Information Geometry and Manifold Learning: A Novel Framework for Analyzing Alzheimer’s Disease MRI Data

Diagnostics, Journal Year: 2025, Volume and Issue: 15(2), P. 153 - 153

Published: Jan. 10, 2025

Background: Alzheimer’s disease is a progressive neurological condition marked by decline in cognitive abilities. Early diagnosis crucial but challenging due to overlapping symptoms among impairment stages, necessitating non-invasive, reliable diagnostic tools. Methods: We applied information geometry and manifold learning analyze grayscale MRI scans classified into No Impairment, Very Mild, Moderate Impairment. Preprocessed images were reduced via Principal Component Analysis (retaining 95% variance) converted statistical manifolds using estimated mean vectors covariance matrices. Geodesic distances, computed with the Fisher Information metric, quantified class differences. Graph Neural Networks, including Convolutional Networks (GCN), Attention (GAT), GraphSAGE, utilized categorize levels graph-based representations of data. Results: Significant differences structures observed, increased variability stronger feature correlations at higher levels. distances between Impairment Mild (58.68, p<0.001) (58.28, are statistically significant. GCN GraphSAGE achieve perfect classification accuracy (precision, recall, F1-Score: 1.0), correctly identifying all instances across classes. GAT attains an overall 59.61%, variable performance Conclusions: Integrating geometry, learning, GNNs effectively differentiates AD stages from The strong indicates their potential assist clinicians early identification tracking progression.

Language: Английский

---

MHNet: Multi-view High-Order Network for Diagnosing Neurodevelopmental Disorders Using Resting-State fMRI

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 28, 2025

Abstract Deep learning models have shown promise in diagnosing neurodevelopmental disorders (NDD) like ASD and ADHD. However, many either use graph neural networks (GNN) to construct single-level brain functional (BFNs) or employ spatial convolution filtering for local information extraction from rs-fMRI data, often neglecting high-order features crucial NDD classification. We introduce a Multi-view High-order Network (MHNet) capture hierarchical multi-view BFNs derived data prediction. MHNet has two branches: the Euclidean Space Features Extraction (ESFE) module Non-Euclidean (Non-ESFE) module, followed by Feature Fusion-based Classification (FFC) identification. ESFE includes Functional Connectivity Generation (FCG) Convolutional Neural (HCNN) extract space. Non-ESFE comprises Generic Internet-like Brain Hierarchical (G-IBHN-G) Graph (HGNN) topological non-Euclidean Experiments on three public datasets show that outperforms state-of-the-art methods using both AAL1 Brainnetome Atlas templates. Extensive ablation studies confirm superiority of effectiveness fMRI features. Our study also offers atlas options constructing more sophisticated explains association between key regions NDD. leverages feature spaces, incorporating enhance classification performance.

Language: Английский

---

Open-World Semi-Supervised Learning for fMRI Analysis to Diagnose Psychiatric Disease

Information, Journal Year: 2025, Volume and Issue: 16(3), P. 171 - 171

Published: Feb. 25, 2025

Due to the incomplete nature of cognitive testing data and human subjective biases, accurately diagnosing mental disease using functional magnetic resonance imaging (fMRI) poses a challenging task. In clinical diagnosis disorders, there often arises problem limited labeled due factors such as large volumes cumbersome labeling processes, leading emergence unlabeled with new classes, which can result in misdiagnosis. context graph-based disorder classification, open-world semi-supervised learning for node classification aims classify nodes into known classes or potentially presenting practical yet underexplored issue within graph community. To improve representation fMRI under low-label settings, we propose novel approach tailored analysis, termed Open-World Semi-Supervised Learning Analysis (OpenfMA). Specifically, employ spectral augmentation self-supervised dynamic concept contrastive achieve guided by pseudo-labels, construct hard positive sample pairs enhance network’s focus on potential pairs. Experiments conducted public datasets validate superior performance this method psychiatric domain.

Language: Английский

---

Incorporating attributes and multi-scale structures for heterogeneous graph contrastive learning

Information Fusion, Journal Year: 2025, Volume and Issue: unknown, P. 103220 - 103220

Published: April 1, 2025

Language: Английский

---

Robust fMRI time-varying functional connectivity analysis using multivariate swarm decomposition

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 130404 - 130404

Published: May 1, 2025

Language: Английский

---

Application of multi-modal data fusion based on deep learning in diagnosis of depression

Data & Metadata, Journal Year: 2025, Volume and Issue: 4, P. 863 - 863

Published: April 4, 2025

Depression is a frequent mental condition requiring precise diagnosis in its early onset. Traditional methods are less than accurate and occur late. Following these deficits, this investigates the multi-modal data fusion Deep Learning (DL) with purpose of enhancing accuracy for diagnosis. A new DL model, Dynamic Dolphin Echolocation-tuned Effective Temporal Convolutional Networks (DDE-ETCN), utilized depression Different sources data, such as physiological signals (EEG, heart rate), behavioral indicators (facial expressions), biometric (activity levels), fused. Data preprocessing includes wavelet transformation normalization median filtering to provide smooth inputs. Feature extraction performed through Fast Fourier Transform (FFT) obtain frequency-domain features indicators. Feature-level good all sources, which improves model's performance. The DDE tuning mechanism temporal convolution layers improve ability detecting sequential changes. proposed DDE-ETCN model highly when it developed Python. attains an RMSE 3.59 MAE 3.09. It has 98.72% accuracy, 98.13% precision, 97.65% F1-score, 97.81% recall, outperforming conventional diagnostic models other deep learning-based models. outcomes show efficiency rendering more objective Its higher performance justifies potential practical use, providing enhanced reliability compared traditional approaches. This innovation emphasizes necessity incorporating learning health evaluations.

Language: Английский

---

ORC-GNN: A novel open set recognition based on graph neural network for multi-class classification of psychiatric disorders

Information Fusion, Journal Year: 2024, Volume and Issue: unknown, P. 102887 - 102887

Published: Dec. 1, 2024

Language: Английский

---