AI-Enabled 'Future' FDG-PET Brain Scans Predict Brain Changes in Alzheimer Patients

Previous research has shown that artificial intelligence (AI) can predict clinical symptomatic changes in neuropsychiatric disorders based on baseline neuroimaging data. However, successful studies predicting actual longitudinal changes in the entire brain are relatively few compared to those focusing on specific longitudinal alterations like hippocampal volume. Now, a preliminary study indicates that a deep learning-based algorithm can accurately predict brain development up to six years following an initial Alzheimer’s disease assessment via FDG-PET scans.

Researchers at the German Center for Neurodegenerative Diseases (DZNE, Göttingen, Germany) employed a convolutional neural network (CNN) to train an algorithm on the first two FDG-PET scans to predict the third scan acquired in elderly (>+ 55 years) participants, who received FDG-PET imaging in three consecutive years. The algorithm successfully predicted the overall future FDG-PET signal for the entire brain—namely, the metabolic reduction, which indicates neuronal activity. The tool was also capable of anticipating a future signal decline, or metabolic reduction, reflecting a loss of neuronal activity.

Image: An algorithm can forecast brain development from images obtained in FDG-PET examinations (Photo courtesy of Freepik)

The algorithm's capabilities could be extended to predict FDG-PET outcomes up to six years following the initial scan, by sequentially using model output as input for subsequent-year predictions. Additionally, the tool seemed to detect ongoing neurodegenerative processes at baseline as it predicted a significant signal decline in year 2 in Alzheimer’s disease (AD) patients, especially in AD-prone regions such as the bilateral inferior temporal and parietal regions, and the posterior cingulate cortex. Possessing a tool that forecasts longitudinal FDG-PET scans based on scans obtained at baseline and one year later could enhance patient care. This study explores new territory, as the prediction of longitudinal metabolic changes in the brain, as measured by FDG-PET, has rarely been examined before.

“Such an algorithm would allow physicians to read an anticipated ‘future’ FDG-PET brain scan as they would in their normal routine, but years in advance,” said Elena Doering, a Ph.D. student at DZNE. “We hope that our work can provide clinical benefit in two ways: improving early diagnosis or providing reliable prognosis; and allowing individual prediction of brain pathological changes over time.”

Related Links:
DZNE