[Paper Review] Classification of Alzheimer's Disease using fMRI Data and Deep Learning Convolutional Neural Networks

Saman Sarraf, Ghassem Tofighi|arXiv (Cornell University)|Mar 29, 2016

Brain Tumor Detection and Classification127 citations

TL;DR

The paper uses a CNN (LeNet-5) to classify Alzheimer's disease from healthy controls using preprocessed fMRI data from ADNI, achieving about 96.86% test accuracy across 5-fold runs.

ABSTRACT

Over the past decade, machine learning techniques especially predictive modeling and pattern recognition in biomedical sciences from drug delivery system to medical imaging has become one of the important methods which are assisting researchers to have deeper understanding of entire issue and to solve complex medical problems. Deep learning is power learning machine learning algorithm in classification while extracting high-level features. In this paper, we used convolutional neural network to classify Alzheimer's brain from normal healthy brain. The importance of classifying this kind of medical data is to potentially develop a predict model or system in order to recognize the type disease from normal subjects or to estimate the stage of the disease. Classification of clinical data such as Alzheimer's disease has been always challenging and most problematic part has been always selecting the most discriminative features. Using Convolutional Neural Network (CNN) and the famous architecture LeNet-5, we successfully classified functional MRI data of Alzheimer's subjects from normal controls where the accuracy of test data on trained data reached 96.85%. This experiment suggests us the shift and scale invariant features extracted by CNN followed by deep learning classification is most powerful method to distinguish clinical data from healthy data in fMRI. This approach also enables us to expand our methodology to predict more complicated systems.

Motivation & Objective

• Motivate how deep learning can aid in classifying Alzheimer’s disease from fMRI data.
• Develop a CNN-based pipeline to convert fMRI 4D data into image-like inputs for classification.
• Evaluate the performance and robustness of the CNN approach with cross-validation.

Proposed method

• Preprocess fMRI data from ADNI and convert 4D fMRI to 2D JPEG images; label as Alzheimer’s vs Normal.
• Adopt LeNet-5 CNN architecture and train via Caffe DIGITS on an NVIDIA GPU setup.
• Split data into training (60%), validation (20%), and testing (20%); perform 5-fold cross-validation.
• Train for 30 epochs with batch size 64; total samples: 270,900 training and 90,300 validation/testing images.
• Report accuracies across 5 runs and discuss learning-rate schedule and loss/accuracy trends.

Experimental results

Research questions

• RQ1 Can a CNN architecture approximate LeNet-5 effectively on fMRI-derived image inputs for Alzheimer’s vs healthy classification?
• RQ2 What is the achievable accuracy and stability of the CNN-based approach with cross-validation on ADNI-derived fMRI data?
• RQ3 Does the CNN capture discriminative, shift/scale-invariant features suitable for Alzheimer’s classification from rs-fMRI data?

Key findings

• Mean test accuracy across 5 runs is 96.8588% (ranges approximately 96.854%–96.863%).
• Training used 30 epochs with a learning-rate schedule (0.01 initial, divided by 10 every 10 epochs).
• CNN features (filters/kernels) provide visualization of learned representations across layers.
• Compared to prior SVM-based approaches (~84%), the CNN approach achieved a substantial accuracy gain on the same data processing pipeline.