The Dean of the Polytechnic School and Director of the degree in Computer Engineering, of the European University of the Atlantic (UNEATLANTICO) Manuel Masías, participates in a study that proposes a new system that ensures early treatment of type 2 diabetes, is called DiabSense.
Type 2 diabetes, also known as non-insulin dependent diabetes mellitus (NIDDM), represents one of the biggest health crises of the 21st century. With more than 500 million people affected in 2021, this figure is projected to reach 12.2% of the world’s population by 2045. NIDDM accounts for more than 90% of diabetes cases globally. This chronic disease, characterized by high blood sugar levels, can lead to severe complications such as blindness if not properly treated. Alarmingly, approximately 45% of diabetic patients live without a diagnosis, delaying treatment and exacerbating complications.
In response to this growing threat, this study has developed an innovative system called DiabSense, a system that uses smartphone-based human activity recognition technology and diabetic retinopathy analysis with graphical neural networks. DiabSense combines both graphical neural networks: the Graph Attention Network (GAT) for human activity recognition and the Convolutional Graph Network (CGN) for diabetic retinopathy analysis. The system uses a wide range of 23 human activities that resemble diabetic symptoms for human activity recognition. In addition, it analyzes retinal images of patients to detect the presence of diabetic retinopathy, a common diabetic complication.
The system was tested on four experimental subjects, generating reports of diabetic retinopathy and assessing daily activities over a 30-day period. The GAT achieved 98.32% accuracy in detecting human activities from sensor data, outperforming other state-of-the-art models. For its part, CGN achieved an accuracy of 84.48% in retinal image analysis for diabetic retinopathy report generation.
Once the results of the two graphical neural networks were obtained, the daily activities of diabetic patients were compared with those of the experimental subjects. This made it possible to identify risk factors and recommend early diagnosis, even in the absence of apparent symptoms. The results obtained with DiabSense were compared with clinical diagnostic reports using the A1C test, confirming the accuracy of the system in the early diagnosis of diabetes.
The development of this system marks a milestone in the use of technology to address global health issues such as diabetes. The combination of both graphical neural networks makes it possible to identify the disease in its early stages. With its accuracy and efficiency, it has the potential to improve the lives of millions of people worldwide by ensuring early treatment of the disease.
To learn more about this study, click here.
To read more research, consult the UNEATLANTICO repository.