Artificial intelligence for early detection of dementia

Alzheimer’s disease is the most common cause of dementia and affects millions of elderly people around the world. In fact, in Spain alone, it is estimated that there are more than 900,000 people affected by this disease, which has become a public health priority. Early detection is key to improving the quality of life of those affected and their families, but identifying it in its initial stages is not always easy. In order to try to improve in this field and achieve better results, an international team, which includes researchers from the Polytechnic University of Madrid (UPM) in Spain, has applied machine learning techniques (a type of artificial intelligence) to the analysis of different types of medical images used for the diagnosis of neurological diseases.

“From a clinical point of view, magnetic resonance imaging (MRI) and positron emission tomography (PET) are the two types of medical imaging used in the diagnosis of this type of disease, as they provide complementary information on the anatomical and metabolic aspects of the disease. Unfortunately, however, these tests are not performed synchronously, which makes their integration and the proper interpretation of their results by medical professionals difficult,” explains Consuelo Gonzalo, a researcher at the UPM Biomedical Technology Centre and one of the authors of this research.

Addressing this problem is the objective set by the UPM researchers and the proposal to do so was to develop a methodology that uses convolutional neural networks, a machine learning technique that enhances image analysis and computer vision tasks, allowing for obtaining significant information from digital images, videos and other visual inputs, as well as taking measures based on those inputs.

To this end, the UPM researchers performed a systematic analysis of MRI and PET images for the assessment of dementia status, using different fusion techniques (early, late and intermediate fusion). They then designed and implemented a solution entirely based on 3D convolutional neural networks () that extracted features from the entire brain volume in three dimensions. Once this was done, they proposed a training strategy capable of handling a highly imbalanced and incomplete data set.

“As far as we know, the proposed methodology represents the first work that provides an analysis of different fusion techniques based on multimodal deep learning for the assessment of dementia severity,” explains the UPM researcher. “The type of solutions developed in this work can be a decision-making support tool of enormous practical interest for neurologists,” she adds.

The research team has applied artificial intelligence to the analysis of different types of medical images used to diagnose neurological diseases that cause cognitive impairment. In the photograph, researchers from the Polytechnic University of Madrid working on it. (Photo: UPM)

The team that carried out the study also includes researchers from the Italian universities of Naples and Rome, and the University of Umea in Sweden. The first author of the study is Michela Gravina, from the University of Naples.

In future work, the researchers intend to continue exploring the fusion of different modalities, further analyzing the properties of the representation of shared features. “Approaches that aim to improve the integration of heterogeneous data should be investigated, generalizing them to case studies with more than two imaging modalities. The explainability of the implemented models should also be addressed, evaluating the decisions made by the networks in comparison to the clinical diagnosis,” they conclude.

The study is titled “Multi input–Multi output 3D for dementia severity assessment with incomplete multimodal data”. It has been published in the academic journal Artificial Intelligence in Medicine. (Source: UPM)