Search for: All records

Objective: This study develops new machine learning architectures that are more adept at detecting interictal epileptiform discharges (IEDs) in scalp EEG. A comparison of results using the average precision (AP) metric is made with the proposed models on two datasets obtained from Baptist Hospital of Miami and Temple University Hospital. Methods: Applying graph neural networks (GNNs) on functional connectivity (FC) maps of different frequency sub-bands to yield a novel architecture we call FC-GNN. Attention mechanism is applied on a complete graph to let the neural network select its important edges, hence bypassing the extraction of features, a model we refer to as CA-GNN. Results: On the Baptist Hospital dataset, the results were as follows: Vanilla Self-Attention → 0.9029 ± 0.0431, Hierarchical Attention → 0.8546 ± 0.0587, Vanilla Visual Geometry Group (VGG) → 0.92 ± 0.0618, Satelight → 0.9219 ± 0.046, FC-GNN → 0.9731 ± 0.0187, and CA-GNN → 0.9788 ± 0.0125. In the same order, the results on the Temple University Hospital dataset are 0.9692, 0.9113, 0.97, 0.9575, 0.963, and 0.9879. Conclusion: Based on the good results they yield, GNNs prove to have a strong potential in detecting epileptogenic activity. Significance: This study opens the door for the discovery of the powerful role played by GNNs in capturing IEDs, which is an essential step for identifying the epileptogenic networks of the affected brain and hence improving the prospects for more accurate 3D source localization. 

Pediatric epilepsy due to drug-resistant Focal Cortical Dysplasia (FCD) presents significant healthcare challenges. Precise preoperative identification of FCD lesions is imperative for surgical planning and patient outcomes. This paper presents a proof-of-concept for an integrated methodology that combines Electroencephalogram (EEG)-based functional connectivity analysis with Magnetic Resonance Imaging (MRI)-derived cortical thickness measurements to identify FCD lesions in pediatric epileptic patients. We examined a single-case clinical scenario from Oregon Health Science and University, consistently identifying the Caudal Middle Frontal (cMFG) region across both EEG and MRI modalities, a finding that was confirmed in the postoperative MRI scan. This cross-validation underscores the potential of the precision of our approach in pinpointing the surgical target region. Despite being constrained by its preliminary nature, our research offers a valuable foundation for a personalized, rigorous method of detecting the location of the FCD lesions. It holds significant clinical implications for managing FCD-related epilepsy. It also portends broader applications in neurology and precision medicine. Nonetheless, further large-scale studies are needed to validate and fine-tune our methodology. Clinical Relevance - This study offers clinicians an advanced, integrated approach to preoperative assessment of FCD lesions, potentially improving the precision of surgical planning in pediatric epilepsy. 

Epilepsy is a brain disorder that causes seizures, affecting nearly half a million children in the US alone. In this study, we aimed to use a nonlinear driven method to characterize scalp EEG recordings of pediatric epilepsy patients (PE: n=7) compared to pediatric control subjects (PC: n=7) in a clinical environment. A time-varying approach was used to construct functional connectivity networks (FCNs) of all subjects. Next, the FCNs are mapped into the form of undirected graphs that are subjected to the extraction of graph theory-based features. An unsupervised clustering technique based on K-mean is used to delineate the PE from the PC group. Our findings show a statistically significant difference in the mean FCNs between PC and PE groups (t(340)=- 15.9899, p<< 0.0001). Performance results showed an accuracy of 92.5% with a sensitivity of 90% and a specificity of 95.3%. This approach can help improve and validate the early diagnosis of PE by applying non-invasive scalp EEG signals. 

Early detection of Alzheimer’s disease (AD) during the Mild Cognitive Impairment (MCI) stage could enable effective intervention to slow down disease progression. Computer-aided diagnosis of AD relies on a sufficient amount of biomarker data. When this requirement is not fulfilled, transfer learning can be used to transfer knowledge from a source domain with more amount of labeled data than available in the desired target domain. In this study, an instance-based transfer learning framework is presented based on the gradient boosting machine (GBM). In GBM, a sequence of base learners is built, and each learner focuses on the errors (residuals) of the previous learner. In our transfer learning version of GBM (TrGB), a weighting mechanism based on the residuals of the base learners is defined for the source instances. Consequently, instances with different distribution than the target data will have a lower impact on the target learner. The proposed weighting scheme aims to transfer as much information as possible from the source domain while avoiding negative transfer. The target data in this study was obtained from the Mount Sinai dataset which is collected and processed in a collaborative 5-year project at the Mount Sinai Medical Center. The Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset was used as the source domain. The experimental results showed that the proposed TrGB algorithm could improve the classification accuracy by 1.5 and 4.5% for CN vs. MCI and multiclass classification, respectively, as compared to the conventional methods. Also, using the TrGB model and transferred knowledge from the CN vs. AD classification of the source domain, the average score of early MCI vs. late MCI classification improved by 5%. 

