Improving Prognostic Prediction from Mild Cognitive Impairment to Alzheimer’s Disease Using Genetic Algorithms

Alzheimer’s disease is becoming a global epidemic. Its impact is devastating for patients’, their families and the economy. As such, it is important to build good prognostic models that can predict conversion to dementia so that treatment measures could be taken. In this work, we applied a genetic algorithm to choose the most relevant neuropsychological and demographic features for prognostic prediction. The results show improvements over other feature selection methods, with our model being able to predict conversion to dementia with AUC and sensitivity of 88% . Moreover, we found that with only 7 features it is possible to achieve good classification results. These results could help physicians to adjust treatment and select which exams should be performed regularly to increase efficiency in clinical practice.