Frequency-domain inverse Monte Carlo simulation for the diagnosis of the early cervical cancer based on NIR diffuse measurement

摘要

This article aims at the optical parameter reconstruction technology for the frequency- domain measurement of near-infrared diffused light. For mimicking the cervix, a cylindrical model with hole in the middle is used in the simulation and experiments. Concerning the structure of the cervix, Monte-Carlo simulation is adopted for describing the photon migration in tissue and Perturbation Monte-Carlo is used for the reconstruction of the optical properties of cervix. The difficulties in the reconstruction of cervical optical properties with frequency domain measurement are the description of the tissue boundary, expression of the frequency-domain signal, and development of rapid reconstruction method for clinical use. To get the frequency domain signal in Monte Carlos simulation, discrete Fourier transformation of the photon migration history in time-domain is employed. By combining the perturbation Monte-Carlo simulation and the LM optimization technology, a rapid reconstruction algorithm is constructed, by which only one Monte-Carlo simulation is needed. The reconstruction method is validated by simulation and experiments on solid phantom. Simulation results show that the inaccuracy in reconstruction of absorption coefficient is less than 3% for a certain range of optical properties. The algorithm is also proved to be robust to the initial guess of optical properties and noise. Experimental results showed that the absorption coefficient can be reconstructed with inaccuracy of less than 10%. The absorption coefficient reconstruction for one set of measurement data can be fulfilled within one minute.