Prediction of Children's Math Learning Disability Using Longitudinal Brain Data and Machine Learning

Descripción del proyecto

Análisis de la actividad encefálica en niños para detectar dificultades de aprendizaje de las matemáticas

La detección de dificultades de aprendizaje de las matemáticas (DAM) no es tarea fácil. No obstante, es fundamental para poder proporcionar un apoyo adecuado a los niños. Los métodos computacionales actuales para predecir las DAM son limitados. A fin de abordar esta laguna, el proyecto MathDevBML, financiado con fondos europeos, creará un modelo computacional destinado predecir las DAM antes de que los niños empiecen la primaria. En concreto, combinará datos encefálicos de preescolares con técnicas de aprendizaje automático de última generación. En experimentos con imágenes por resonancia magnética, se mostrarán estímulos visuales compuestos por series de puntos a preescolares de cinco años de edad. Se volverá a analizar su actividad encefálica a los siete años de edad. Se aplicarán distintos algoritmos a los datos encefálicos a la edad de cinco años para predecir la manifestación de DAM a la edad de siete años.

Objetivo

Mathematics is the fundamental basis of modern science and technology. However, individuals differ in mathematical ability, and 5%–7% of the population suffers from a math learning disability (MLD). To provide appropriate support for children with MLD, detecting MLD before entering the formal education system is essential. Previous studies have identified some of the neural correlates of MLD; however, computational approaches to predict MLD have been limited. Also, most studies recruited children who were enrolled in elementary school, which is problematic because negative math experience may worsen the difficulties. This research project aims to address these gaps. By combining brain data of preschoolers with state-of-the-art machine learning techniques, I will construct a computational model aiming at predicting MLD before children enter elementary school. The host laboratory of Dr. Jérôme Prado is currently conducting magnetic resonance imaging (MRI) experiments in 5-year-old preschoolers. Participants are presented with visual stimuli consisting of dot patterns, and their brain activity is measured using functional MRI. I will repeat the same MRI task two years later (when children are 7). The math skills of participants will be measured at the age of 7. Multiple algorithms (model-based and model-free approaches) will be applied to the brain data at the age of 5 to predict the occurrence of MLD at the age of 7. Computational models will be applied to other cognitive abilities (language, reasoning), and the influence on atypical math development will be examined. I will benefit from the strong administrative support and advanced neuroimaging resources at the Lyon Neuroscience Research Center, where I will receive training in technical and leadership skills. This research project is an excellent opportunity for me and the host to contribute to the growth of an innovative research field combining developmental neuroscience and machine learning techniques.