Abstract:Several studies using machine learning have recently looked at the potential of diffusion tensor imaging (DTI) for Alzheimer's disease (AD) classification. However, classification accuracies are not directly comparable across studies because of differences in feature extraction, patient selection and validation schemes. Samper-Gonzalez et al (2017) proposed a reproducible framework for classification of AD from T1 MRI and PET from ADNI. Here, we extend this framework to DTI. The framework is composed of the following components. Tools were implemented to automatically convert original ADNI diffusion MRI into the BIDS format. Standardized artifact correction procedures were designed. The DTI model was then fitted generating FA and MD maps. These maps were nonlinearly registered to MNI space and masked using the WM, GM or GM+WM binarized maps. Classification was performed using a linear support vector machine (SVM). A repeated holdout cross-validation (250 runs) with a 10-fold inner grid search for hyperparameter optimization was performed. Additionally, optimal margin hyperplane coefficient maps were reported to characterize potential anatomical patterns (Cuingnet et al, 2013). We demonstrate the use of the framework with 46 AD patients, 46 cognitively normal (CN) subjects, 97 mild cognitive impairment (MCI) subjects, including 54 stable MCI (sMCI) and 24 progressive MCI (pMCI) at 36 months. Additionally, a random undersampling technique with 10 repetitions was used to assess the impact of imbalanced data (e.g 54 sMCI vs 24 pMCI) on the classification performance. Classification results are shown in tables 1 to 4. We found high accuracies for AD vs CN classification but lower accuracies for classifications involving MCI. Balanced and imbalanced classifications gave comparable accuracies. The hippocampus and medial temporal cortex are the main discriminative regions using WM+GM-MD features (Fig 1). Normalized coefficient maps chosen by higher accuracies obtained from four tasks with all available data. The classification accuracy was 75% for CN vs AD with GM+WM-MD, 62% for CN vs MCI with GM+WM-MD, 66% for CN vs pMCI with GM+WM-FA, and 52% for sMCI vs pMCI with GM+WM-FA. In each voxel, high coefficient means higher likelihood of classification into A) AD, B) MCI, C) pMCI and D) pMCI. We presented a framework that enables the comparability and reproducibility of DTI-based AD classification. Classification results are in line with the state of the art. The low accuracies obtained in discriminating pMCI from sMCI or CN are potentially caused by the small data samples, rather than by imbalanced data. The code will be made publicly available at https://gitlab.icm-institute.org/aramislab/AD-ML.

Automated Multiclass Artifact Detection in Diffusion MRI Volumes via 3D Residual Squeeze-and-Excitation Convolutional Neural Networks

3D-QCNet - A pipeline for automated artifact detection in diffusion MRI images

Unsupervised dMRI Artifact Detection via Angular Resolution Enhancement and Cycle Consistency Learning

Motion Artifact Detection for T1-Weighted Brain MR Images Using Convolutional Neural Networks

An efficient semi-supervised quality control system trained using physics-based MRI-artefact generators and adversarial training

A Lightweight Deep Learning Framework for Automatic MRI Data Sorting and Artifacts Detection

Automated detection of motion artifacts in brain MR images using deep learning and explainable artificial intelligence

Automated detection of critical findings in multi-parametric brain MRI using a system of 3D neural networks

Automatic artifact removal of resting-state fMRI with Deep Neural Networks

Objective and Automated Detection of Diffuse White Matter Abnormality in Preterm Infants Using Deep Convolutional Neural Networks.

Automated artifact detection and removal for improved tensor estimation in motion-corrupted DTI data sets using the combination of local binary patterns and 2D partial least squares

Diffusion MRI with Machine Learning

Automatic motion artefact detection in brain T1-weighted magnetic resonance images from a clinical data warehouse using synthetic data

IC‐P‐113: USING DIFFUSION MRI FOR CLASSIFICATION AND PREDICTION OF ALZHEIMER'S DISEASE: A REPRODUCIBLE STUDY

Automated MRI Quality Assessment of Brain T1-weighted MRI in Clinical Data Warehouses: A Transfer Learning Approach Relying on Artefact Simulation

Data Driven-based Motion Artifact Detection from Brain Computed Tomography Images

Iterative Learning for Joint Image Denoising and Motion Artifact Correction of 3D Brain MRI

A latent space for unsupervised MR image quality control via artifact assessment

3D CNN-based classification using sMRI and MD-DTI images for Alzheimer disease studies

Transformer Convolutional Neural Networks for Automated Artifact Detection in Scalp EEG