the old classes. Experimental evaluations on CIFAR-10 and CIFAR-100 datasets show that MGBCIL outperforms other methods in most incremental settings. Specifically, when storing 3 exemplars on CIFAR-10 with Base2 Inc2 setting, the average accuracy is improved by up to 9.59% and the forgetting rate is reduced by up to 25.45%.

Epilepsy is one of the most common clinical diseases, which is caused by abnormal discharge of brain nerves. Around 30% of patients will develop drug-resistant epilepsy that are hard to be cured by anti-epileptic drug treatment. This patient cohort are ideal candidate for surgical resection of the epileptic focus. For safety and maximum effective rate, the key to success of the operation is to identify the focus area and normal functional area accurately in the preoperative evaluation stage. Intracranial EEG (iEEG) has attracted much attention for its precise capture of the state of rapid brain activity and its strong locality. To automate the process of iEEG inspection and surgical evaluation, this paper propose a Gated Recurrent Unit-Granger Causality (GRU-GC) algorithm to detect effective connectivity between channels and construct a directed graph. From six local features, the top five feature combinations were selected to differentiate between epileptic foci and non-epileptic regions. Experiments indicate that these features are most discriminative during the ictal phase, yielding superior classification accuracy. Compared to traditional time-series-based methods, this study shows that GRU-GC algorithm is efficient in building effective graph model for improving preoperative epilepsy evaluations.

Human brain signal processing and finger's movement coordination is a complex mechanism. In this mechanism finger's movement is mostly performed for every day's task. It is well known that to capture such movement EEG or ECoG signals are used. In this order to find the patterns from these signals is important. The BCI competition 4 dataset 4 is one such standard dataset of ECoG signals for individual finger movement provided by University of Washington, USA. In this work, this dataset is, statistically analyzed to understand the nature of data and outliers in it. Effectiveness of pre-processing algorithm is then visualized. The cleaned dataset has dual polarity and gaussian distribution nature which makes Tanh activation function suitable for the neural network BC4D4 model. BC4D4 uses Convolutional neural network for feature extraction, dense neural network for pattern identification and incorporating dropout & regularization making the proposed model more resilient. Our model outperforms the state of the art work on the dataset 4 achieving 0.85 correlation value that is 1.85X (Winner of BCI competition 4, 2012) & 1.25X (Finger Flex model, 2022).

This study focuses on classifying multiple sessions of loving kindness meditation (LKM) and non-meditation electroencephalography (EEG) data. This novel study focuses on using multiple sessions of EEG data from a single individual to train a machine learning pipeline, and then using a new session data from the same individual for the classification. Here, two meditation techniques, LKM-Self and LKM-Others were compared with non-meditation EEG data for 12 participants. Among many tested, three BCI pipelines we built produced promising results, successfully detecting features in meditation/ non-meditation EEG data. While testing different feature extraction algorithms, a common neural network structure was used as the classification algorithm to compare the performance of the feature extraction algorithms. For two of those pipelines, Common Spatial Patterns (CSP) and Short Time Fourier Transform (STFT) were successfully used as feature extraction algorithms where both these algorithms are significantly new for meditation EEG. As a novel concept, the third BCI pipeline used a feature extraction algorithm that fused the features of CSP and STFT, achieving the highest classification accuracies among all tested pipelines. Analyses were conducted using EEG data of 3, 4 or 5 sessions, totaling 3960 tests on the entire dataset. At the end of the study, when considering all the tests, the overall classification accuracy using SCP alone was 67.1%, and it was 67.8% for STFT alone. The algorithm combining the features of CSP and STFT achieved an overall classification accuracy of 72.9% which is more than 5% higher than the other two pipelines. At the same time, the highest mean classification accuracy for the 12 participants was achieved using the pipeline with the combination of CSP STFT algorithm, reaching 75.5% for LKM-Self/ non-meditation for the case of 5 sessions of data. Additionally, the highest individual classification accuracy of 88.9% was obtained by the participant no. 14. Furthermore, the results showed that the classification accuracies for all three pipelines increased with the number of training sessions increased from 2 to 3 and then to 4. The study was successful in classifying a new session of EEG meditation/ non-meditation data after training machine learning algorithms using a different set of session data, and this achievement will be beneficial in the development of algorithms that support meditation.