Artificial intelligence based health indicator extraction and disease symptoms identification using medical...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Data Mining

Background:

• Patient health record analysis is crucial for understanding medical needs and disease challenges.
• Existing medical data analysis schemes offer baseline solutions but lack optimal hypothesis-based approaches for clinical decision-making.
• The field of medical electronics requires advanced health indicator extraction models.

Purpose of the Study:

• To propose an optimal hypothesis model for health indicator extraction from electronic medical records (EMR) and International Classification of Diseases (ICD-10) databases.
• To introduce the Akin Method and Friendship method for building hypothesis structures and trait-based feature extraction.
• To enhance the accuracy of medical hypothesis models using deep learning.

Main Methods:

• Developed the Composite Akin Friendship Model (CAFM) using the Akin Method and Friendship method for evidence-based hypothesis generation and information association.
• Implemented a Deep Learning based Akin Friendship Method (DLAFM) integrating Convolutional Neural Networks (CNN) and a Legacy Prediction Model for Health Indicator (LPHI).
• Utilized evidence checking procedures to collect medical evidence from patient examination reports for health indicator and disease symptom extraction (HIDSE).

Main Results:

• The proposed DLAFM model demonstrated an 8-10% improvement in system performance for health indicator extraction compared to existing techniques.
• The CAFM model provided a framework for developing medical hypothesis systems through various test cases.
• The DLAFM approach enhanced the accuracy of the medical hypothesis model, addressing limitations of previous methods.

Conclusions:

• The DLAFM model offers a significant advancement in health indicator extraction from EMR and ICD-10 data.
• This approach aids health sectors in clinical decision-making by providing more accurate health indications.
• The integration of deep learning with Akin-Friendship principles presents a promising direction for medical data analysis and hypothesis generation.