Standard-based personalized healthcare delivery for kidney illness using deep learning

Area of Science:

• Medical Informatics
• Machine Learning
• Artificial Intelligence

Background:

• Predictive analytics in healthcare uses patient data for improved services and treatment planning.
• Deep learning, a subset of machine learning, utilizes deep artificial neural networks for pattern recognition, significantly improving predictive accuracy over traditional models.

Purpose of the Study:

• To analyze the impact of deep learning on a standardized Electronic Health Records dataset for diagnosing kidney-related diseases.
• To evaluate a novel deep learning architecture for its effectiveness in a real-world healthcare scenario.

Main Methods:

• Employed a modularized deep learning architecture, Encoder-Combiner-Decoder (ECD), known for its robust framework and adaptability.
• Trained the ECD model using the openEHR Benchmark Dataset (ORBDA), a real-world dataset from the Brazilian Public Health System (DATASUS).

Main Results:

• The deep learning model demonstrated high precision, recall, and F1 scores in identifying kidney-related diseases from the ORBDA dataset.
• The model's performance indicates a strong capability for accurate and consistent diagnosis of kidney conditions.

Conclusions:

• The developed deep learning model offers a novel approach to analyzing standardized healthcare data for disease diagnosis.
• This model shows potential for deployment in medical institutions, enabling medical professionals to assess its clinical utility and performance.