DeSa COVID-19: Deep salient COVID-19 image-based quality assessment

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Image Quality Assessment

Background:

• Accurate assessment of COVID-19 medical images is crucial for diagnosis and treatment.
• Existing full-reference image quality assessment (FR-IQA) methods may not fully capture the nuances of medical image quality.
• Deep convolutional neural networks (DCNNs) show promise in image analysis tasks.

Purpose of the Study:

• To develop an advanced method for evaluating COVID-19 image quality.
• To incorporate a salient COVID-19 image map with a DCNN for enhanced FR-IQA.
• To compare the proposed method's performance against existing FR-IQA techniques.

Main Methods:

• Developed DeSa COVID-19, a DCNN model incorporating a salient COVID-19 image map.
• Employed the n-convex method for full-reference image quality assessment (FR-IQA).
• Evaluated performance on COVID-chestxray and COVID-CT datasets, including distorted and undistorted images.

Main Results:

• DeSa COVID-19 demonstrated remarkable accomplishment on COVID-chestxray and COVID-CT datasets.
• The proposed DCNN and DeSa COVID-19 methods outperformed other advanced FR-IQA techniques.
• DeSa COVID-19 achieved superior performance in Spearman's rank order correlation coefficient (SROCC) and linear correlation coefficient (LCC) measures, outperforming the recommended DCNN and other advanced FR-MIQA methods by significant margins.
• Reduced computational complexity through optimized function operations.

Conclusions:

• The DeSa COVID-19 method offers a superior approach to COVID-19 image quality assessment.
• The DCNN architecture provides enhanced performance in medical image quality evaluation.
• The study highlights the potential of advanced AI models in improving diagnostic imaging quality.