Joint Image Processing with Learning-Driven Data Representation and Model Behavior for Non-Intrusive Anemia Diagnosis...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Pediatric Diagnostics

Background:

• Anemia diagnosis in children is vital for development but traditional blood tests present challenges like discomfort and infection risk.
• There is a critical need for non-intrusive diagnostic methods to overcome the limitations of current invasive techniques.
• Image processing combined with advanced machine learning offers a potential solution for non-invasive pediatric anemia detection.

Purpose of the Study:

• To develop and validate a novel, non-intrusive method for diagnosing anemia in pediatric patients using image analysis.
• To introduce a new deep network, the recurrent expansion network (RexNet), for enhanced anemia classification.
• To compare the performance of RexNet against traditional methods and existing deep learning models.

Main Methods:

• An image-processing pipeline extracted 181 features from pediatric images, followed by feature selection.
• A deep multilayered long short-term memory (LSTM) network was trained using Bayesian hyperparameter optimization.
• The LSTM model was integrated into a recurrent expansion rules-based model to create RexNet, applied to conjunctival, palmar, and fingernail images.

Main Results:

• RexNet achieved a high overall evaluation of 99.83 ± 0.02% across all classification metrics.
• The proposed method demonstrated significant improvements in diagnostic accuracy and generalization compared to LSTM networks and other existing methods.
• The study successfully applied the non-intrusive approach to conjunctival, palmar, and fingernail images from children up to 6 years old.

Conclusions:

• The developed recurrent expansion network (RexNet) shows high efficacy for non-intrusive anemia diagnosis in pediatric patients.
• RexNet offers a promising, accurate, and less invasive alternative to conventional blood-based anemia testing.
• This innovative approach has the potential to significantly improve the diagnostic process for anemia in young children.