Recommender System for the Efficient Treatment of COVID-19 Using a Convolutional Neural Network Model and Image...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Epidemic Management

Background:

• Hospitals struggle with resource allocation during epidemics, particularly with sudden patient influxes like during COVID-19.
• Efficient management of medical resources (doctors, medicines) is critical for patient care during health crises.

Purpose of the Study:

• To design a treatment recommender system (RS) for efficient hospital resource and human capital management during epidemics.
• To leverage deep learning and image search paradigms for an effective RS.

Main Methods:

• Utilized a Convolutional Neural Network (CNN) for feature extraction from chest X-ray images to identify similar patients.
• Employed similarity metrics to compute image similarity for patient matching.
• Developed a methodology to recommend doctors, medicines, and resources based on identified similar patients.

Main Results:

• The proposed RS, using ResNet-50 CNN and Maxwell-Boltzmann similarity, demonstrated high performance.
• Achieved a mean average precision exceeding 0.90 for similarity thresholds between 0.7 and 0.9.
• Attained an average highest cosine similarity of over 0.95, validating the system's efficacy.

Conclusions:

• A recommender system integrating CNN models and image similarity is an effective tool for resource management in pandemics.
• The proposed system can be adopted in clinical settings to optimize hospital operations during peak demand periods.
• This approach enhances the capacity of hospitals to manage patient needs efficiently during widespread outbreaks.