Leveraging Big Data to Provide a Web Service That Provides the Likelihood of Developing Psychological Conditions after a Concussion

• Main Author: Dabek, Filip
• Format: Conference Proceeding
• Language: English
• Subjects: Big data; Biological neural networks; concussion; Hadoop; machine learning; Medical services; Predictive models; preventive healthcare; Psychology; scalability; web service; Web services
• ISSN: 2329-6453
• DOI: 10.1109/MobServ.2016.32

Preventive care attempts to provide an early diagnosis by seeking to inform the patient and their physician of potential complications and diagnoses to expect. With the recent increase in big data present in the healthcare domain, from medical organizations modernizing their operations through electronic health records (EHRs) and deploying new health information technology (HIT) systems, there exists an opportunity for applying predictive models on this vast amount of EHR data and developing a web service to increase the potential for preventive care to be applied. In this paper we present our web service that uses previous work on predicting psychological conditions in concussion patients to provide doctors and patients with the ability to identify the likelihood of said developing psychological conditions. The existing predictive work that we base our web service on utilizes a neural network model which is very slow and inefficient by nature. Due to the inherent costly computation associated with this neural network model we ensure that our web service is fast, efficient, and scalable to meet the demands of prospective users. We provide users with both an API and a frontend interface for predicting psychological conditions 60, 90, 180, 365, and 730 days post concussion, and we utilize various technologies including Hadoop, Pybrain, and Django to form the architecture of our fast, scalable, and efficient prediction web service.