Abstract
Purpose: A differential analysis with fitted magnetization transfer and lipid signals (DIGITAL) method is proposed for eliminating fat artifacts and accurately extracting amide proton transfer (APT) contrast in lipid‐rich regions. Theory and Methods: The DIGITAL method integrates a two‐pool water‐magnetization transfer (MT) Bloch‐McConnell (BM) model with a six‐peak fat Bloch model. Renormalization is employed to correct fat‐induced Z‐spectrum distortion, and APT # signals are calculated by subtracting the fitted Z‐spectrum from the raw Z‐spectrum. The proposed DIGITAL method was evaluated using numerical simulations, phantoms, and breast tumor data, with its performance compared against prior multi‐pool Lorentzian (MPL) and numerical fitting of extrapolated semisolid magnetization transfer reference signals (NEMR) methods. We also developed a voxel‐wise reliability map to assess data and fit quality. Results: In simulations and experiments, renormalization effectively corrected fat‐induced signal scaling and restored the Z‐spectrum value at 0 ppm to near 0. The proposed DIGITAL method demonstrated superior uniformity in APT # maps compared to MPL and NEMR. Moreover, DIGITAL was minimally affected by fat, exhibiting high robustness against variations in fat concentrations and a strong linear relationship with amide proton concentrations. In addition, the reliability map functioned as an essential quality assurance tool, enabling the identification of outliers and improving the overall precision of APT analysis. Conclusion: The proposed DIGITAL method can effectively eliminate fat artifacts and extract accurate APT signals without requiring sequence modifications, offering a promising solution to the long‐standing challenge of lipid contamination in body APT imaging.
Funding Information
  • National Key Research and Development Program of China (2024YFC2707700)
  • National Key Research and Development Program of China (2023YFE0210300)
  • Fundamental Research Funds for the Central Universities (2025ZFJH01)

This publication has 60 references indexed in Scilit: