Deep Learning in Alzheimer's disease: Diagnostic Classification and Prognostic Prediction using Neuroimaging Data (1905.00931v4)

Abstract: Deep learning has shown outstanding performance in identifying intricate structures in complex high-dimensional data, especially in the domain of computer vision. The application of deep learning to early detection and automated classification of Alzheimer's disease (AD) has recently gained considerable attention, as rapid progress in neuroimaging techniques has generated large-scale multimodal neuroimaging data. A systematic review of publications using deep learning approaches and neuroimaging data for diagnostic classification of AD was performed. A PubMed and Google Scholar search was used to identify deep learning papers on AD published between January 2013 and July 2018. These papers were reviewed, evaluated, and classified by algorithm and neuroimaging type, and the findings were summarized. Of 16 studies meeting full inclusion criteria, 4 used a combination of deep learning and traditional machine learning approaches, and 12 used only deep learning approaches. The combination of traditional machine learning for classification and stacked auto-encoder (SAE) for feature selection produced accuracies of up to 98.8% for AD classification and 83.7% for prediction of conversion from mild cognitive impairment (MCI), a prodromal stage of AD, to AD. Deep learning approaches, such as convolutional neural network (CNN) or recurrent neural network (RNN), that use neuroimaging data without preprocessing for feature selection have yielded accuracies of up to 96.0% for AD classification and 84.2% for MCI conversion prediction. The best classification performance was obtained when multimodal neuroimaging and fluid biomarkers were combined. AD research that uses deep learning is still evolving, improving performance by incorporating additional hybrid data types, increasing transparency with explainable approaches that add knowledge of specific disease-related features and mechanisms.

Deep Learning in Alzheimer’s Disease: Diagnostic Classification and Prognostic Prediction Using Neuroimaging Data

This paper focuses on employing deep learning methodologies to advance the diagnostic classification and prognostic prediction of Alzheimer's Disease (AD) using neuroimaging data. Given the intricate nature of high-dimensional medical imaging, particularly from modalities such as MRI and PET, deep learning offers automated feature extraction, thereby circumventing some limitations of traditional machine learning techniques.

Methodological Overview

A systematic review was conducted on publications from 2013 to 2018 focusing on deep learning applications for AD classification and progression prediction. The review specifically examined 16 studies adhering to the inclusion criteria, highlighting algorithm selection and neuroimaging modalities used. These studies predominantly utilized either pure deep learning approaches or hybrid methods combining deep learning with machine learning classifiers like SVM.

Key Findings

1. Algorithm Performance:
   • Hybrid approaches combining stacked auto-encoders (SAE) and traditional machine learning achieved the highest accuracy of 98.8% in AD classification.
   • Pure deep learning approaches, using architectures like CNN and RNN without preprocessing, yielded accuracies up to 96.0% for AD classification and 84.2% for predicting conversion from mild cognitive impairment (MCI) to AD.
2. Multimodal Data Integration:
   • Combining multiple neuroimaging modalities often improves classification accuracies. PET, both FDG-PET and amyloid PET, provided better results in AD/CN classification compared to single modalities such as MRI.
3. Data Sensitivity:
   • Deep learning's requirement for extensive data highlights the value of hybrid methods when data are limited. These methods integrate traditional classification techniques with deep learning for improved outcomes.

Implications and Future Directions

The paper indicates that, while deep learning demonstrates high accuracy, issues such as interpretability, transparency, and reproducibility persist. The approaches discussed hold potential for early AD diagnosis, essential for timely intervention and management. However, transitioning these methods to clinical settings demands addressing challenges related to model transparency and data limitations.

Future advancements may involve:

• Exploring hybrid models that integrate diverse data types, including genetic and -omics data, to enhance classification robustness.
• Adopting Generative Adversarial Networks (GANs) for generating synthetic imaging data to augment training datasets.
• Utilizing reinforcement learning to adapt models dynamically to real-world clinical data, potentially improving their contextual applicability.

As computational resources and clinical data repositories expand, there is a clear trajectory towards exclusive reliance on deep learning for AD research, moving beyond hybrid models. This progression will necessitate models capable of integrating heterogeneous data types without extensive preprocessing.

Overall, this research underscores the evolving nature of deep learning in AD diagnostics, showing promise in managing the complexities of multimodal neuroimaging data and elevating diagnostic capabilities through refined automatic feature extraction.