[Object Separation from Medical X-Ray Images Based on ICA]

Area of Science:

• Medical Imaging
• Signal Processing
• Biomedical Engineering

Background:

• Traditional X-ray images suffer from noise, limited depth perception, and organ overlap, hindering accurate medical diagnosis.
• Developing advanced imaging techniques is crucial for overcoming these limitations and improving diagnostic accuracy.

Purpose of the Study:

• To propose and evaluate a novel method combining multi-spectrum X-ray imaging and independent component analysis (ICA) for enhanced medical image analysis.
• To improve the separation and reconstruction of target objects in X-ray images, addressing issues of noise and aliasing.

Main Methods:

• Image de-noising preprocessing using sparse code shrinkage to ensure accurate target extraction.
• Application of independent component analysis (ICA) for separating aliasing organs based on pixel-level analysis.
• Reconstruction of target objects using ICA's blind separation theory and convergence matrix.

Main Results:

• Successful separation of target objects was achieved when the number of components in the ICA algorithm exceeded 40.
• Optimal image contrast and minimal distortion were observed when the amplitude scale was within the [25, 45] interval.
• Peak Signal to Noise Ratio (PSNR) analysis indicated that convergence time and amplitude significantly impact image quality.

Conclusions:

• The proposed method effectively enhances X-ray image quality by reducing noise and improving target object separation.
• Achieving optimal results requires careful selection of ICA parameters, specifically convergence times (e.g., 85) and amplitudes (e.g., 35).
• This technique holds significant potential for improving diagnostic accuracy in medical imaging applications.