Abstract Alzheimer’s disease (AD) is a neurogenerative condition characterized by sharp cognitive decline with no confirmed effective treatment or cure. This makes it critically important to identify the symptoms of Alzheimer’s disease in its early stages before significant cognitive deterioration has taken hold and even before any brain morphology and neuropathology are noticeable. In this study, five different multimodal deep neural networks (MDNN), with different architectures, in search of an optimal model for predicting the cognitive test scores for the Mini-Mental State Examination (MMSE) and the modified Alzheimer’s Disease Assessment Scale (ADAS-CoG13) over a span of 60 months (5 years). The multimodal data utilized to train and test the proposed models were obtained from the Alzheimer’s Disease Neuroimaging Initiative study and includes cerebrospinal fluid (CSF) levels of tau and beta-amyloid, structural measures from magnetic resonance imaging (MRI), functional and metabolic measures from positron emission tomography (PET), and cognitive scores from the neuropsychological tests (Cog). The models developed herein delve into two main issues: (1) application merits of single-task vs. multitask for predicting future cognitive scores and (2) whether time-varying input data are better suited than specific timepoints for optimizing prediction results. This model yields a high of 90.27% (SD = 1.36) prediction accuracy (correlation) at 6 months after the initial visit to a lower 79.91% (SD = 8.84) prediction accuracy at 60 months. The analysis provided is comprehensive as it determines the predictions at all other timepoints and all MDNN models include converters in the CN and MCI groups (CNc, MCIc) and all the unstable groups in the CN and MCI groups (CNun and MCIun) that reverted to CN from MCI and to MCI from AD, so as not to bias the results. The results show that the best performance is achieved by a multimodal combined single-task long short-term memory (LSTM) regressor with an input sequence length of 2 data points (2 visits, 6 months apart) augmented with a pretrained Neural Network Estimator to fill in for the missing values. 

With the advances in machine learning for the diagnosis of Alzheimer’s disease (AD), most studies have focused on either identifying the subject’s status through classification algorithms or on predicting their cognitive scores through regression methods, neglecting the potential association between these two tasks. Motivated by the need to enhance the prospects for early diagnosis along with the ability to predict future disease states, this study proposes a deep neural network based on modality fusion, kernelization, and tensorization that perform multiclass classification and longitudinal regression simultaneously within a unified multitask framework. This relationship between multiclass classification and longitudinal regression is found to boost the efficacy of the final model in dealing with both tasks. Different multimodality scenarios are investigated, and complementary aspects of the multimodal features are exploited to simultaneously delineate the subject’s label and predict related cognitive scores at future timepoints using baseline data. The main intent in this multitask framework is to consolidate the highest accuracy possible in terms of precision, sensitivity, F1 score, and area under the curve (AUC) in the multiclass classification task while maintaining the highest similarity in the MMSE score as measured through the correlation coefficient and the RMSE for all time points under the prediction task, with both tasks, run simultaneously under the same set of hyperparameters. The overall accuracy for multiclass classification of the proposed KTMnet method is 66.85 ± 3.77. The prediction results show an average RMSE of 2.32 ± 0.52 and a correlation of 0.71 ± 5.98 for predicting MMSE throughout the time points. These results are compared to state-of-the-art techniques reported in the literature. A discovery from the multitasking of this consolidated machine learning framework is that a set of hyperparameters that optimize the prediction results may not necessarily be the same as those that would optimize the multiclass classification. In other words, there is a breakpoint beyond which enhancing further the results of one process could lead to the downgrading in accuracy for the other. 

Background: Machine learning is a promising tool for biomarker-based diagnosis of Alzheimer’s disease (AD). Performing multimodal feature selection and studying the interaction between biological and clinical AD can help to improve the performance of the diagnosis models. Objective: This study aims to formulate a feature ranking metric based on the mutual information index to assess the relevance and redundancy of regional biomarkers and improve the AD classification accuracy. Methods: From the Alzheimer’s Disease Neuroimaging Initiative (ADNI), 722 participants with three modalities, including florbetapir-PET, flortaucipir-PET, and MRI, were studied. The multivariate mutual information metric was utilized to capture the redundancy and complementarity of the predictors and develop a feature ranking approach. This was followed by evaluating the capability of single-modal and multimodal biomarkers in predicting the cognitive stage. Results: Although amyloid-β deposition is an earlier event in the disease trajectory, tau PET with feature selection yielded a higher early-stage classification F1-score (65.4%) compared to amyloid-β PET (63.3%) and MRI (63.2%). The SVC multimodal scenario with feature selection improved the F1-score to 70.0% and 71.8% for the early and late-stage, respectively. When age and risk factors were included, the scores improved by 2 to 4%. The Amyloid-Tau-Neurodegeneration [AT(N)] framework helped to interpret the classification results for different biomarker categories. Conclusion: The results underscore the utility of a novel feature selection approach to reduce the dimensionality of multimodal datasets and enhance model performance. The AT(N) biomarker framework can help to explore the misclassified cases by revealing the relationship between neuropathological biomarkers and cognition. 

We propose a novel pipeline for the real-time detection of myocardial infarction from a single heartbeat of a 12-lead electrocardiograms. We do so by merging a real-time R-spike detection algorithm with a deep learning Long-Short Term Memory (LSTM) network-based classifier. A comparative assessment of the classification performance of the resulting system is performed and provided. The proposed algorithm achieves an inter-patient classification accuracy of 95.76% (with a 95% Confidence Interval (CI) of ±2.4%), a recall of 96.67% (±2.4% 95% CI), specificity of 93.64% (±5.7% 95% CI), and the average J-Score is 90.31% (±6.2% 95% CI). These state-of-the-art myocardial infarction detection metrics are extremely promising and could pave the wave for the early detection of myocardial infarctions. This high accuracy is achieved with a processing time of 40 milliseconds, which is most appropriate for online classification as the time between fast heartbeats is around 300 milliseconds.