COVID-19 diagnosis using chest CT scans and deep convolutional neural networks evolved by IP-based sine-cosine...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Accurate COVID-19 diagnosis is crucial, especially when polymerase chain reaction (PCR) tests are negative but respiratory symptoms persist.
• Computed tomography (CT) scans of the lungs are recommended in such cases for timely diagnosis.
• Deep learning models, particularly deep convolution neural networks (DCNNs), show promise for analyzing medical images like CT scans.

Purpose of the Study:

• To optimize a DCNN structure for improved accuracy in diagnosing COVID-19 from lung CT images.
• To enhance the efficiency and speed of DCNN models in COVID-19 detection.
• To develop a novel optimization algorithm for DCNNs tailored for medical image analysis.

Main Methods:

• The study proposes an optimized DCNN structure using the sine-cosine algorithm (SCA).
• Three key improvements were introduced to the standard SCA: an internet protocol (IP) address-based encoding approach and an enfeebled layer for variable-length DCNN generation.
• The proposed DCNN-IPSCA model was evaluated on the COVID-CT and SARS-CoV-2 datasets and compared against standard DCNN and other variable-length models.

Main Results:

• The DCNN-IPSCA model achieved high diagnostic accuracy, with 98.32% on the SARS-CoV-2 dataset and 98.01% on the COVID-CT dataset.
• Excellent sensitivity (97.22% and 96.23%) and specificity (96.77% and 96.44%) were recorded on the respective datasets.
• The DCNN-IPSCA demonstrated significantly faster training times compared to standard DCNN, being 387.69 times faster on GPU and 63.10 times faster on CPU.

Conclusions:

• The proposed DCNN-IPSCA model offers a highly accurate and efficient method for COVID-19 diagnosis using lung CT images.
• The novel optimization techniques improve both the performance and computational speed of DCNNs for medical image analysis.
• This approach provides a valuable tool for rapid and reliable COVID-19 detection, particularly in challenging diagnostic scenarios.