Current diagnosis of Alzheimer’s Disease (AD) and its precursors is largely based on qualitative readings and assessment of higher nervous functions. These methods present challenges due to the low specificity and variability of the clinical symptoms as well as the continuous spectrum of brain abnormality, ranging from normal cognition to mild cognitive impairment (MCI) to AD. Presently, there is no definite, quantitative biomarker established for clear diagnosis or for prediction of MCI progression. This has led to high rates of misdiagnosis, deterrence of early treatment, and slow development of new drugs and treatments.
My work aims to use deep learning techniques, namely Convolutional Neural Networks (CNNs), to determine specific disease patterns in the brains of AD patients and to identify a quantitative biomarker from 18F-FDG PET images that would aid clinician’s image assessment for early and accurate diagnosis of Alzheimer’s Disease. I aim to train a CNN which can distinguish between AD and other brain abnormalities and predict the conversion of MCI to AD, providing not only specific and early diagnosis, but additional insight into the functional nature of the disease.
In addition to this research, I am also involved in helping to establish the Networks of Imaging eXcellence (NIX) Alliance. The NIX Alliance is an international collaborative effort to accelerate research in quantitative medical imaging. By centralizing shared imaging data sets and offering dependable image analysis tools for expert use, NIX aims to provide an environment for the derivation of reliable, validated quantitative imaging biomarkers. This, in turn, will help shift medical imaging from qualitative interpretation to quantitative data science and empower the effective translation of quantitative imaging into clinical research and practice.