This invention discloses a
Markov chain Monte Carlo sampling method based on a dual-constraint set projection mechanism, which solves the problems of poor
target distribution adaptability,
slow convergence, and sensitivity to initial values in traditional MCMC sampling in personalized
drug delivery scenarios. The method includes: sampling initialization, using the ppk+MAPB
algorithm to construct a four-layer nested target probability
distribution model; determining a multivariate normal distribution as the proposed distribution; introducing a dual-constraint mechanism of
Euclidean distance constraint set and orientation constraint set during iterative sampling, mapping candidate samples to the intersection of the two sets through a projection operator, and sampling numerical parameters θ and
covariance matrix Σ in stages using the MH criterion; and performing convergence judgment. This invention improves the
clinical effectiveness and accuracy of the sampled samples, achieves faster convergence, and can be widely applied to parameter
estimation and protocol optimization scenarios in personalized
drug delivery.