The IIS La Fe advances in the development of an algorithm that allows the identification of patients with neurodegenerative diseases through medical imaging

September 21, 2022
news 2.png (88 KB)
News brought to you by Linea Directa 
Cool 728x90_Expats_Home-3.jpg (76 KB)

The Biomedical Imaging and Alzheimer's Disease Research groups of the La Fe Health Research Institute (IIS La Fe), in collaboration with the Molecular Imaging Instrumentation Institute of the Polytechnic University of Valencia (UPV), are working on the development of an algorithm that allows, in a non-invasive way, to identify patients with mild cognitive impairment who may develop neurodegenerative diseases such as Alzheimer's. Every day on September 21, the world day of this disease is commemorated.

The objective of the project is to develop and validate this deep learning algorithm based on neural networks that allows the identification of patients with neurodegenerative diseases such as Alzheimer's Disease, Frontotemporal Degeneration and dementia with Lewy bodies, among a group of people with Mild Cognitive Impairment. , through brain images of Positron Emission Tomography with Fluorodeoxyglucose (PET FDG).

FDG PET is a non-invasive technique that, together with Artificial Intelligence, could reduce the number of patients undergoing much more invasive procedures such as lumbar puncture, a method through which cerebrospinal fluid biomarkers are obtained that allow distinguish Alzheimer's disease in patients with Mild Cognitive Impairment.

Thus, the purpose of the algorithm, on which the research staff is working, is to reduce the use of this invasive technique in patients who have a reduced probability of presenting a neurodegenerative disease.

For the development of this artificial intelligence algorithm, which has managed to go from an experimental phase to an operational validation within the hospital environment, it has been based on a database containing PET FDG images called Alzheimer's Disease Neuroimaging (ADNI).

Thus, for the prior training of the neural network, a total of 822 subjects were used, of which 472 had Alzheimer's disease and 350 had mild cognitive impairment. On the other hand, for the validation phase, an independent database made up of 90 patients from the Hospital Universitari i Politècnic La Fe de València was used.

The quantitative information extracted from PET images is very useful for the diagnosis and evaluation of dementia. However, it is sometimes visually complex for doctors, which is why artificial intelligence can be of great support for diagnosis and currently provides satisfactory results.

In this sense, this algorithm that extracts and uses information from PET FDG images has allowed the prediction at an early stage of neurodegenerative diseases in patients with mild cognitive impairment with an accuracy of 80%. Even so, the project has to continue advancing in improvements in the accuracy of prediction and the adaptation of the software for clinical use.

20 new cases of Alzheimer's daily in the Community

Early diagnosis is crucial, also in this disease, because it allows orientation from the beginning to the appropriate treatment that can favor the evolution both in the short and in the medium and long term.

The Ministry of Universal Health and Public Health diagnoses an average of 20 new cases of Alzheimer's every day. It is a degenerative disease that affects memory and behavior, and that affects women more, in a proportion of 7 females for every 3 males, approximately. Regarding age, the majority of new patients are over 65 years old.