Applying Internet of Things and machine-learning for personalized healthcare: issues and challenges
Date
2018
Authors
Ahamed, F.
Farid, F.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Conference paper
Citation
2018 International Conference on Machine Learning and Data Engineering (iCMLDE), 2018, pp.19-21
Statement of Responsibility
Conference Name
2018 International Conference on Machine Learning and Data Engineering (iCMLDE) (3 Dec 2018 - 7 Dec 2018 : Sydney, Australia)
Abstract
Personalized Healthcare (PH) is a new patient-oriented healthcare approach which expects to improve the traditional healthcare system. The focus of this new advancement is the patient data collected from patient Electronic health records (EHR), Internet of Things (IoT) sensor devices, wearables and mobile devices, web-based information and social media. PH applies Artificial Intelligence (AI) techniques to the collected dataset to improve disease progression technique, disease prediction, patient self-management and clinical intervention. Machine learning techniques are widely used in this regard to develop analytic models. These models are integrated into different healthcare service applications and clinical decision support systems. These models mainly analyse the collected data from sensor devices and other sources to identify behavioral patterns and clinical conditions of the patient.
For example, these models analyse the collected data to identify the patient's improvements, habits and anomaly in daily routine, changes in sleeping and mobility, eating, drinking and digestive pattern. Based on those patterns the healthcare applications and the clinical decision support systems recommend lifestyle advice, special treatment and care plans for the patient. The doctors and caregivers can also be engaged in the care plan process to validate lifestyle advice. However, there are many uncertainties and a grey area when it comes to applying machine learning in this context. Clinical, behaviour and lifestyle data in nature are very sensitive. There could be different types of biased involved in the process of data collection and interpretation. The training data model could have an older version of the dataset. All these could lead to an incorrect decision from the system without the user's knowledge. In this paper, some of the standards of the ML models reported in the recent research trends, identify the reliability issues and propose improvements.
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright 2019 IEEE