Diagnosis of Alzheimer's disease via optimized lightweight convolution-attention and structural MRI

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 171, P. 108116 - 108116

Published: Feb. 8, 2024

Alzheimer's disease (AD) poses a substantial public health challenge, demanding accurate screening and diagnosis. Identifying AD in its early stages, including mild cognitive impairment (MCI) healthy control (HC), is crucial given the global aging population. Structural magnetic resonance imaging (sMRI) essential for understanding brain's structural changes due to atrophy. While current deep learning networks overlook voxel long-term dependencies, vision transformers (ViT) excel at recognizing such dependencies images, making them valuable Our proposed method integrates convolution-attention mechanisms transformer-based classifiers brain datasets, enhancing performance without excessive computing resources. Replacing multi-head attention with lightweight self-attention (LMHSA), employing inverted residual (IRU) blocks, introducing local feed-forward (LFFN) yields exceptional results. Training on datasets gradient-centralized optimizer Adam achieves an impressive accuracy rate of 94.31% multi-class classification, rising 95.37% binary classification (AD vs. HC) 92.15% HC MCI. These outcomes surpass existing diagnosis approaches, showcasing model's efficacy. key regions aids future clinical solutions neurodegenerative diseases. However, this study focused exclusively Neuroimaging Initiative (ADNI) cohort, emphasizing need more robust, generalizable approach incorporating diverse databases beyond ADNI research.

Language: Английский

---

Attention-map augmentation for hypercomplex breast cancer classification

Pattern Recognition Letters, Journal Year: 2024, Volume and Issue: 182, P. 140 - 146

Published: April 18, 2024

Breast cancer is the most widespread neoplasm among women and early detection of this disease critical. Deep learning techniques have become great interest to improve diagnostic performance. However, distinguishing between malignant benign masses in whole mammograms poses a challenge, as they appear nearly identical an untrained eye, region (ROI) constitutes only small fraction entire image. In paper, we propose framework, parameterized hypercomplex attention maps (PHAM), overcome these problems. Specifically, deploy augmentation step based on computing maps. Then, are used condition classification by constructing multi-dimensional input comprised original breast image corresponding map. step, neural network (PHNN) employed perform classification. The framework offers two main advantages. First, provide critical information regarding ROI allow model concentrate it. Second, architecture has ability local relations dimensions thanks algebra rules, thus properly exploiting provided We demonstrate efficacy proposed both mammography images well histopathological ones. surpass attention-based state-of-the-art networks real-valued counterpart our approach. code work available at https://github.com/elelo22/AttentionBCS.

Language: Английский

---

A quantitatively interpretable model for Alzheimer’s disease prediction using deep counterfactuals

NeuroImage, Journal Year: 2025, Volume and Issue: unknown, P. 121077 - 121077

Published: Feb. 1, 2025

Language: Английский

---

Enhancing Alzheimer’s disease diagnosis and staging: a multistage CNN framework using MRI

Frontiers in Psychiatry, Journal Year: 2024, Volume and Issue: 15

Published: June 24, 2024

This study addresses the pervasive and debilitating impact of Alzheimer’s disease (AD) on individuals society, emphasizing crucial need for timely diagnosis. We present a multistage convolutional neural network (CNN)-based framework AD detection sub-classification using brain magnetic resonance imaging (MRI). After preprocessing, 26-layer CNN model was designed to differentiate between healthy patients with dementia. detecting dementia, reutilized concept transfer learning further subclassify dementia into mild, moderate, severe Leveraging frozen weights developed correlated medical images facilitated process sub-classifying classes. An online dataset is used verify performance proposed CNN-based framework. The approach yielded noteworthy accuracy 98.24% in identifying classes, whereas it achieved 99.70% subclassification. Another validate framework, resulting 100% performance. Comparative evaluations against pre-trained models current literature were also conducted, highlighting usefulness superiority presenting as robust effective subclassification method.

