Modular Learning in Neural Networks: A Modularized Approach to Neural Network Classification
Modular Learning in Neural Networks: A Modularized Approach to Neural Network Classification
Systems Analysis Modelling Simulation - Special issue: Intelligent systems, models and databases for environmental research
IDEAL'07 Proceedings of the 8th international conference on Intelligent data engineering and automated learning
EUROCAST'11 Proceedings of the 13th international conference on Computer Aided Systems Theory - Volume Part II
Hi-index | 0.01 |
In a clinical context, dementia refers to a syndrome of acquired cognitive deterioration that can be associated with various potential stages of the disease. The two most common variations of this disease are Alzheimer type dementia and Vascular type dementia, although there are other forms known as mixed dementia. All of these forms can be associated with different patterns of anatomical affectation, different risk factors, multiple diagnostic characteristics and multiple profiles of neuropsychological tests, making the differential diagnosis of dementias (DDD) very complex. In this paper we propose new diagnostic tools based on a data fusion scheme using artificial neural networks and ensemble systems, which offer important advantages referring to other computational solutions. We have designed two HUMANS based systems, with capacity of processing missing data. We explore their ability for DDD using a battery of cognitive and functional/instrumental neuropsychological tests. We carried out a comparative study between these systems and an clinical expert, reaching these systems a higher level of performance than the expert. Our proposal is a smart and effective complementary method to assist the diagnosis of dementia both in specialized care as well as in primary care centres.