Eigenbin compression for reducing photon-counting CT data size.

Area of Science:

• Medical Imaging
• Computed Tomography
• Data Compression

Background:

• Photon-counting CT (PCCT) systems offer advanced image quality through multi-spectral data acquisition.
• High data volume from PCCT poses challenges for gantry slip ring transfer.

Purpose of the Study:

• To develop and evaluate a lossy compression method (eigenbin) for PCCT data using eigenvector analysis.
• To enable rapid initial reconstructions for applications like anatomical verification and automated analysis.
• To assess the method's performance on a clinical silicon PCCT prototype.

Main Methods:

• Principal Component Analysis (PCA) was applied to PCCT calibration measurements to identify principal components (eigenvectors).
• Data dimensionality was reduced by retaining M eigenvectors with the highest eigenvalues, generating 'eigenbin' values.
• Both pixel-specific and pixel-general eigenbin methods were evaluated on phantom data, comparing reconstructed images (basis and VMI) with original data.

Main Results:

• The pixel-specific eigenbin method achieved a 4x data reduction with <5% change in mean values and <12% noise increase.
• The pixel-general method yielded a 2.67x data reduction with <5% change in mean values and <10% noise penalty.
• Virtual monoenergetic images (VMIs) showed less noise penalty and errors compared to basis images, with <5% change in Contrast-to-Noise Ratio (CNR).

Conclusions:

• The eigenbin compression method effectively reduces PCCT data size (2-4x) while preserving essential image information.
• The method demonstrates acceptable noise penalties (<10-20%) and minimal impact on CNR (<5% change in VMIs).
• Eigenbin compression is suitable for applications requiring rapid PCCT reconstructions, reducing transfer time and storage needs.