Using data from the first MRI scans of multiple sclerosis patients from the neurology department of the Hospital Clínico de Santiago de Compostela, the program used artificial intelligence to predict the development of these patients over a period of 10 years with great precision, achieving an accuracy of almost 90 percent. This is the result of a new study by the Translational Research Group for Neurological Diseases (ITEN) of IDIS and IIS Galicia.
The study, which was published in PLOS One, analyzed 446 data sets from patients with at least one year of follow-up.
This machine learning model predicts disability progression in MS patients using baseline MRI data and clinical assessments via the Expanded Disability Status Scale.
"The study proposes new models to describe the progression of patients using AI programs that predict their progression with these descriptors. In addition, it provides insights into factors that contribute to this progression, such as age of onset or lesions," explains the study's lead author, Silvia Campanioni.
This research will thus make it possible to optimize the dosage and duration of multiple sclerosis treatment and tailor treatments to individual patient profiles, while improving progression through personalized ML predictors.
One of the most remarkable findings of the study is that the "age of onset" is one of the most influential features for the developed regressor models. In addition, the number of brain lesions greater than or equal to nine on the first MRI proved to be the most influential variable for the classification model decisions.
"This work is important not only in scientific and technical terms, but also in economic and social terms, as it has implications for health, quality of life and development cooperation," explains IDIS researcher and lead author of the project, Roberto Agís. "We could obtain objective evidence and indicators for preventive measures that help predict the therapeutic effectiveness of treatments," he explains.
Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease that leads to demyelination and long-term disability. Autoreactive activation of adaptive immunity plays an important role in MS. Although the cause that triggers the disease is not known, it is thought to result from an interaction between genes and environmental, viral and lifestyle-related risk factors.
According to the research team, the AI study can provide versatile and powerful tools for the treatment of MS. "AI technologies such as deep learning and machine learning could support the integration of biological, psychological and social factors in the prevention, diagnosis and treatment of MS and other diseases," explains César Veiga.
Therapeutic decision-making in MS is still based on the integration of the same demographic, clinical and paraclinical variables as in previous years, e.g. MRI images and the presence of oligoclonal bands. "There are still many challenges in this field, and improvements are coming from different lines of convergence, such as the integration of datasets that can improve the personalization and predictive ability of AI algorithms in healthcare," says ITEN group leader Jose María Prieto.
