Deep-Precognitive Diagnosis: Preventing Future Pandemics by Novel Disease Detection With Biologically-Inspired...

Area of Science:

• Artificial Intelligence in Medicine
• Medical Image Analysis
• Infectious Disease Surveillance

Background:

• Supervised deep learning models for Computer-Aided Diagnosis (CAD) struggle with novel diseases not present in training data.
• Early detection of emerging infectious diseases is critical for controlling outbreaks and preventing pandemics.
• Conventional CAD model development requires extensive data post-outbreak, hindering rapid response.

Purpose of the Study:

• To propose a novel class of CAD models, termed Deep-Precognitive Diagnosis, capable of identifying unknown diseases.
• To develop a biologically-inspired Conv-Fuzzy network for real-time classification of novel diseases.
• To address the limitations of supervised learning with scarce data for emerging pathogens.

Main Methods:

• Development of a novel, biologically-inspired Conv-Fuzzy network architecture.
• Training the model on Chest X-Ray (CXR) scans for normal and bacterial pneumonia classification.
• Testing the model's ability to detect unseen novel diseases, including COVID-19, SARS-CoV-1, and MERS-CoV.

Main Results:

• The Deep-Precognitive Diagnosis model successfully identified COVID-19 as a novel disease in CXR scans not included in its training.
• The model achieved state-of-the-art accuracy when tested on unseen SARS-CoV-1 and MERS-CoV samples.
• The proposed model dynamically creates new classes for novel diseases in real-time, eliminating the need for re-training.

Conclusions:

• Deep-Precognitive Diagnosis offers a groundbreaking solution for early detection of pandemic-potential diseases.
• The Conv-Fuzzy network effectively handles novel disease identification and classification with limited labeled data.
• This approach significantly enhances preparedness and response capabilities for future infectious disease outbreaks.