Design and Application of Automated Algorithms for Diagnosis and Treatment Optimization in Neurodegenerative Diseases

Abstract

Neurodegenerative diseases represent a growing healthcare problem, mainly related to an aging population worldwide and thus their increasing prevalence. In particular, Alzheimer's disease (AD) and Parkinson's disease (PD) are leading neurodegenerative diseases. To aid their diagnosis and optimize treatment, we have developed a classification algorithm for AD to manipulate magnetic resonance images (MRI) stored in a large database of patients, containing 1,200 images. The algorithm can predict whether a patient is healthy, has mild cognitive impairment, or already has AD. We then applied this classification algorithm to therapeutic outcomes in PD after treatment with deep brain stimulation (DBS), to assess which stereotactic variables were the most important to consider when performing surgery in this indication. Here, we describe the stereotactic system used for DBS procedures, and compare different planning methods with the gold standard normally used (i.e., neurophysiological coordinates recorded intraoperatively). We used information collected from database of 72 DBS electrodes implanted in PD patients, and assessed the potentially most beneficial ranges of deviation within planning and neurophysiological coordinates from the operating room, to provide neurosurgeons with additional landmarks that may help to optimize outcomes: we observed that x coordinate deviation within CT scan and gold standard intra-operative neurophysiological coordinates is a robust matric to pre-assess positive therapy outcomes- "good therapy" prediction if deviation is higher than 2.5 mm. When being less than 2.5 mm, adding directly calculated variables deviation (on Y and Z axis) would lead to specific assessment of "very good therapy".