<p>Natural gamma-ray spectroscopy logging quantitatively measures K, U, and Th contents for lithology identification in unconventional reservoirs. To address large errors in LWD element yield determination, this paper proposes a high-precision gamma-ray spectrum analysis method based on the detector response matrix and maximum likelihood estimation. A 256 × 256 CeBr₃ detector response matrix is established using Monte Carlo simulation, and spectra are reconstructed using the Boosted-Gold algorithm to suppress statistical fluctuations prior to MLE-based elemental quantification. To suppress statistical fluctuations and stabilize spectral inversion under low-count conditions, the Boosted-Gold unfolding algorithm was applied prior to elemental quantification based on maximum likelihood estimation (MLE).Simulated tests under various conditions show significantly improved accuracy compared with traditional methods, demonstrating strong potential for low-count natural gamma spectral analysis. Compared with the traditional weighted least squares (WLS) method, the mean relative errors of K, U and Th determined by this method are reduced from 65.6%, 8.4% and 20.0% to 31.5%, 3.2% and 14.1%, respectively, with the accuracy of U and K analysis improved by more than 60%.</p>