Language: Английский

---

Ensemble approach of transfer learning and vision transformer leveraging explainable AI for disease diagnosis: An advancement towards smart healthcare 5.0

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 179, P. 108874 - 108874

Published: July 15, 2024

Language: Английский

---

Applications of deep learning in Alzheimer’s disease: a systematic literature review of current trends, methodologies, challenges, innovations, and future directions

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 58(2)

Published: Dec. 20, 2024

Language: Английский

---

Transformer’s Role in Brain MRI: A Scoping Review

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 108876 - 108896

Published: Jan. 1, 2024

Magnetic Resonance Imaging (MRI) is a critical imaging technique that provides detailed visualization of internal structures without harmful radiation. This review focuses on key MRI modalities, including T1-weighted and T2-weighted imaging, functional (fMRI), diffusion (dMRI). images offer precise anatomical details, whereas are essential for highlighting abnormalities such as tumors inflammation. Functional (fMRI) captures blood flow changes related to neural activity, (dMRI) tracks the movement water molecules within brain tissues. Our synthesizes insights from 173 studies across major databases, PubMed, ACM Digital Library, IEEE Xplore, Google Scholar. We emphasize versatility transformer architectures in neuroimaging applications, segmentation, detection, reconstruction, super-resolution, with particular focus tumor segmentation notable achievement. Despite these successes, there remains significant gap research, need further collaborative efforts fully realize potential transformers applications. Following PRISMA-ScR guidelines, this analysis explores current trends, dataset availability, overall research landscape. It calls scientific community investigate underexplored capabilities transformers, aiming inspire comprehensive could revolutionize advance fields medical neuroscience.

Language: Английский

---

Multiple Inputs and Mixed Data for Alzheimer’s Disease Classification Based on 3D Vision Transformer

Mathematics, Journal Year: 2024, Volume and Issue: 12(17), P. 2720 - 2720

Published: Aug. 31, 2024

The current methods for diagnosing Alzheimer’s Disease using Magnetic Resonance Imaging (MRI) have significant limitations. Many previous studies used 2D Transformers to analyze individual brain slices independently, potentially losing critical 3D contextual information. Region of interest-based models often focus on only a few regions despite affecting multiple areas. Additionally, most classification rely single test, whereas requires multifaceted approach integrating diverse data sources more accurate assessment. This study introduces novel methodology called the Multiple Inputs and Mixed Data Vision Transformer (MIMD-3DVT). method processes consecutive together capture feature dimensions spatial information, fuses ROI imaging inputs, integrates mixed from demographic factors, cognitive assessments, imaging. proposed was experimentally evaluated combined dataset that included Neuroimaging Initiative (ADNI), Australian Imaging, Biomarker, Lifestyle Flagship Study Ageing (AIBL), Open Access Series Studies (OASIS). Our MIMD-3DVT, utilizing or ROIs, achieved an accuracy 97.14%, outperforming state-of-the-art in distinguishing between Normal Cognition Disease.

Language: Английский

---

Advancements in Alzheimer's disease classification using deep learning frameworks for multimodal neuroimaging: A comprehensive review

Computers & Electrical Engineering, Journal Year: 2024, Volume and Issue: 120, P. 109796 - 109796

Published: Nov. 1, 2024

Language: Английский

---

Ensemble Vision Transformer for Dementia Diagnosis

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2024, Volume and Issue: 28(9), P. 5551 - 5561

Published: June 18, 2024

In recent years, deep learning has gained momentum in computer-aided Alzheimer's Disease (AD) diagnosis. This study introduces a novel approach, Monte Carlo Ensemble Vision Transformer (MC-ViT), which develops an ensemble approach with transformer (ViT). Instead of using traditional methods that deploy multiple learners, our employs single vision learner. By harnessing sampling, this method produces broad spectrum classification decisions, enhancing the MC-ViT performance. technique adeptly overcomes limitation 3D patch convolutional neural networks only characterize partial whole brain anatomy, paving way for network adept at discerning inter-feature correlations. Evaluations Neuroimaging Initiative (ADNI) dataset 7199 scans and Open Access Series Imaging Studies-3 (OASIS-3) 1992 showcased its With minimal preprocessing, achieved impressive 90% accuracy AD classification, surpassing both 2D-slice CNNs CNNs.

Language: Английский

---