The healthcare sector has undergone significant technological transformation in recent years, with Portugal demonstrating leadership in adopting health tech solutions. However, critical challenges persist, such as limited data integration, inefficiencies in patient care workflows, and variability in treatment outcomes. These issues are exacerbated by the individualized nature of medicine, where treatment efficacy varies due to genetic, demographic, and lifestyle factors.
This article addresses these challenges through the study of an integrated health information system leveraging wearable technology and artificial intelligence (AI). This system aims to centralize patient data, streamline diagnostics, and enable personalized medical care.
Globally, the integration of advanced technologies in healthcare has been met with varied success. Blockchain, Internet of Things (IoT), and AI-based solutions are gaining traction for their ability to enhance data security, interoperability, and analytical capabilities.
Platforms like Gem Healthcare Network and Patientory utilize blockchain for secure patient data management, offering real-time data sharing across stakeholders. Wearable devices such as Fitbit and FDA-approved Apple Watch are transforming health monitoring by providing vital statistics directly to clinicians. These technologies, however, face barriers such as data standardization, interoperability, and regulatory compliance.
Study Details
This study aimed to address inefficiencies in healthcare data management and diagnosis by creating a prototype system capable of:
- Aggregating and normalizing health data from diverse sources, including wearables.
- Providing secure, centralized access to patient health records through blockchain.
- Utilizing AI for predictive diagnostics, focusing on early diabetes detection.
- Establishing interoperability standards for seamless data sharing across healthcare providers.
The ultimate goal is to enable personalized, efficient, and preventive healthcare delivery, reducing costs while improving patient outcomes.
Health data from wearable devices, such as Fitbit and OneTouch, was integrated into a centralized platform. Data normalization processes ensured compatibility across diverse data formats, converting values (e.g., glucose levels) into a standardized structure.
A hybrid blockchain model was developed, combining on-chain and off-chain storage to comply with GDPR requirements. Smart contracts facilitated data access control, enabling patients to manage permissions transparently.
Multi-Layer Perceptron (MLP) Neural Networks were trained on datasets from the MIMIC-III clinical database. Attributes such as age, BMI, glucose levels, and activity patterns were utilized for model training and testing. A predictive accuracy of 82% was achieved, demonstrating the potential for reliable early diabetes detection.
Data was structured using the HL7 FHIR standard, ensuring compatibility across systems and simplifying integration with healthcare providers.
Data from Fitbit and OneTouch devices was captured using APIs and normalized into meaningful health indicators. A scheduling system ensured periodic data synchronization, reducing manual interventions.
The MLP neural network was trained using structured datasets, with preprocessing to eliminate anomalies and optimize feature selection. The model identified patterns indicative of diabetes with high confidence and accuracy.
A dashboard was developed to display patient data, including vital metrics, historical trends, and diagnostic results. Interactive features allowed patients to manage data sharing permissions and schedule medical consultations.
This study sums the potential of integrating advanced technologies into healthcare, offering a solution that aligns with the industry's shift toward precision medicine and patient-centric care.
