An improved crow search optimization with Bi-LSTM model for identification and classification of COVID-19 infection from chest X-Ray images

Abstract

Deep learning has become an effective detection method as coronavirus disease 2019 (COVID-19) incidences are increasing quickly. Nevertheless, finding the best accurate models for describing COVID-19 patients is difficult since comparing the outcomes of different data kinds and collecting procedures is difficult. X-ray scans from patients with verified COVID-19 illness and healthy people were combined to produce a dataset. The dataset's noisy, duplicated, and inappropriate characteristics harm the effectiveness of classification algorithms utilized in deep learning. As a result, the data should be pre-processed, and essential parts should be chosen to minimize the dimension of datasets by choosing the most crucial qualities while improving classification accuracy. Speed Up-Robust Feature (SURF) method is applied for the feature extraction from the dataset. The required features are selected after feature extraction based on Crow Search Optimization (CSO). The Bi-LSTM mechanism completes the classification procedure at the final stage. Quality metrics like accuracy, recall, precision, and F1-score are contrasted with some other cutting-edge classifiers. The findings demonstrate the effectiveness of the suggested paradigm for COVID-19 identification and classification, demonstrating high specificity, high sensitivity, and low computing complexity.

Keywords

• COVID-19
• Image classification
• X-ray images
• Crow Search Optimization (CSO)
• Bi-directional Long short-term memory (Bi-LSTM)