DeepCSFusion: Deep Compressive Sensing Fusion for Efficient COVID-19 Classification.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computational Biology

Background:

• The COVID-19 pandemic has caused millions of deaths globally.
• Deep learning models are crucial for analyzing medical data, but large model sizes pose deployment challenges on resource-constrained devices.

Purpose of the Study:

• To propose a novel deep feature compression strategy for accurate COVID-19 classification from CT scans.
• To develop a model that reduces computational time and storage demands without sacrificing accuracy.

Main Methods:

• A novel compression strategy was developed to compress deep features by 10-90%.
• The DeepCSFusion model was proposed, integrating feature compression and fusion techniques.
• The model was validated on the publicly available "SARS-CoV-2 CT" dataset (1252 scans).

Main Results:

• The DeepCSFusion model achieved an overall classification accuracy of 99.3% for COVID-19 detection.
• The compression strategy significantly reduced computational time and feature requirements.
• The model outperformed existing state-of-the-art methods in various classification metrics.

Conclusions:

• DeepCSFusion offers an effective solution for deploying deep learning models for COVID-19 detection on resource-limited platforms.
• The proposed method demonstrates high accuracy and efficiency in classifying COVID-19 from CT scans.