Effective processing pipeline PACE 2.0 for enhancing chest x-ray contrast and diagnostic interpretability.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Digital medical image analysis requires effective preprocessing to address artifacts like low contrast and intensity inhomogeneities.
• Previous methods, such as the PACE procedure, improved chest X-ray (CXR) images for COVID-19 pneumonia detection but struggled with oversaturated areas in specific conditions.

Purpose of the Study:

• To develop an improved image processing technique, PACE2.0, that enhances the quality of chest X-ray images, particularly in challenging cases.
• To evaluate the performance of PACE2.0 in improving image quality metrics and its impact on pneumonia classification accuracy.

Main Methods:

• PACE2.0 integrates 2D image decomposition, non-local means denoising, gamma correction, and recursive algorithms for image enhancement.
• The enhanced images were quantitatively assessed using contrast improvement index (CII), information entropy (ENT), effective measure of enhancement (EME), and BRISQUE.
• A pre-trained DenseNet-121 model was utilized for transfer learning to evaluate the impact of enhanced images on pneumonia classification.

Main Results:

• PACE2.0 demonstrated significant improvements across image quality metrics, with average increases of 35% in CII, 7.5% in ENT, and 95.6% in EME.
• The system achieved substantial reductions in artifacts, with a 13% improvement in BRISQUE score.
• Pneumonia classification accuracy increased from 80% to 94%, and recall improved from 89% to 97% when using PACE2.0-enhanced images.

Conclusions:

• PACE2.0 significantly enhances medical image quality, particularly for chest X-rays, addressing limitations of previous methods.
• The improved image quality directly translates to better performance in AI-driven pneumonia detection, increasing diagnostic accuracy and recall.
• PACE2.0 shows potential as a valuable clinical decision support tool for healthcare professionals in accurately detecting pneumonia.