Studying the Manifold Structure of Alzheimer's Disease: A Deep Learning Approach Using Convolutional Autoencoders

  title={Studying the Manifold Structure of Alzheimer's Disease: A Deep Learning Approach Using Convolutional Autoencoders},
  author={Francisco J. Mart{\'i}nez-Murcia and Andr{\'e}s Ortiz and Juan Manuel G{\'o}rriz and Javier Ram{\'i}rez and Diego Castillo-Barnes},
  journal={IEEE Journal of Biomedical and Health Informatics},
Many classical machine learning techniques have been used to explore Alzheimer's disease (AD), evolving from image decomposition techniques such as principal component analysis toward higher complexity, non-linear decomposition algorithms. With the arrival of the deep learning paradigm, it has become possible to extract high-level abstract features directly from MRI images that internally describe the distribution of data in low-dimensional manifolds. In this work, we try a new exploratory data… 

Figures and Tables from this paper

Convolutional Autoencoder based Deep Learning Approach for Alzheimer's Disease Diagnosis using Brain MRI
The 2D neural network is based upon a pre-trained 2D convolutional autoencoder to capture latent representations in structural brain magnetic resonance imaging (MRI) scans and outperforms a number of traditional classifiers in terms of accuracy using a single slice.
It is demonstrated that using 3D convolutional neural networks with sparse auto encoder outperform several other classifiers stated in the literature.
Three-Dimensional Convolutional Autoencoder Extracts Features of Structural Brain Images With a “Diagnostic Label-Free” Approach: Application to Schizophrenia Datasets
This study investigated the efficacy of a 3D convolutional autoencoder (CAE) for extracting features related to psychiatric disorders without diagnostic labels and suggested that a model with eight convolution layers might be relevant to extract features for predicting the dose of medication and symptom severity in schizophrenia.
Alzheimer's disease detection using convolutional neural networks and transfer learning based methods
A deep based method for the detection of AD (i.e. classify brain images into normal brain or brain with AD) is proposed and the obtained results show that the classification of images using Transfer Learning with 92.86 % outperformed the CNN's classification rate.
CrossViT Wide Residual Squeeze-and-Excitation Network for Alzheimer's disease classification with self attention ProGAN data augmentation
A new method based on the combination of the Cross ViT and Wide Residual Squeeze-and-Excitation Network to overcome the limitation of the data and achieve 99% classification accuracy and outperforms CNN models.
Quantification of Cognitive Function in Alzheimer’s Disease Based on Deep Learning
The algorithm proposed in this paper can automatically extract and fuse the features of multi-modality and multi-regions layer by layer, and the visual analysis results show that the abnormally changed regions affected by Alzheimer’s disease provide important information for clinical quantification.
Multi-task Learning and Ensemble Approach to Predict Cognitive Scores for Patients with Alzheimer’s Disease
A considerable performance improvement from the initial multi-task U-Net on the ADAS-cog 11 prediction task was achieved across the authors' experiments, and the final model integrates the deep learning results with a Gradient Boosting model trained on demographic data.
XADLiME: eXplainable Alzheimer's Disease Likelihood Map Estimation via Clinically-guided Prototype Learning
This work proposes a novel deep-learning approach through eXplainable AD Likelihood Map Estimation (XADLiME) for AD progression modeling over 3D sMRIs using clinically-guided prototype learning and promotes the explainability of such a likelihood map by revealing a comprehensible overview from two perspectives: clinical and morphological.


Ensembles of Deep Learning Architectures for the Early Diagnosis of the Alzheimer's Disease
Computer Aided Diagnosis (CAD) constitutes an important tool for the early diagnosis of Alzheimer's Disease (AD), which, in turn, allows the application of treatments that can be simpler and more
Latent feature representation with stacked auto-encoder for AD/MCI diagnosis
A deep learning-based latent feature representation with a stacked auto-encoder (SAE) helps build a robust model in AD/MCI classification, with high diagnostic accuracy, and the applicability of this method to brain disease diagnosis is presented.
Functional Brain Imaging Synthesis Based on Image Decomposition and Kernel Modeling: Application to Neurodegenerative Diseases
These results prove that these images are suitable for standardizing the evaluation of CAD pipelines, and providing data augmentation in machine learning systems -e.g. in deep learning-, or even to train future professionals at medical school.
Learning Longitudinal MRI Patterns by SICE and Deep Learning: Assessing the Alzheimer's Disease Progression
This paper presents a method to construct longitudinal models from subjects using a series of MRI images that can be used to classify between Controls (CN) and AD subjects with 94% of accuracy, outperforming other previous methods.
Predicting Alzheimer's Disease - A Neuroimaging Study with 3D Convolutional Neural Networks
This paper uses deep learning methods, and in particular sparse autoencoders and 3D convolutional neural networks, to build an algorithm that can predict the disease status of a patient, based on an MRI scan of the brain.
Convolutional Neural Networks for Neuroimaging in Parkinson's Disease: Is Preprocessing Needed?
This work has trained four different CNN models based on well-established architectures, using or not different spatial and intensity normalization preprocessings, and shows that a sufficiently complex model such as the three-dimensional version of the ALEXNET can effectively account for spatial differences.
SVM feature selection for classification of SPECT images of Alzheimer's disease using spatial information
The proposed formulation incorporates proximity information about the features and generates a classifier that does not just select the most relevant voxels but the mostrelevant “areas” for classification resulting in more robust classifiers that are better suitable for interpretation.