AI Diagnosis of Mild Cognitive Impairment in Neurodegenerative Diseases from FDG Pet Images

Objective This study aimed to develop and validate a deep learning FDG-PET scanning algorithm that may differentiate between patients who have developed moderate cognitive decline compared to the patients having been diagnosed with one of the following neuronal diseases: frontotemporal degeneration and dementia with lewy bodies, and both. A 3D Convolutional brain network that was proposed and trained using images from the Brain scans Institute for Alzheimer’s collection. It contained a dataset used in training and testing the model: the ADNI, which holds data of 822 participants. Also, a dataset from Technological Institute and La Fe Institute was used for the evaluation, which consisted of 91 participants with MCI, of which 71 acquired neurologic condition afterward, while the remaining 19 do not. The model, when tested on the 10% ADNI dataset, identified neurologic illness in patients with an accuracy of 79%, an awareness of 88%, and a specificity of 71%. It further achieved an area around the receiver operating curves of 0.90. Moreover, the model retained its 80% balance accuracy, 75% awareness, 84% specificity, and 0.86 AUC in the external verification. With an overall accuracy of 80% for its balanced classification, this binary classifier algorithm developed through deep learning in PET scan images can become instrumental for the early findings of a neurologic disease in patients with MCI under a standard clinical setting. Definition of Mild Cognitive Impairment MCI is the transitional stage between normal cognitive ageing and dementia, most commonly caused by neurodegenerative diseases such as Alzheimer’s disease. Successful and early identification of patients with MCI at risk for converting their disorder into a neurodegenerative condition is mandatory for timely intervention and management.