Preventive and curative health profiling based on augmented Event-Related Potentials (ERP) and machine learning

Abstract

There are two profiles regarding a healthy lifestyle, which are preventive and curative. Preventive people that avoid potential health problem or treat them early have lower medical costs compared to curative people. According to the report through the Ministry of Health in 2015, the Government of Malaysia has spent 98% of the total amount of RM 11870 million hospital expenditure for curative care. Existing techniques using self-assessment and reported test have been conducted to assess health lifestyle profile, but these techniques susceptible to produce a response bias that leads to misclassification. Another alternative method emerges based on the knowledge that the different individual profile is influenced by the perception formed by each individual, which comes from the human brain. EEG can measure brainwave activity. However, in its raw form, EEG is very difficult to assess the highly specific neural processes that are the focus of cognitive neuroscience. Thus, another method by averaging of the raw signal, Event-Related Potentials (ERPs) became the primary tool of the cognitive neuroscientist and make the technique ideal for studying perception and attention. This research captured the brain activities using electroencephalography (EEG) during receiving healthy and unhealthy food images which act as a stimulus associated with health. These EEG signals converted mathematically into the ERP signals and fed into the classification interface as input. This research aims to identify and classify individual profile, namely preventive and curative using ERP brain signals. In term of classification, the methodology used was the dynamic evolving Spiking Neural Network (deSSN) based in the Neucube architecture. The conclusions from the finding confirm a strong association of perceptions of food images and health profiles are clearly expressed. The results of the ERP analysis shown the mean amplitudes of P300 and LPP components in Parietal and Occipital lobe were higher for healthy foods in the preventive groups. Whereas within curative groups it has been shown the higher for unhealthy foods. These results are suspected to reflect their preferences in choosing food in their daily lifestyle. However, the classification results have shown that unhealthy food stimuli in LPP wave show superior results compared to data analysis in other conditions. Therefore, this study proven the proposed method to do profiling for preventive and curative by using ERP data work on Neucube framework. The classification with ERP data is believed to support the results of the self-assessment and build a more accurate and reliable profiling method. It is hoped that the research findings can lead people more towards a healthy lifestyle.