Evaluation of Transfer Learning Methods for Detecting Alzheimer’s Disease with Brain MRI

Abstract

Deep neural networks show great promise for classifying brain diseases and making prognostic assessments based on neuroimaging data, but large, labeled training datasets are often required to achieve high predictive accuracy. Here we evaluated a range of transfer learning or pre-training strategies to create useful MRI representations for downstream tasks that lack large amounts of training data, such as Alzheimer's disease (AD) classification. To test our models, we analyzed 4,098 3D T1-weighted brain MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort and independently validated our proposed methods for detecting AD with an out-of-distribution test set of 600 scans from the Open Access Series of Imaging Studies (OASIS3) cohort. First, we trained 3D and 2D convolutional neural network (CNN) architectures. We tested combinations of multiple pre-training strategies based on (1) supervised, (2) contrastive learning, and (3) self-supervised learning - using pre-training data within versus outside the MRI domain. In our experiments, the 3D CNN pre-trained with contrastive learning provided the best overall results - when fine-tuned on T1-weighted scans for AD classification - outperformed the baseline by 2.8% when trained with all of the training data from ADNI. We also show test performance as a function of the training dataset size and the chosen pre-training method. Transfer learning offered significant benefits in low data regimes, with a performance boost of 7.7%. When the pre-trained model was used for AD classification, we were able to visualize an improved clustering of test subjects' diagnostic groups, as illustrated via a uniform manifold approximation (UMAP) projection of the high-dimensional model embedding space. Further, saliency maps indicate the additional activation regions in the brain scan using pre-training, that then maximally contributed towards the final prediction score.

Deep causal feature extraction and inference with neuroimaging genetic data

“…The sample size of the ADNI, around 800, is still small. Other techniques, such as transfer learning, 44 data augmentation 17 and self‐supervised learning can be explored and incorporated in the future for better performance. Third, perhaps most importantly, it is still quite challenging to interpret CNN‐extracted features.…”

Brain Age Analysis and Dementia Classification using Convolutional Neural Networks trained on Diffusion MRI: Tests in Indian and North American Cohorts

“…To evaluate the performance of our deep learning techniques, we first trained and tested CNNs on the task of predicting brain age. In this work, we focus on CNNs as they are widely used and well understood, but other networks (such as vision transformers [15]) have also been used for these and related tasks. While a person's chronological age is typically known and may not hold immediate clinical utility, predicting brain age in healthy control subjects is a standard benchmarking task, as ground truth is known.…”

Structural biomarker‐based Alzheimer's disease detection via ensemble learning techniques

“…Additionally, vision transformer models have shown promise in enhancing performance for larger‐scale population neuroimaging diagnostic and prognostic tasks 27 . ML algorithms based on structural MRI have also proven effective in detecting AD, demonstrating significant sensitivity, specificity, and diagnostic odds ratio 28 …”

“…27 ML algorithms based on structural MRI have also proven effective in detecting AD, demonstrating significant sensitivity, specificity, and diagnostic odds ratio. 28 Researchers have developed advanced technologies like multi-relation reasoning networks (MRNs) and three-dimensional Jacobian domain convolutional neural networks (JD-CNNs) to enhance the P rec of AD diagnosis. These technologies analyze the brain regions in sMRI data and improve the A cc of AD diagnosis.…”