Using 18F-fluorodeoxy-glucose positron emission tomography (FDG-PET) and Machine learning to Predict dementia in short term and long term in an ageing population
Mushtaq, Ayisha (2025-08-13)
Using 18F-fluorodeoxy-glucose positron emission tomography (FDG-PET) and Machine learning to Predict dementia in short term and long term in an ageing population
(13.08.2025)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
avoin
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2025092297139
https://urn.fi/URN:NBN:fi-fe2025092297139
Tiivistelmä
Dementia is a major health concern in aging populations. It requires early detection along with accurate prediction so that necessary steps for intervention can be taken. This is important for improving the patient as well as their loved one’s quality of life. Brain imaging method such as 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) can show that how the brains metabolism changes in dementia patients. At the same time machine learning methods gives us powerful tools to predict that who might develop the disease. In this study the potential of FDG-PET and machine learning to predict neurological disorders (including dementia) and to understand brain metabolism in aging is explored.
This study used brain imaging (18F-FDG PET scans) and machine learning to predict neurological diseases (including dementia) in the short term (1–5 years) and long term (>5 years) in an aging population. It also examined how age affects brain metabolism and whether brain metabolism patterns can point out differences between healthy people (Control), those with memory disorders (e.g., Alzheimer’s disease), and others with different neurological conditions.
Using data from 1,447 participants in the AIVO database, it was found that older age is linked to lower brain metabolism across all 25 brain regions with the strongest effects in areas important for thinking and memory. Clear differences in brain metabolism between the Control and Memory Disorder groups could not be found possibly due to the small sample size (147 participants with complete data) and differences in age and gender. The machine learning models were built with all 25 brain regions. It showed that the long-term model worked well (AUC = 0.8718) and accurately predicting onset of neurological diseases after 5 years, with the bilateral inferior temporal gyri coming out as a key feature. The short-term model performed less well (AUC = 0.6571) to detect early onset of disease. This study highlights the role of brain metabolism in aging, memory disorders and neurological diseases (including dementia) and the value of using comprehensive brain data for prediction. This preliminary analysis of the dataset needs to be extended, the pre-existing image database needs to be further consolidated to obtain more valid cases for the analysis. Further research in the topic is required. Use of external validation cohorts might be needed.
This study used brain imaging (18F-FDG PET scans) and machine learning to predict neurological diseases (including dementia) in the short term (1–5 years) and long term (>5 years) in an aging population. It also examined how age affects brain metabolism and whether brain metabolism patterns can point out differences between healthy people (Control), those with memory disorders (e.g., Alzheimer’s disease), and others with different neurological conditions.
Using data from 1,447 participants in the AIVO database, it was found that older age is linked to lower brain metabolism across all 25 brain regions with the strongest effects in areas important for thinking and memory. Clear differences in brain metabolism between the Control and Memory Disorder groups could not be found possibly due to the small sample size (147 participants with complete data) and differences in age and gender. The machine learning models were built with all 25 brain regions. It showed that the long-term model worked well (AUC = 0.8718) and accurately predicting onset of neurological diseases after 5 years, with the bilateral inferior temporal gyri coming out as a key feature. The short-term model performed less well (AUC = 0.6571) to detect early onset of disease. This study highlights the role of brain metabolism in aging, memory disorders and neurological diseases (including dementia) and the value of using comprehensive brain data for prediction. This preliminary analysis of the dataset needs to be extended, the pre-existing image database needs to be further consolidated to obtain more valid cases for the analysis. Further research in the topic is required. Use of external validation cohorts might be needed.