Generalizable deep learning model for early Alzheimer’s disease detection from structural MRIs

Abstract

Early diagnosis of Alzheimer’s disease plays a pivotal role in patient care and clinical trials. In this study, we have developed a new approach based on 3D deep convolutional neural networks to accurately differentiate mild Alzheimer’s disease dementia from mild cognitive impairment and cognitively normal individuals using structural MRIs. For comparison, we have built a reference model based on the volumes and thickness of previously reported brain regions that are known to be implicated in disease progression. We validate both models on an internal held-out cohort from The Alzheimer's Disease Neuroimaging Initiative (ADNI) and on an external independent cohort from The National Alzheimer's Coordinating Center (NACC). The deep-learning model is accurate, achieved an area-under-the-curve (AUC) of 85.12 when distinguishing between cognitive normal subjects and subjects with either MCI or mild Alzheimer’s dementia. In the more challenging task of detecting MCI, it achieves an AUC of 62.45. It is also significantly faster than the volume/thickness model in which the volumes and thickness need to be extracted beforehand. The model can also be used to forecast progression: subjects with mild cognitive impairment misclassified as having mild Alzheimer’s disease dementia by the model were faster to progress to dementia over time. An analysis of the features learned by the proposed model shows that it relies on a wide range of regions associated with Alzheimer's disease. These findings suggest that deep neural networks can automatically learn to identify imaging biomarkers that are predictive of Alzheimer's disease, and leverage them to achieve accurate early detection of the disease.

CAD-ALZ: A Blockwise Fine-Tuning Strategy on Convolutional Model and Random Forest Classifier for Recognition of Multistage Alzheimer’s Disease

“…The suggested framework obtained 97.5% classification accuracy on the ADNI dataset in multi-classification studies. A unique CNN design that can discriminate between people with normal cognition, MCI, and mild Alzheimer's disease dementia was described in [51]. The authors used the ADNI dataset to evaluate the CNN architecture.…”

Convolution Neural Networks and Self-Attention Learners for Alzheimer Dementia Diagnosis from Brain MRI

“…Since it depends on many training parameters, using DL on 3D brain volumes significantly increases the computational cost [ 10 ]. In addition, the availability of 3D data is limited, and its use may suffer from the curse of dimensionality [ 16 ], limiting the ability to create accurate models. In addition, pre-trained 3D models are not as widely available as 2D models (trained on huge image datasets) [ 17 ].…”

Self-supervised learning for early detection of neurodegenerative diseases with small data

“…CNNs have been used to classify MRI scans from subjects at different stages on the AD spectrum [365][366][367][368][369][370][371]. High accuracy can be achieved on the AD vs CN task, as it compares samples from two ends of the disease spectrum.…”

“…Recent works use a deep learning approach to extract the image features at the level of granularity which optimizes performance. CNNs are trained end-to-end and process each MRI in its entirety [365][366][367][368][369] or at patch level [370,371]. Multimodal machine learning has also been explored to integrate tabular (e.g., clinical test scores) [376,377] or Positron Emission Tomography (PET) [378,379] features to complement the image-based features extracted from MRI.…”

“…Convolutional Neural Network (CNN) architectures have seen widespread adoption in the domain of MRI [366,368] due to their state-of-the-art performance on image recognition tasks. Adapting CNNs to classify MRI scans is challenging due to the low signal-to-noise ratio and heterogeneity of pathological patterns.…”